(upbeat music) >> Hello everyone. Welcome to Silicon Angle news breaking story here. Amazon Web Services, expanding their relationship with Hugging Face, breaking news here on Silicon Angle. I'm John Furrier, Silicon Angle reporter, founder and also co-host of theCUBE. And I have with me Swami from Amazon Web Services, vice president of database analytics machine learning with AWS. Swami, great to have you on for this breaking news segment on AWS's big news. Thanks for coming on, taking the time. >> Hey John, pleasure to be here. >> We've had many conversations on theCUBE over the years. We've watched Amazon really move fast into the large data modeling. You SageMaker became a very smashing success. Obviously you've been on this for a while, now with Chat GPT, open AI, a lot of buzz going mainstream, takes it from behind the curtain, inside the ropes, if you will, in the industry to a mainstream. And so this is a big moment I think in the industry. I want to get your perspective because your news with Hugging Face, I think is a is another tell sign that we're about to tip over into a new accelerated growth around making AI now application aware application centric, more programmable, more API access. What's the big news about with AWS Hugging Face, you know, what's going on with this announcement? >> Yeah, first of all, they're very excited to announce our expanded collaboration with Hugging Face because with this partnership, our goal, as you all know, I mean Hugging Face I consider them like the GitHub for machine learning. And with this partnership, Hugging Face and AWS will be able to democratize AI for a broad range of developers, not just specific deep AI startups. And now with this we can accelerate the training, fine tuning, and deployment of these large language models and vision models from Hugging Face in the cloud. So, and the broader context, when you step back and see what customer problem we are trying to solve with this announcement, essentially if you see these foundational models are used to now create like a huge number of applications, suggest like tech summarization, question answering, or search image generation, creative, other things. And these are all stuff we are seeing in the likes of these Chat GPT style applications. But there is a broad range of enterprise use cases that we don't even talk about. And it's because these kind of transformative generative AI capabilities and models are not available to, I mean, millions of developers. And because either training these elements from scratch can be very expensive or time consuming and need deep expertise, or more importantly, they don't need these generic models. They need them to be fine tuned for the specific use cases. And one of the biggest complaints we hear is that these models, when they try to use it for real production use cases, they are incredibly expensive to train and incredibly expensive to run inference on, to use it at a production scale, so And unlike search, web search style applications where the margins can be really huge, here in production use cases and enterprises, you want efficiency at scale. That's where a Hugging Face and AWS share our mission. And by integrating with Trainium and Inferentia, we're able to handle the cost efficient training and inference at scale. I'll deep dive on it and by training teaming up on the SageMaker front now the time it takes to build these models and fine tune them as also coming down. So that's what makes this partnership very unique as well. So I'm very excited. >> I want to get into the, to the time savings and the cost savings as well on the on the training and inference. It's a huge issue. But before we get into that, just how long have you guys been working with Hugging Face? I know this is a previous relationship. This is an expansion of that relationship. Can you comment on the what's different about what's happened before and then now? >> Yeah, so Hugging Face, we have had an great relationship in the past few years as well where they have actually made their models available to run on AWS in a fashion, even inspect their Bloom project was something many of our customers even used. Bloom Project for context is their open source project, which builds a GPT three style model. And now with this expanded collaboration, now Hugging Face selected AWS for that next generation of this generative AI model, building on their highly successful Bloom project as well. And the nice thing is now by direct integration with Trainium and Inferentia, where you get cost savings in a really significant way. Now for instance, tier 1 can provide up to 50% cost to train savings, and Inferentia can deliver up to 60% better costs and Forex more higher throughput. Now these models, especially as they train that next generation generated AI model, it is going to be not only more accessible to all the developers who use it in open. So it'll be a lot cheaper as well. And that's what makes this moment really exciting because yeah, we can't democratize AI unless we make it broadly accessible and cost efficient, and easy to program and use as well. >> Okay, thanks Swami. We really appreciate. Swami's a Cube alumni, but also vice President, database analyst machine learning web services breaking down the Hugging Face announcement. Obviously the relationship he called it the GitHub of machine learning. This is the beginning of what we will see, a continuing competitive battle with Microsoft. Microsoft launching OpenAI. Amazon's been doing it for years. They got Alexa, they know what they're doing. It's going to be very interesting to see how this all plays out. You're watching Silicon Angle News, breaking here. I'm John Furrier, host of the Cube. Thanks for watching. (ethereal music)

>> From theCUBE Studios in Palo Alto in Boston, bringing you data-driven insights from theCUBE and ETR. This is Breaking Analysis with Dave Vellante. >> OpenAI The company, and ChatGPT have taken the world by storm. Microsoft reportedly is investing an additional 10 billion dollars into the company. But in our view, while the hype around ChatGPT is justified, we don't believe OpenAI will lock up the market with its first mover advantage. Rather, we believe that success in this market will be directly proportional to the quality and quantity of data that a technology company has at its disposal, and the compute power that it could deploy to run its system. Hello and welcome to this week's Wikibon CUBE insights, powered by ETR. In this Breaking Analysis, we unpack the excitement around ChatGPT, and debate the premise that the company's early entry into the space may not confer winner take all advantage to OpenAI. And to do so, we welcome CUBE collaborator, alum, Sarbjeet Johal, (chuckles) and John Furrier, co-host of the Cube. Great to see you Sarbjeet, John. Really appreciate you guys coming to the program. >> Great to be on. >> Okay, so what is ChatGPT? Well, actually we asked ChatGPT, what is ChatGPT? So here's what it said. ChatGPT is a state-of-the-art language model developed by OpenAI that can generate human-like text. It could be fine tuned for a variety of language tasks, such as conversation, summarization, and language translation. So I asked it, give it to me in 50 words or less. How did it do? Anything to add? >> Yeah, think it did good. It's large language model, like previous models, but it started applying the transformers sort of mechanism to focus on what prompt you have given it to itself. And then also the what answer it gave you in the first, sort of, one sentence or two sentences, and then introspect on itself, like what I have already said to you. And so just work on that. So it it's self sort of focus if you will. It does, the transformers help the large language models to do that. >> So to your point, it's a large language model, and GPT stands for generative pre-trained transformer. >> And if you put the definition back up there again, if you put it back up on the screen, let's see it back up. Okay, it actually missed the large, word large. So one of the problems with ChatGPT, it's not always accurate. It's actually a large language model, and it says state of the art language model. And if you look at Google, Google has dominated AI for many times and they're well known as being the best at this. And apparently Google has their own large language model, LLM, in play and have been holding it back to release because of backlash on the accuracy. Like just in that example you showed is a great point. They got almost right, but they missed the key word. >> You know what's funny about that John, is I had previously asked it in my prompt to give me it in less than a hundred words, and it was too long, I said I was too long for Breaking Analysis, and there it went into the fact that it's a large language model. So it largely, it gave me a really different answer the, for both times. So, but it's still pretty amazing for those of you who haven't played with it yet. And one of the best examples that I saw was Ben Charrington from This Week In ML AI podcast. And I stumbled on this thanks to Brian Gracely, who was listening to one of his Cloudcasts. Basically what Ben did is he took, he prompted ChatGPT to interview ChatGPT, and he simply gave the system the prompts, and then he ran the questions and answers into this avatar builder and sped it up 2X so it didn't sound like a machine. And voila, it was amazing. So John is ChatGPT going to take over as a cube host? >> Well, I was thinking, we get the questions in advance sometimes from PR people. We should actually just plug it in ChatGPT, add it to our notes, and saying, "Is this good enough for you? Let's ask the real question." So I think, you know, I think there's a lot of heavy lifting that gets done. I think the ChatGPT is a phenomenal revolution. I think it highlights the use case. Like that example we showed earlier. It gets most of it right. So it's directionally correct and it feels like it's an answer, but it's not a hundred percent accurate. And I think that's where people are seeing value in it. Writing marketing, copy, brainstorming, guest list, gift list for somebody. Write me some lyrics to a song. Give me a thesis about healthcare policy in the United States. It'll do a bang up job, and then you got to go in and you can massage it. So we're going to do three quarters of the work. That's why plagiarism and schools are kind of freaking out. And that's why Microsoft put 10 billion in, because why wouldn't this be a feature of Word, or the OS to help it do stuff on behalf of the user. So linguistically it's a beautiful thing. You can input a string and get a good answer. It's not a search result. >> And we're going to get your take on on Microsoft and, but it kind of levels the playing- but ChatGPT writes better than I do, Sarbjeet, and I know you have some good examples too. You mentioned the Reed Hastings example. >> Yeah, I was listening to Reed Hastings fireside chat with ChatGPT, and the answers were coming as sort of voice, in the voice format. And it was amazing what, he was having very sort of philosophy kind of talk with the ChatGPT, the longer sentences, like he was going on, like, just like we are talking, he was talking for like almost two minutes and then ChatGPT was answering. It was not one sentence question, and then a lot of answers from ChatGPT and yeah, you're right. I, this is our ability. I've been thinking deep about this since yesterday, we talked about, like, we want to do this segment. The data is fed into the data model. It can be the current data as well, but I think that, like, models like ChatGPT, other companies will have those too. They can, they're democratizing the intelligence, but they're not creating intelligence yet, definitely yet I can say that. They will give you all the finite answers. Like, okay, how do you do this for loop in Java, versus, you know, C sharp, and as a programmer you can do that, in, but they can't tell you that, how to write a new algorithm or write a new search algorithm for you. They cannot create a secretive code for you to- >> Not yet. >> Have competitive advantage. >> Not yet, not yet. >> but you- >> Can Google do that today? >> No one really can. The reasoning side of the data is, we talked about at our Supercloud event, with Zhamak Dehghani who's was CEO of, now of Nextdata. This next wave of data intelligence is going to come from entrepreneurs that are probably cross discipline, computer science and some other discipline. But they're going to be new things, for example, data, metadata, and data. It's hard to do reasoning like a human being, so that needs more data to train itself. So I think the first gen of this training module for the large language model they have is a corpus of text. Lot of that's why blog posts are, but the facts are wrong and sometimes out of context, because that contextual reasoning takes time, it takes intelligence. So machines need to become intelligent, and so therefore they need to be trained. So you're going to start to see, I think, a lot of acceleration on training the data sets. And again, it's only as good as the data you can get. And again, proprietary data sets will be a huge winner. Anyone who's got a large corpus of content, proprietary content like theCUBE or SiliconANGLE as a publisher will benefit from this. Large FinTech companies, anyone with large proprietary data will probably be a big winner on this generative AI wave, because it just, it will eat that up, and turn that back into something better. So I think there's going to be a lot of interesting things to look at here. And certainly productivity's going to be off the charts for vanilla and the internet is going to get swarmed with vanilla content. So if you're in the content business, and you're an original content producer of any kind, you're going to be not vanilla, so you're going to be better. So I think there's so much at play Dave (indistinct). >> I think the playing field has been risen, so we- >> Risen and leveled? >> Yeah, and leveled to certain extent. So it's now like that few people as consumers, as consumers of AI, we will have a advantage and others cannot have that advantage. So it will be democratized. That's, I'm sure about that. But if you take the example of calculator, when the calculator came in, and a lot of people are, "Oh, people can't do math anymore because calculator is there." right? So it's a similar sort of moment, just like a calculator for the next level. But, again- >> I see it more like open source, Sarbjeet, because like if you think about what ChatGPT's doing, you do a query and it comes from somewhere the value of a post from ChatGPT is just a reuse of AI. The original content accent will be come from a human. So if I lay out a paragraph from ChatGPT, did some heavy lifting on some facts, I check the facts, save me about maybe- >> Yeah, it's productive. >> An hour writing, and then I write a killer two, three sentences of, like, sharp original thinking or critical analysis. I then took that body of work, open source content, and then laid something on top of it. >> And Sarbjeet's example is a good one, because like if the calculator kids don't do math as well anymore, the slide rule, remember we had slide rules as kids, remember we first started using Waze, you know, we were this minority and you had an advantage over other drivers. Now Waze is like, you know, social traffic, you know, navigation, everybody had, you know- >> All the back roads are crowded. >> They're car crowded. (group laughs) Exactly. All right, let's, let's move on. What about this notion that futurist Ray Amara put forth and really Amara's Law that we're showing here, it's, the law is we, you know, "We tend to overestimate the effect of technology in the short run and underestimate it in the long run." Is that the case, do you think, with ChatGPT? What do you think Sarbjeet? >> I think that's true actually. There's a lot of, >> We don't debate this. >> There's a lot of awe, like when people see the results from ChatGPT, they say what, what the heck? Like, it can do this? But then if you use it more and more and more, and I ask the set of similar question, not the same question, and it gives you like same answer. It's like reading from the same bucket of text in, the interior read (indistinct) where the ChatGPT, you will see that in some couple of segments. It's very, it sounds so boring that the ChatGPT is coming out the same two sentences every time. So it is kind of good, but it's not as good as people think it is right now. But we will have, go through this, you know, hype sort of cycle and get realistic with it. And then in the long term, I think it's a great thing in the short term, it's not something which will (indistinct) >> What's your counter point? You're saying it's not. >> I, no I think the question was, it's hyped up in the short term and not it's underestimated long term. That's what I think what he said, quote. >> Yes, yeah. That's what he said. >> Okay, I think that's wrong with this, because this is a unique, ChatGPT is a unique kind of impact and it's very generational. People have been comparing it, I have been comparing to the internet, like the web, web browser Mosaic and Netscape, right, Navigator. I mean, I clearly still remember the days seeing Navigator for the first time, wow. And there weren't not many sites you could go to, everyone typed in, you know, cars.com, you know. >> That (indistinct) wasn't that overestimated, the overhyped at the beginning and underestimated. >> No, it was, it was underestimated long run, people thought. >> But that Amara's law. >> That's what is. >> No, they said overestimated? >> Overestimated near term underestimated- overhyped near term, underestimated long term. I got, right I mean? >> Well, I, yeah okay, so I would then agree, okay then- >> We were off the charts about the internet in the early days, and it actually exceeded our expectations. >> Well there were people who were, like, poo-pooing it early on. So when the browser came out, people were like, "Oh, the web's a toy for kids." I mean, in 1995 the web was a joke, right? So '96, you had online populations growing, so you had structural changes going on around the browser, internet population. And then that replaced other things, direct mail, other business activities that were once analog then went to the web, kind of read only as you, as we always talk about. So I think that's a moment where the hype long term, the smart money, and the smart industry experts all get the long term. And in this case, there's more poo-pooing in the short term. "Ah, it's not a big deal, it's just AI." I've heard many people poo-pooing ChatGPT, and a lot of smart people saying, "No this is next gen, this is different and it's only going to get better." So I think people are estimating a big long game on this one. >> So you're saying it's bifurcated. There's those who say- >> Yes. >> Okay, all right, let's get to the heart of the premise, and possibly the debate for today's episode. Will OpenAI's early entry into the market confer sustainable competitive advantage for the company. And if you look at the history of tech, the technology industry, it's kind of littered with first mover failures. Altair, IBM, Tandy, Commodore, they and Apple even, they were really early in the PC game. They took a backseat to Dell who came in the scene years later with a better business model. Netscape, you were just talking about, was all the rage in Silicon Valley, with the first browser, drove up all the housing prices out here. AltaVista was the first search engine to really, you know, index full text. >> Owned by Dell, I mean DEC. >> Owned by Digital. >> Yeah, Digital Equipment >> Compaq bought it. And of course as an aside, Digital, they wanted to showcase their hardware, right? Their super computer stuff. And then so Friendster and MySpace, they came before Facebook. The iPhone certainly wasn't the first mobile device. So lots of failed examples, but there are some recent successes like AWS and cloud. >> You could say smartphone. So I mean. >> Well I know, and you can, we can parse this so we'll debate it. Now Twitter, you could argue, had first mover advantage. You kind of gave me that one John. Bitcoin and crypto clearly had first mover advantage, and sustaining that. Guys, will OpenAI make it to the list on the right with ChatGPT, what do you think? >> I think categorically as a company, it probably won't, but as a category, I think what they're doing will, so OpenAI as a company, they get funding, there's power dynamics involved. Microsoft put a billion dollars in early on, then they just pony it up. Now they're reporting 10 billion more. So, like, if the browsers, Microsoft had competitive advantage over Netscape, and used monopoly power, and convicted by the Department of Justice for killing Netscape with their monopoly, Netscape should have had won that battle, but Microsoft killed it. In this case, Microsoft's not killing it, they're buying into it. So I think the embrace extend Microsoft power here makes OpenAI vulnerable for that one vendor solution. So the AI as a company might not make the list, but the category of what this is, large language model AI, is probably will be on the right hand side. >> Okay, we're going to come back to the government intervention and maybe do some comparisons, but what are your thoughts on this premise here? That, it will basically set- put forth the premise that it, that ChatGPT, its early entry into the market will not confer competitive advantage to >> For OpenAI. >> To Open- Yeah, do you agree with that? >> I agree with that actually. It, because Google has been at it, and they have been holding back, as John said because of the scrutiny from the Fed, right, so- >> And privacy too. >> And the privacy and the accuracy as well. But I think Sam Altman and the company on those guys, right? They have put this in a hasty way out there, you know, because it makes mistakes, and there are a lot of questions around the, sort of, where the content is coming from. You saw that as your example, it just stole the content, and without your permission, you know? >> Yeah. So as quick this aside- >> And it codes on people's behalf and the, those codes are wrong. So there's a lot of, sort of, false information it's putting out there. So it's a very vulnerable thing to do what Sam Altman- >> So even though it'll get better, others will compete. >> So look, just side note, a term which Reid Hoffman used a little bit. Like he said, it's experimental launch, like, you know, it's- >> It's pretty damn good. >> It is clever because according to Sam- >> It's more than clever. It's good. >> It's awesome, if you haven't used it. I mean you write- you read what it writes and you go, "This thing writes so well, it writes so much better than you." >> The human emotion drives that too. I think that's a big thing. But- >> I Want to add one more- >> Make your last point. >> Last one. Okay. So, but he's still holding back. He's conducting quite a few interviews. If you want to get the gist of it, there's an interview with StrictlyVC interview from yesterday with Sam Altman. Listen to that one it's an eye opening what they want- where they want to take it. But my last one I want to make it on this point is that Satya Nadella yesterday did an interview with Wall Street Journal. I think he was doing- >> You were not impressed. >> I was not impressed because he was pushing it too much. So Sam Altman's holding back so there's less backlash. >> Got 10 billion reasons to push. >> I think he's almost- >> Microsoft just laid off 10000 people. Hey ChatGPT, find me a job. You know like. (group laughs) >> He's overselling it to an extent that I think it will backfire on Microsoft. And he's over promising a lot of stuff right now, I think. I don't know why he's very jittery about all these things. And he did the same thing during Ignite as well. So he said, "Oh, this AI will write code for you and this and that." Like you called him out- >> The hyperbole- >> During your- >> from Satya Nadella, he's got a lot of hyperbole. (group talks over each other) >> All right, Let's, go ahead. >> Well, can I weigh in on the whole- >> Yeah, sure. >> Microsoft thing on whether OpenAI, here's the take on this. I think it's more like the browser moment to me, because I could relate to that experience with ChatG, personally, emotionally, when I saw that, and I remember vividly- >> You mean that aha moment (indistinct). >> Like this is obviously the future. Anything else in the old world is dead, website's going to be everywhere. It was just instant dot connection for me. And a lot of other smart people who saw this. Lot of people by the way, didn't see it. Someone said the web's a toy. At the company I was worked for at the time, Hewlett Packard, they like, they could have been in, they had invented HTML, and so like all this stuff was, like, they just passed, the web was just being passed over. But at that time, the browser got better, more websites came on board. So the structural advantage there was online web usage was growing, online user population. So that was growing exponentially with the rise of the Netscape browser. So OpenAI could stay on the right side of your list as durable, if they leverage the category that they're creating, can get the scale. And if they can get the scale, just like Twitter, that failed so many times that they still hung around. So it was a product that was always successful, right? So I mean, it should have- >> You're right, it was terrible, we kept coming back. >> The fail whale, but it still grew. So OpenAI has that moment. They could do it if Microsoft doesn't meddle too much with too much power as a vendor. They could be the Netscape Navigator, without the anti-competitive behavior of somebody else. So to me, they have the pole position. So they have an opportunity. So if not, if they don't execute, then there's opportunity. There's not a lot of barriers to entry, vis-a-vis say the CapEx of say a cloud company like AWS. You can't replicate that, Many have tried, but I think you can replicate OpenAI. >> And we're going to talk about that. Okay, so real quick, I want to bring in some ETR data. This isn't an ETR heavy segment, only because this so new, you know, they haven't coverage yet, but they do cover AI. So basically what we're seeing here is a slide on the vertical axis's net score, which is a measure of spending momentum, and in the horizontal axis's is presence in the dataset. Think of it as, like, market presence. And in the insert right there, you can see how the dots are plotted, the two columns. And so, but the key point here that we want to make, there's a bunch of companies on the left, is he like, you know, DataRobot and C3 AI and some others, but the big whales, Google, AWS, Microsoft, are really dominant in this market. So that's really the key takeaway that, can we- >> I notice IBM is way low. >> Yeah, IBM's low, and actually bring that back up and you, but then you see Oracle who actually is injecting. So I guess that's the other point is, you're not necessarily going to go buy AI, and you know, build your own AI, you're going to, it's going to be there and, it, Salesforce is going to embed it into its platform, the SaaS companies, and you're going to purchase AI. You're not necessarily going to build it. But some companies obviously are. >> I mean to quote IBM's general manager Rob Thomas, "You can't have AI with IA." information architecture and David Flynn- >> You can't Have AI without IA >> without, you can't have AI without IA. You can't have, if you have an Information Architecture, you then can power AI. Yesterday David Flynn, with Hammersmith, was on our Supercloud. He was pointing out that the relationship of storage, where you store things, also impacts the data and stressablity, and Zhamak from Nextdata, she was pointing out that same thing. So the data problem factors into all this too, Dave. >> So you got the big cloud and internet giants, they're all poised to go after this opportunity. Microsoft is investing up to 10 billion. Google's code red, which was, you know, the headline in the New York Times. Of course Apple is there and several alternatives in the market today. Guys like Chinchilla, Bloom, and there's a company Jasper and several others, and then Lena Khan looms large and the government's around the world, EU, US, China, all taking notice before the market really is coalesced around a single player. You know, John, you mentioned Netscape, they kind of really, the US government was way late to that game. It was kind of game over. And Netscape, I remember Barksdale was like, "Eh, we're going to be selling software in the enterprise anyway." and then, pshew, the company just dissipated. So, but it looks like the US government, especially with Lena Khan, they're changing the definition of antitrust and what the cause is to go after people, and they're really much more aggressive. It's only what, two years ago that (indistinct). >> Yeah, the problem I have with the federal oversight is this, they're always like late to the game, and they're slow to catch up. So in other words, they're working on stuff that should have been solved a year and a half, two years ago around some of the social networks hiding behind some of the rules around open web back in the days, and I think- >> But they're like 15 years late to that. >> Yeah, and now they got this new thing on top of it. So like, I just worry about them getting their fingers. >> But there's only two years, you know, OpenAI. >> No, but the thing (indistinct). >> No, they're still fighting other battles. But the problem with government is that they're going to label Big Tech as like a evil thing like Pharma, it's like smoke- >> You know Lena Khan wants to kill Big Tech, there's no question. >> So I think Big Tech is getting a very seriously bad rap. And I think anything that the government does that shades darkness on tech, is politically motivated in most cases. You can almost look at everything, and my 80 20 rule is in play here. 80% of the government activity around tech is bullshit, it's politically motivated, and the 20% is probably relevant, but off the mark and not organized. >> Well market forces have always been the determining factor of success. The governments, you know, have been pretty much failed. I mean you look at IBM's antitrust, that, what did that do? The market ultimately beat them. You look at Microsoft back in the day, right? Windows 95 was peaking, the government came in. But you know, like you said, they missed the web, right, and >> so they were hanging on- >> There's nobody in government >> to Windows. >> that actually knows- >> And so, you, I think you're right. It's market forces that are going to determine this. But Sarbjeet, what do you make of Microsoft's big bet here, you weren't impressed with with Nadella. How do you think, where are they going to apply it? Is this going to be a Hail Mary for Bing, or is it going to be applied elsewhere? What do you think. >> They are saying that they will, sort of, weave this into their products, office products, productivity and also to write code as well, developer productivity as well. That's a big play for them. But coming back to your antitrust sort of comments, right? I believe the, your comment was like, oh, fed was late 10 years or 15 years earlier, but now they're two years. But things are moving very fast now as compared to they used to move. >> So two years is like 10 Years. >> Yeah, two years is like 10 years. Just want to make that point. (Dave laughs) This thing is going like wildfire. Any new tech which comes in that I think they're going against distribution channels. Lina Khan has commented time and again that the marketplace model is that she wants to have some grip on. Cloud marketplaces are a kind of monopolistic kind of way. >> I don't, I don't see this, I don't see a Chat AI. >> You told me it's not Bing, you had an interesting comment. >> No, no. First of all, this is great from Microsoft. If you're Microsoft- >> Why? >> Because Microsoft doesn't have the AI chops that Google has, right? Google is got so much core competency on how they run their search, how they run their backends, their cloud, even though they don't get a lot of cloud market share in the enterprise, they got a kick ass cloud cause they needed one. >> Totally. >> They've invented SRE. I mean Google's development and engineering chops are off the scales, right? Amazon's got some good chops, but Google's got like 10 times more chops than AWS in my opinion. Cloud's a whole different story. Microsoft gets AI, they get a playbook, they get a product they can render into, the not only Bing, productivity software, helping people write papers, PowerPoint, also don't forget the cloud AI can super help. We had this conversation on our Supercloud event, where AI's going to do a lot of the heavy lifting around understanding observability and managing service meshes, to managing microservices, to turning on and off applications, and or maybe writing code in real time. So there's a plethora of use cases for Microsoft to deploy this. combined with their R and D budgets, they can then turbocharge more research, build on it. So I think this gives them a car in the game, Google may have pole position with AI, but this puts Microsoft right in the game, and they already have a lot of stuff going on. But this just, I mean everything gets lifted up. Security, cloud, productivity suite, everything. >> What's under the hood at Google, and why aren't they talking about it? I mean they got to be freaked out about this. No? Or do they have kind of a magic bullet? >> I think they have the, they have the chops definitely. Magic bullet, I don't know where they are, as compared to the ChatGPT 3 or 4 models. Like they, but if you look at the online sort of activity and the videos put out there from Google folks, Google technology folks, that's account you should look at if you are looking there, they have put all these distinctions what ChatGPT 3 has used, they have been talking about for a while as well. So it's not like it's a secret thing that you cannot replicate. As you said earlier, like in the beginning of this segment, that anybody who has more data and the capacity to process that data, which Google has both, I think they will win this. >> Obviously living in Palo Alto where the Google founders are, and Google's headquarters next town over we have- >> We're so close to them. We have inside information on some of the thinking and that hasn't been reported by any outlet yet. And that is, is that, from what I'm hearing from my sources, is Google has it, they don't want to release it for many reasons. One is it might screw up their search monopoly, one, two, they're worried about the accuracy, 'cause Google will get sued. 'Cause a lot of people are jamming on this ChatGPT as, "Oh it does everything for me." when it's clearly not a hundred percent accurate all the time. >> So Lina Kahn is looming, and so Google's like be careful. >> Yeah so Google's just like, this is the third, could be a third rail. >> But the first thing you said is a concern. >> Well no. >> The disruptive (indistinct) >> What they will do is do a Waymo kind of thing, where they spin out a separate company. >> They're doing that. >> The discussions happening, they're going to spin out the separate company and put it over there, and saying, "This is AI, got search over there, don't touch that search, 'cause that's where all the revenue is." (chuckles) >> So, okay, so that's how they deal with the Clay Christensen dilemma. What's the business model here? I mean it's not advertising, right? Is it to charge you for a query? What, how do you make money at this? >> It's a good question, I mean my thinking is, first of all, it's cool to type stuff in and see a paper get written, or write a blog post, or gimme a marketing slogan for this or that or write some code. I think the API side of the business will be critical. And I think Howie Xu, I know you're going to reference some of his comments yesterday on Supercloud, I think this brings a whole 'nother user interface into technology consumption. I think the business model, not yet clear, but it will probably be some sort of either API and developer environment or just a straight up free consumer product, with some sort of freemium backend thing for business. >> And he was saying too, it's natural language is the way in which you're going to interact with these systems. >> I think it's APIs, it's APIs, APIs, APIs, because these people who are cooking up these models, and it takes a lot of compute power to train these and to, for inference as well. Somebody did the analysis on the how many cents a Google search costs to Google, and how many cents the ChatGPT query costs. It's, you know, 100x or something on that. You can take a look at that. >> A 100x on which side? >> You're saying two orders of magnitude more expensive for ChatGPT >> Much more, yeah. >> Than for Google. >> It's very expensive. >> So Google's got the data, they got the infrastructure and they got, you're saying they got the cost (indistinct) >> No actually it's a simple query as well, but they are trying to put together the answers, and they're going through a lot more data versus index data already, you know. >> Let me clarify, you're saying that Google's version of ChatGPT is more efficient? >> No, I'm, I'm saying Google search results. >> Ah, search results. >> What are used to today, but cheaper. >> But that, does that, is that going to confer advantage to Google's large language (indistinct)? >> It will, because there were deep science (indistinct). >> Google, I don't think Google search is doing a large language model on their search, it's keyword search. You know, what's the weather in Santa Cruz? Or how, what's the weather going to be? Or you know, how do I find this? Now they have done a smart job of doing some things with those queries, auto complete, re direct navigation. But it's, it's not entity. It's not like, "Hey, what's Dave Vellante thinking this week in Breaking Analysis?" ChatGPT might get that, because it'll get your Breaking Analysis, it'll synthesize it. There'll be some, maybe some clips. It'll be like, you know, I mean. >> Well I got to tell you, I asked ChatGPT to, like, I said, I'm going to enter a transcript of a discussion I had with Nir Zuk, the CTO of Palo Alto Networks, And I want you to write a 750 word blog. I never input the transcript. It wrote a 750 word blog. It attributed quotes to him, and it just pulled a bunch of stuff that, and said, okay, here it is. It talked about Supercloud, it defined Supercloud. >> It's made, it makes you- >> Wow, But it was a big lie. It was fraudulent, but still, blew me away. >> Again, vanilla content and non accurate content. So we are going to see a surge of misinformation on steroids, but I call it the vanilla content. Wow, that's just so boring, (indistinct). >> There's so many dangers. >> Make your point, cause we got to, almost out of time. >> Okay, so the consumption, like how do you consume this thing. As humans, we are consuming it and we are, like, getting a nicely, like, surprisingly shocked, you know, wow, that's cool. It's going to increase productivity and all that stuff, right? And on the danger side as well, the bad actors can take hold of it and create fake content and we have the fake sort of intelligence, if you go out there. So that's one thing. The second thing is, we are as humans are consuming this as language. Like we read that, we listen to it, whatever format we consume that is, but the ultimate usage of that will be when the machines can take that output from likes of ChatGPT, and do actions based on that. The robots can work, the robot can paint your house, we were talking about, right? Right now we can't do that. >> Data apps. >> So the data has to be ingested by the machines. It has to be digestible by the machines. And the machines cannot digest unorganized data right now, we will get better on the ingestion side as well. So we are getting better. >> Data, reasoning, insights, and action. >> I like that mall, paint my house. >> So, okay- >> By the way, that means drones that'll come in. Spray painting your house. >> Hey, it wasn't too long ago that robots couldn't climb stairs, as I like to point out. Okay, and of course it's no surprise the venture capitalists are lining up to eat at the trough, as I'd like to say. Let's hear, you'd referenced this earlier, John, let's hear what AI expert Howie Xu said at the Supercloud event, about what it takes to clone ChatGPT. Please, play the clip. >> So one of the VCs actually asked me the other day, right? "Hey, how much money do I need to spend, invest to get a, you know, another shot to the openAI sort of the level." You know, I did a (indistinct) >> Line up. >> A hundred million dollar is the order of magnitude that I came up with, right? You know, not a billion, not 10 million, right? So a hundred- >> Guys a hundred million dollars, that's an astoundingly low figure. What do you make of it? >> I was in an interview with, I was interviewing, I think he said hundred million or so, but in the hundreds of millions, not a billion right? >> You were trying to get him up, you were like "Hundreds of millions." >> Well I think, I- >> He's like, eh, not 10, not a billion. >> Well first of all, Howie Xu's an expert machine learning. He's at Zscaler, he's a machine learning AI guy. But he comes from VMware, he's got his technology pedigrees really off the chart. Great friend of theCUBE and kind of like a CUBE analyst for us. And he's smart. He's right. I think the barriers to entry from a dollar standpoint are lower than say the CapEx required to compete with AWS. Clearly, the CapEx spending to build all the tech for the run a cloud. >> And you don't need a huge sales force. >> And in some case apps too, it's the same thing. But I think it's not that hard. >> But am I right about that? You don't need a huge sales force either. It's, what, you know >> If the product's good, it will sell, this is a new era. The better mouse trap will win. This is the new economics in software, right? So- >> Because you look at the amount of money Lacework, and Snyk, Snowflake, Databrooks. Look at the amount of money they've raised. I mean it's like a billion dollars before they get to IPO or more. 'Cause they need promotion, they need go to market. You don't need (indistinct) >> OpenAI's been working on this for multiple five years plus it's, hasn't, wasn't born yesterday. Took a lot of years to get going. And Sam is depositioning all the success, because he's trying to manage expectations, To your point Sarbjeet, earlier. It's like, yeah, he's trying to "Whoa, whoa, settle down everybody, (Dave laughs) it's not that great." because he doesn't want to fall into that, you know, hero and then get taken down, so. >> It may take a 100 million or 150 or 200 million to train the model. But to, for the inference to, yeah to for the inference machine, It will take a lot more, I believe. >> Give it, so imagine, >> Because- >> Go ahead, sorry. >> Go ahead. But because it consumes a lot more compute cycles and it's certain level of storage and everything, right, which they already have. So I think to compute is different. To frame the model is a different cost. But to run the business is different, because I think 100 million can go into just fighting the Fed. >> Well there's a flywheel too. >> Oh that's (indistinct) >> (indistinct) >> We are running the business, right? >> It's an interesting number, but it's also kind of, like, context to it. So here, a hundred million spend it, you get there, but you got to factor in the fact that the ways companies win these days is critical mass scale, hitting a flywheel. If they can keep that flywheel of the value that they got going on and get better, you can almost imagine a marketplace where, hey, we have proprietary data, we're SiliconANGLE in theCUBE. We have proprietary content, CUBE videos, transcripts. Well wouldn't it be great if someone in a marketplace could sell a module for us, right? We buy that, Amazon's thing and things like that. So if they can get a marketplace going where you can apply to data sets that may be proprietary, you can start to see this become bigger. And so I think the key barriers to entry is going to be success. I'll give you an example, Reddit. Reddit is successful and it's hard to copy, not because of the software. >> They built the moat. >> Because you can, buy Reddit open source software and try To compete. >> They built the moat with their community. >> Their community, their scale, their user expectation. Twitter, we referenced earlier, that thing should have gone under the first two years, but there was such a great emotional product. People would tolerate the fail whale. And then, you know, well that was a whole 'nother thing. >> Then a plane landed in (John laughs) the Hudson and it was over. >> I think verticals, a lot of verticals will build applications using these models like for lawyers, for doctors, for scientists, for content creators, for- >> So you'll have many hundreds of millions of dollars investments that are going to be seeping out. If, all right, we got to wrap, if you had to put odds on it that that OpenAI is going to be the leader, maybe not a winner take all leader, but like you look at like Amazon and cloud, they're not winner take all, these aren't necessarily winner take all markets. It's not necessarily a zero sum game, but let's call it winner take most. What odds would you give that open AI 10 years from now will be in that position. >> If I'm 0 to 10 kind of thing? >> Yeah, it's like horse race, 3 to 1, 2 to 1, even money, 10 to 1, 50 to 1. >> Maybe 2 to 1, >> 2 to 1, that's pretty low odds. That's basically saying they're the favorite, they're the front runner. Would you agree with that? >> I'd say 4 to 1. >> Yeah, I was going to say I'm like a 5 to 1, 7 to 1 type of person, 'cause I'm a skeptic with, you know, there's so much competition, but- >> I think they're definitely the leader. I mean you got to say, I mean. >> Oh there's no question. There's no question about it. >> The question is can they execute? >> They're not Friendster, is what you're saying. >> They're not Friendster and they're more like Twitter and Reddit where they have momentum. If they can execute on the product side, and if they don't stumble on that, they will continue to have the lead. >> If they say stay neutral, as Sam is, has been saying, that, hey, Microsoft is one of our partners, if you look at their company model, how they have structured the company, then they're going to pay back to the investors, like Microsoft is the biggest one, up to certain, like by certain number of years, they're going to pay back from all the money they make, and after that, they're going to give the money back to the public, to the, I don't know who they give it to, like non-profit or something. (indistinct) >> Okay, the odds are dropping. (group talks over each other) That's a good point though >> Actually they might have done that to fend off the criticism of this. But it's really interesting to see the model they have adopted. >> The wildcard in all this, My last word on this is that, if there's a developer shift in how developers and data can come together again, we have conferences around the future of data, Supercloud and meshs versus, you know, how the data world, coding with data, how that evolves will also dictate, 'cause a wild card could be a shift in the landscape around how developers are using either machine learning or AI like techniques to code into their apps, so. >> That's fantastic insight. I can't thank you enough for your time, on the heels of Supercloud 2, really appreciate it. All right, thanks to John and Sarbjeet for the outstanding conversation today. Special thanks to the Palo Alto studio team. My goodness, Anderson, this great backdrop. You guys got it all out here, I'm jealous. And Noah, really appreciate it, Chuck, Andrew Frick and Cameron, Andrew Frick switching, Cameron on the video lake, great job. And Alex Myerson, he's on production, manages the podcast for us, Ken Schiffman as well. Kristen Martin and Cheryl Knight help get the word out on social media and our newsletters. Rob Hof is our editor-in-chief over at SiliconANGLE, does some great editing, thanks to all. Remember, all these episodes are available as podcasts. All you got to do is search Breaking Analysis podcast, wherever you listen. Publish each week on wikibon.com and siliconangle.com. Want to get in touch, email me directly, david.vellante@siliconangle.com or DM me at dvellante, or comment on our LinkedIn post. And by all means, check out etr.ai. They got really great survey data in the enterprise tech business. This is Dave Vellante for theCUBE Insights powered by ETR. Thanks for watching, We'll see you next time on Breaking Analysis. (electronic music)

>> From theCUBE Studios in Palo Alto and Boston, bringing you data-driven insights from theCUBE and ETR. This is Breaking Analysis with Dave Vellante. >> Spark became a top level Apache project in 2014, and then shortly thereafter, burst onto the big data scene. Spark, along with the cloud, transformed and in many ways, disrupted the big data market. Databricks optimized its tech stack for Spark and took advantage of the cloud to really cleverly deliver a managed service that has become a leading AI and data platform among data scientists and data engineers. However, emerging customer data requirements are shifting into a direction that will cause modern data platform players generally and Databricks, specifically, we think, to make some key directional decisions and perhaps even reinvent themselves. Hello and welcome to this week's wikibon theCUBE Insights, powered by ETR. In this Breaking Analysis, we're going to do a deep dive into Databricks. We'll explore its current impressive market momentum. We're going to use some ETR survey data to show that, and then we'll lay out how customer data requirements are changing and what the ideal data platform will look like in the midterm future. We'll then evaluate core elements of the Databricks portfolio against that vision, and then we'll close with some strategic decisions that we think the company faces. And to do so, we welcome in our good friend, George Gilbert, former equities analyst, market analyst, and current Principal at TechAlpha Partners. George, good to see you. Thanks for coming on. >> Good to see you, Dave. >> All right, let me set this up. We're going to start by taking a look at where Databricks sits in the market in terms of how customers perceive the company and what it's momentum looks like. And this chart that we're showing here is data from ETS, the emerging technology survey of private companies. The N is 1,421. What we did is we cut the data on three sectors, analytics, database-data warehouse, and AI/ML. The vertical axis is a measure of customer sentiment, which evaluates an IT decision maker's awareness of the firm and the likelihood of engaging and/or purchase intent. The horizontal axis shows mindshare in the dataset, and we've highlighted Databricks, which has been a consistent high performer in this survey over the last several quarters. And as we, by the way, just as aside as we previously reported, OpenAI, which burst onto the scene this past quarter, leads all names, but Databricks is still prominent. You can see that the ETR shows some open source tools for reference, but as far as firms go, Databricks is very impressively positioned. Now, let's see how they stack up to some mainstream cohorts in the data space, against some bigger companies and sometimes public companies. This chart shows net score on the vertical axis, which is a measure of spending momentum and pervasiveness in the data set is on the horizontal axis. You can see that chart insert in the upper right, that informs how the dots are plotted, and net score against shared N. And that red dotted line at 40% indicates a highly elevated net score, anything above that we think is really, really impressive. And here we're just comparing Databricks with Snowflake, Cloudera, and Oracle. And that squiggly line leading to Databricks shows their path since 2021 by quarter. And you can see it's performing extremely well, maintaining an elevated net score and net range. Now it's comparable in the vertical axis to Snowflake, and it consistently is moving to the right and gaining share. Now, why did we choose to show Cloudera and Oracle? The reason is that Cloudera got the whole big data era started and was disrupted by Spark. And of course the cloud, Spark and Databricks and Oracle in many ways, was the target of early big data players like Cloudera. Take a listen to Cloudera CEO at the time, Mike Olson. This is back in 2010, first year of theCUBE, play the clip. >> Look, back in the day, if you had a data problem, if you needed to run business analytics, you wrote the biggest check you could to Sun Microsystems, and you bought a great big, single box, central server, and any money that was left over, you handed to Oracle for a database licenses and you installed that database on that box, and that was where you went for data. That was your temple of information. >> Okay? So Mike Olson implied that monolithic model was too expensive and inflexible, and Cloudera set out to fix that. But the best laid plans, as they say, George, what do you make of the data that we just shared? >> So where Databricks has really come up out of sort of Cloudera's tailpipe was they took big data processing, made it coherent, made it a managed service so it could run in the cloud. So it relieved customers of the operational burden. Where they're really strong and where their traditional meat and potatoes or bread and butter is the predictive and prescriptive analytics that building and training and serving machine learning models. They've tried to move into traditional business intelligence, the more traditional descriptive and diagnostic analytics, but they're less mature there. So what that means is, the reason you see Databricks and Snowflake kind of side by side is there are many, many accounts that have both Snowflake for business intelligence, Databricks for AI machine learning, where Snowflake, I'm sorry, where Databricks also did really well was in core data engineering, refining the data, the old ETL process, which kind of turned into ELT, where you loaded into the analytic repository in raw form and refine it. And so people have really used both, and each is trying to get into the other. >> Yeah, absolutely. We've reported on this quite a bit. Snowflake, kind of moving into the domain of Databricks and vice versa. And the last bit of ETR evidence that we want to share in terms of the company's momentum comes from ETR's Round Tables. They're run by Erik Bradley, and now former Gartner analyst and George, your colleague back at Gartner, Daren Brabham. And what we're going to show here is some direct quotes of IT pros in those Round Tables. There's a data science head and a CIO as well. Just make a few call outs here, we won't spend too much time on it, but starting at the top, like all of us, we can't talk about Databricks without mentioning Snowflake. Those two get us excited. Second comment zeros in on the flexibility and the robustness of Databricks from a data warehouse perspective. And then the last point is, despite competition from cloud players, Databricks has reinvented itself a couple of times over the year. And George, we're going to lay out today a scenario that perhaps calls for Databricks to do that once again. >> Their big opportunity and their big challenge for every tech company, it's managing a technology transition. The transition that we're talking about is something that's been bubbling up, but it's really epical. First time in 60 years, we're moving from an application-centric view of the world to a data-centric view, because decisions are becoming more important than automating processes. So let me let you sort of develop. >> Yeah, so let's talk about that here. We going to put up some bullets on precisely that point and the changing sort of customer environment. So you got IT stacks are shifting is George just said, from application centric silos to data centric stacks where the priority is shifting from automating processes to automating decision. You know how look at RPA and there's still a lot of automation going on, but from the focus of that application centricity and the data locked into those apps, that's changing. Data has historically been on the outskirts in silos, but organizations, you think of Amazon, think Uber, Airbnb, they're putting data at the core, and logic is increasingly being embedded in the data instead of the reverse. In other words, today, the data's locked inside the app, which is why you need to extract that data is sticking it to a data warehouse. The point, George, is we're putting forth this new vision for how data is going to be used. And you've used this Uber example to underscore the future state. Please explain? >> Okay, so this is hopefully an example everyone can relate to. The idea is first, you're automating things that are happening in the real world and decisions that make those things happen autonomously without humans in the loop all the time. So to use the Uber example on your phone, you call a car, you call a driver. Automatically, the Uber app then looks at what drivers are in the vicinity, what drivers are free, matches one, calculates an ETA to you, calculates a price, calculates an ETA to your destination, and then directs the driver once they're there. The point of this is that that cannot happen in an application-centric world very easily because all these little apps, the drivers, the riders, the routes, the fares, those call on data locked up in many different apps, but they have to sit on a layer that makes it all coherent. >> But George, so if Uber's doing this, doesn't this tech already exist? Isn't there a tech platform that does this already? >> Yes, and the mission of the entire tech industry is to build services that make it possible to compose and operate similar platforms and tools, but with the skills of mainstream developers in mainstream corporations, not the rocket scientists at Uber and Amazon. >> Okay, so we're talking about horizontally scaling across the industry, and actually giving a lot more organizations access to this technology. So by way of review, let's summarize the trend that's going on today in terms of the modern data stack that is propelling the likes of Databricks and Snowflake, which we just showed you in the ETR data and is really is a tailwind form. So the trend is toward this common repository for analytic data, that could be multiple virtual data warehouses inside of Snowflake, but you're in that Snowflake environment or Lakehouses from Databricks or multiple data lakes. And we've talked about what JP Morgan Chase is doing with the data mesh and gluing data lakes together, you've got various public clouds playing in this game, and then the data is annotated to have a common meaning. In other words, there's a semantic layer that enables applications to talk to the data elements and know that they have common and coherent meaning. So George, the good news is this approach is more effective than the legacy monolithic models that Mike Olson was talking about, so what's the problem with this in your view? >> So today's data platforms added immense value 'cause they connected the data that was previously locked up in these monolithic apps or on all these different microservices, and that supported traditional BI and AI/ML use cases. But now if we want to build apps like Uber or Amazon.com, where they've got essentially an autonomously running supply chain and e-commerce app where humans only care and feed it. But the thing is figuring out what to buy, when to buy, where to deploy it, when to ship it. We needed a semantic layer on top of the data. So that, as you were saying, the data that's coming from all those apps, the different apps that's integrated, not just connected, but it means the same. And the issue is whenever you add a new layer to a stack to support new applications, there are implications for the already existing layers, like can they support the new layer and its use cases? So for instance, if you add a semantic layer that embeds app logic with the data rather than vice versa, which we been talking about and that's been the case for 60 years, then the new data layer faces challenges that the way you manage that data, the way you analyze that data, is not supported by today's tools. >> Okay, so actually Alex, bring me up that last slide if you would, I mean, you're basically saying at the bottom here, today's repositories don't really do joins at scale. The future is you're talking about hundreds or thousands or millions of data connections, and today's systems, we're talking about, I don't know, 6, 8, 10 joins and that is the fundamental problem you're saying, is a new data error coming and existing systems won't be able to handle it? >> Yeah, one way of thinking about it is that even though we call them relational databases, when we actually want to do lots of joins or when we want to analyze data from lots of different tables, we created a whole new industry for analytic databases where you sort of mung the data together into fewer tables. So you didn't have to do as many joins because the joins are difficult and slow. And when you're going to arbitrarily join thousands, hundreds of thousands or across millions of elements, you need a new type of database. We have them, they're called graph databases, but to query them, you go back to the prerelational era in terms of their usability. >> Okay, so we're going to come back to that and talk about how you get around that problem. But let's first lay out what the ideal data platform of the future we think looks like. And again, we're going to come back to use this Uber example. In this graphic that George put together, awesome. We got three layers. The application layer is where the data products reside. The example here is drivers, rides, maps, routes, ETA, et cetera. The digital version of what we were talking about in the previous slide, people, places and things. The next layer is the data layer, that breaks down the silos and connects the data elements through semantics and everything is coherent. And then the bottom layers, the legacy operational systems feed that data layer. George, explain what's different here, the graph database element, you talk about the relational query capabilities, and why can't I just throw memory at solving this problem? >> Some of the graph databases do throw memory at the problem and maybe without naming names, some of them live entirely in memory. And what you're dealing with is a prerelational in-memory database system where you navigate between elements, and the issue with that is we've had SQL for 50 years, so we don't have to navigate, we can say what we want without how to get it. That's the core of the problem. >> Okay. So if I may, I just want to drill into this a little bit. So you're talking about the expressiveness of a graph. Alex, if you'd bring that back out, the fourth bullet, expressiveness of a graph database with the relational ease of query. Can you explain what you mean by that? >> Yeah, so graphs are great because when you can describe anything with a graph, that's why they're becoming so popular. Expressive means you can represent anything easily. They're conducive to, you might say, in a world where we now want like the metaverse, like with a 3D world, and I don't mean the Facebook metaverse, I mean like the business metaverse when we want to capture data about everything, but we want it in context, we want to build a set of digital twins that represent everything going on in the world. And Uber is a tiny example of that. Uber built a graph to represent all the drivers and riders and maps and routes. But what you need out of a database isn't just a way to store stuff and update stuff. You need to be able to ask questions of it, you need to be able to query it. And if you go back to prerelational days, you had to know how to find your way to the data. It's sort of like when you give directions to someone and they didn't have a GPS system and a mapping system, you had to give them turn by turn directions. Whereas when you have a GPS and a mapping system, which is like the relational thing, you just say where you want to go, and it spits out the turn by turn directions, which let's say, the car might follow or whoever you're directing would follow. But the point is, it's much easier in a relational database to say, "I just want to get these results. You figure out how to get it." The graph database, they have not taken over the world because in some ways, it's taking a 50 year leap backwards. >> Alright, got it. Okay. Let's take a look at how the current Databricks offerings map to that ideal state that we just laid out. So to do that, we put together this chart that looks at the key elements of the Databricks portfolio, the core capability, the weakness, and the threat that may loom. Start with the Delta Lake, that's the storage layer, which is great for files and tables. It's got true separation of compute and storage, I want you to double click on that George, as independent elements, but it's weaker for the type of low latency ingest that we see coming in the future. And some of the threats highlighted here. AWS could add transactional tables to S3, Iceberg adoption is picking up and could accelerate, that could disrupt Databricks. George, add some color here please? >> Okay, so this is the sort of a classic competitive forces where you want to look at, so what are customers demanding? What's competitive pressure? What are substitutes? Even what your suppliers might be pushing. Here, Delta Lake is at its core, a set of transactional tables that sit on an object store. So think of it in a database system, this is the storage engine. So since S3 has been getting stronger for 15 years, you could see a scenario where they add transactional tables. We have an open source alternative in Iceberg, which Snowflake and others support. But at the same time, Databricks has built an ecosystem out of tools, their own and others, that read and write to Delta tables, that's what makes the Delta Lake and ecosystem. So they have a catalog, the whole machine learning tool chain talks directly to the data here. That was their great advantage because in the past with Snowflake, you had to pull all the data out of the database before the machine learning tools could work with it, that was a major shortcoming. They fixed that. But the point here is that even before we get to the semantic layer, the core foundation is under threat. >> Yep. Got it. Okay. We got a lot of ground to cover. So we're going to take a look at the Spark Execution Engine next. Think of that as the refinery that runs really efficient batch processing. That's kind of what disrupted the DOOp in a large way, but it's not Python friendly and that's an issue because the data science and the data engineering crowd are moving in that direction, and/or they're using DBT. George, we had Tristan Handy on at Supercloud, really interesting discussion that you and I did. Explain why this is an issue for Databricks? >> So once the data lake was in place, what people did was they refined their data batch, and Spark has always had streaming support and it's gotten better. The underlying storage as we've talked about is an issue. But basically they took raw data, then they refined it into tables that were like customers and products and partners. And then they refined that again into what was like gold artifacts, which might be business intelligence metrics or dashboards, which were collections of metrics. But they were running it on the Spark Execution Engine, which it's a Java-based engine or it's running on a Java-based virtual machine, which means all the data scientists and the data engineers who want to work with Python are really working in sort of oil and water. Like if you get an error in Python, you can't tell whether the problems in Python or where it's in Spark. There's just an impedance mismatch between the two. And then at the same time, the whole world is now gravitating towards DBT because it's a very nice and simple way to compose these data processing pipelines, and people are using either SQL in DBT or Python in DBT, and that kind of is a substitute for doing it all in Spark. So it's under threat even before we get to that semantic layer, it so happens that DBT itself is becoming the authoring environment for the semantic layer with business intelligent metrics. But that's again, this is the second element that's under direct substitution and competitive threat. >> Okay, let's now move down to the third element, which is the Photon. Photon is Databricks' BI Lakehouse, which has integration with the Databricks tooling, which is very rich, it's newer. And it's also not well suited for high concurrency and low latency use cases, which we think are going to increasingly become the norm over time. George, the call out threat here is customers want to connect everything to a semantic layer. Explain your thinking here and why this is a potential threat to Databricks? >> Okay, so two issues here. What you were touching on, which is the high concurrency, low latency, when people are running like thousands of dashboards and data is streaming in, that's a problem because SQL data warehouse, the query engine, something like that matures over five to 10 years. It's one of these things, the joke that Andy Jassy makes just in general, he's really talking about Azure, but there's no compression algorithm for experience. The Snowflake guy started more than five years earlier, and for a bunch of reasons, that lead is not something that Databricks can shrink. They'll always be behind. So that's why Snowflake has transactional tables now and we can get into that in another show. But the key point is, so near term, it's struggling to keep up with the use cases that are core to business intelligence, which is highly concurrent, lots of users doing interactive query. But then when you get to a semantic layer, that's when you need to be able to query data that might have thousands or tens of thousands or hundreds of thousands of joins. And that's a SQL query engine, traditional SQL query engine is just not built for that. That's the core problem of traditional relational databases. >> Now this is a quick aside. We always talk about Snowflake and Databricks in sort of the same context. We're not necessarily saying that Snowflake is in a position to tackle all these problems. We'll deal with that separately. So we don't mean to imply that, but we're just sort of laying out some of the things that Snowflake or rather Databricks customers we think, need to be thinking about and having conversations with Databricks about and we hope to have them as well. We'll come back to that in terms of sort of strategic options. But finally, when come back to the table, we have Databricks' AI/ML Tool Chain, which has been an awesome capability for the data science crowd. It's comprehensive, it's a one-stop shop solution, but the kicker here is that it's optimized for supervised model building. And the concern is that foundational models like GPT could cannibalize the current Databricks tooling, but George, can't Databricks, like other software companies, integrate foundation model capabilities into its platform? >> Okay, so the sound bite answer to that is sure, IBM 3270 terminals could call out to a graphical user interface when they're running on the XT terminal, but they're not exactly good citizens in that world. The core issue is Databricks has this wonderful end-to-end tool chain for training, deploying, monitoring, running inference on supervised models. But the paradigm there is the customer builds and trains and deploys each model for each feature or application. In a world of foundation models which are pre-trained and unsupervised, the entire tool chain is different. So it's not like Databricks can junk everything they've done and start over with all their engineers. They have to keep maintaining what they've done in the old world, but they have to build something new that's optimized for the new world. It's a classic technology transition and their mentality appears to be, "Oh, we'll support the new stuff from our old stuff." Which is suboptimal, and as we'll talk about, their biggest patron and the company that put them on the map, Microsoft, really stopped working on their old stuff three years ago so that they could build a new tool chain optimized for this new world. >> Yeah, and so let's sort of close with what we think the options are and decisions that Databricks has for its future architecture. They're smart people. I mean we've had Ali Ghodsi on many times, super impressive. I think they've got to be keenly aware of the limitations, what's going on with foundation models. But at any rate, here in this chart, we lay out sort of three scenarios. One is re-architect the platform by incrementally adopting new technologies. And example might be to layer a graph query engine on top of its stack. They could license key technologies like graph database, they could get aggressive on M&A and buy-in, relational knowledge graphs, semantic technologies, vector database technologies. George, as David Floyer always says, "A lot of ways to skin a cat." We've seen companies like, even think about EMC maintained its relevance through M&A for many, many years. George, give us your thought on each of these strategic options? >> Okay, I find this question the most challenging 'cause remember, I used to be an equity research analyst. I worked for Frank Quattrone, we were one of the top tech shops in the banking industry, although this is 20 years ago. But the M&A team was the top team in the industry and everyone wanted them on their side. And I remember going to meetings with these CEOs, where Frank and the bankers would say, "You want us for your M&A work because we can do better." And they really could do better. But in software, it's not like with EMC in hardware because with hardware, it's easier to connect different boxes. With software, the whole point of a software company is to integrate and architect the components so they fit together and reinforce each other, and that makes M&A harder. You can do it, but it takes a long time to fit the pieces together. Let me give you examples. If they put a graph query engine, let's say something like TinkerPop, on top of, I don't even know if it's possible, but let's say they put it on top of Delta Lake, then you have this graph query engine talking to their storage layer, Delta Lake. But if you want to do analysis, you got to put the data in Photon, which is not really ideal for highly connected data. If you license a graph database, then most of your data is in the Delta Lake and how do you sync it with the graph database? If you do sync it, you've got data in two places, which kind of defeats the purpose of having a unified repository. I find this semantic layer option in number three actually more promising, because that's something that you can layer on top of the storage layer that you have already. You just have to figure out then how to have your query engines talk to that. What I'm trying to highlight is, it's easy as an analyst to say, "You can buy this company or license that technology." But the really hard work is making it all work together and that is where the challenge is. >> Yeah, and well look, I thank you for laying that out. We've seen it, certainly Microsoft and Oracle. I guess you might argue that well, Microsoft had a monopoly in its desktop software and was able to throw off cash for a decade plus while it's stock was going sideways. Oracle had won the database wars and had amazing margins and cash flow to be able to do that. Databricks isn't even gone public yet, but I want to close with some of the players to watch. Alex, if you'd bring that back up, number four here. AWS, we talked about some of their options with S3 and it's not just AWS, it's blob storage, object storage. Microsoft, as you sort of alluded to, was an early go-to market channel for Databricks. We didn't address that really. So maybe in the closing comments we can. Google obviously, Snowflake of course, we're going to dissect their options in future Breaking Analysis. Dbt labs, where do they fit? Bob Muglia's company, Relational.ai, why are these players to watch George, in your opinion? >> So everyone is trying to assemble and integrate the pieces that would make building data applications, data products easy. And the critical part isn't just assembling a bunch of pieces, which is traditionally what AWS did. It's a Unix ethos, which is we give you the tools, you put 'em together, 'cause you then have the maximum choice and maximum power. So what the hyperscalers are doing is they're taking their key value stores, in the case of ASW it's DynamoDB, in the case of Azure it's Cosmos DB, and each are putting a graph query engine on top of those. So they have a unified storage and graph database engine, like all the data would be collected in the key value store. Then you have a graph database, that's how they're going to be presenting a foundation for building these data apps. Dbt labs is putting a semantic layer on top of data lakes and data warehouses and as we'll talk about, I'm sure in the future, that makes it easier to swap out the underlying data platform or swap in new ones for specialized use cases. Snowflake, what they're doing, they're so strong in data management and with their transactional tables, what they're trying to do is take in the operational data that used to be in the province of many state stores like MongoDB and say, "If you manage that data with us, it'll be connected to your analytic data without having to send it through a pipeline." And that's hugely valuable. Relational.ai is the wildcard, 'cause what they're trying to do, it's almost like a holy grail where you're trying to take the expressiveness of connecting all your data in a graph but making it as easy to query as you've always had it in a SQL database or I should say, in a relational database. And if they do that, it's sort of like, it'll be as easy to program these data apps as a spreadsheet was compared to procedural languages, like BASIC or Pascal. That's the implications of Relational.ai. >> Yeah, and again, we talked before, why can't you just throw this all in memory? We're talking in that example of really getting down to differences in how you lay the data out on disk in really, new database architecture, correct? >> Yes. And that's why it's not clear that you could take a data lake or even a Snowflake and why you can't put a relational knowledge graph on those. You could potentially put a graph database, but it'll be compromised because to really do what Relational.ai has done, which is the ease of Relational on top of the power of graph, you actually need to change how you're storing your data on disk or even in memory. So you can't, in other words, it's not like, oh we can add graph support to Snowflake, 'cause if you did that, you'd have to change, or in your data lake, you'd have to change how the data is physically laid out. And then that would break all the tools that talk to that currently. >> What in your estimation, is the timeframe where this becomes critical for a Databricks and potentially Snowflake and others? I mentioned earlier midterm, are we talking three to five years here? Are we talking end of decade? What's your radar say? >> I think something surprising is going on that's going to sort of come up the tailpipe and take everyone by storm. All the hype around business intelligence metrics, which is what we used to put in our dashboards where bookings, billings, revenue, customer, those things, those were the key artifacts that used to live in definitions in your BI tools, and DBT has basically created a standard for defining those so they live in your data pipeline or they're defined in their data pipeline and executed in the data warehouse or data lake in a shared way, so that all tools can use them. This sounds like a digression, it's not. All this stuff about data mesh, data fabric, all that's going on is we need a semantic layer and the business intelligence metrics are defining common semantics for your data. And I think we're going to find by the end of this year, that metrics are how we annotate all our analytic data to start adding common semantics to it. And we're going to find this semantic layer, it's not three to five years off, it's going to be staring us in the face by the end of this year. >> Interesting. And of course SVB today was shut down. We're seeing serious tech headwinds, and oftentimes in these sort of downturns or flat turns, which feels like this could be going on for a while, we emerge with a lot of new players and a lot of new technology. George, we got to leave it there. Thank you to George Gilbert for excellent insights and input for today's episode. I want to thank Alex Myerson who's on production and manages the podcast, of course Ken Schiffman as well. Kristin Martin and Cheryl Knight help get the word out on social media and in our newsletters. And Rob Hof is our EIC over at Siliconangle.com, he does some great editing. Remember all these episodes, they're available as podcasts. Wherever you listen, all you got to do is search Breaking Analysis Podcast, we publish each week on wikibon.com and siliconangle.com, or you can email me at David.Vellante@siliconangle.com, or DM me @DVellante. Comment on our LinkedIn post, and please do check out ETR.ai, great survey data, enterprise tech focus, phenomenal. This is Dave Vellante for theCUBE Insights powered by ETR. Thanks for watching, and we'll see you next time on Breaking Analysis.

(upbeat music) >> Hello, everyone, and welcome to theCUBE's presentation of the "AWS Startup Showcase." This is season three, episode one. The focus of this episode is AI/ML: Top Startups Building Foundational Models, Infrastructure, and AI. It's great topics, super-relevant, and it's part of our ongoing coverage of startups in the AWS ecosystem. I'm your host, John Furrier, with theCUBE. Today, we're excited to be joined by Jay Marshall, VP of Business Development at Neural Magic. Jay, thanks for coming on theCUBE. >> Hey, John, thanks so much. Thanks for having us. >> We had a great CUBE conversation with you guys. This is very much about the company focuses. It's a feature presentation for the "Startup Showcase," and the machine learning at scale is the topic, but in general, it's more, (laughs) and we should call it "Machine Learning and AI: How to Get Started," because everybody is retooling their business. Companies that aren't retooling their business right now with AI first will be out of business, in my opinion. You're seeing massive shift. This is really truly the beginning of the next-gen machine learning AI trend. It's really seeing ChatGPT. Everyone sees that. That went mainstream. But this is just the beginning. This is scratching the surface of this next-generation AI with machine learning powering it, and with all the goodness of cloud, cloud scale, and how horizontally scalable it is. The resources are there. You got the Edge. Everything's perfect for AI 'cause data infrastructure's exploding in value. AI is just the applications. This is a super topic, so what do you guys see in this general area of opportunities right now in the headlines? And I'm sure you guys' phone must be ringing off the hook, metaphorically speaking, or emails and meetings and Zooms. What's going on over there at Neural Magic? >> No, absolutely, and you pretty much nailed most of it. I think that, you know, my background, we've seen for the last 20-plus years. Even just getting enterprise applications kind of built and delivered at scale, obviously, amazing things with AWS and the cloud to help accelerate that. And we just kind of figured out in the last five or so years how to do that productively and efficiently, kind of from an operations perspective. Got development and operations teams. We even came up with DevOps, right? But now, we kind of have this new kind of persona and new workload that developers have to talk to, and then it has to be deployed on those ITOps solutions. And so you pretty much nailed it. Folks are saying, "Well, how do I do this?" These big, generational models or foundational models, as we're calling them, they're great, but enterprises want to do that with their data, on their infrastructure, at scale, at the edge. So for us, yeah, we're helping enterprises accelerate that through optimizing models and then delivering them at scale in a more cost-effective fashion. >> Yeah, and I think one of the things, the benefits of OpenAI we saw, was not only is it open source, then you got also other models that are more proprietary, is that it shows the world that this is really happening, right? It's a whole nother level, and there's also new landscape kind of maps coming out. You got the generative AI, and you got the foundational models, large LLMs. Where do you guys fit into the landscape? Because you guys are in the middle of this. How do you talk to customers when they say, "I'm going down this road. I need help. I'm going to stand this up." This new AI infrastructure and applications, where do you guys fit in the landscape? >> Right, and really, the answer is both. I think today, when it comes to a lot of what for some folks would still be considered kind of cutting edge around computer vision and natural language processing, a lot of our optimization tools and our runtime are based around most of the common computer vision and natural language processing models. So your YOLOs, your BERTs, you know, your DistilBERTs and what have you, so we work to help optimize those, again, who've gotten great performance and great value for customers trying to get those into production. But when you get into the LLMs, and you mentioned some of the open source components there, our research teams have kind of been right in the trenches with those. So kind of the GPT open source equivalent being OPT, being able to actually take, you know, a multi-$100 billion parameter model and sparsify that or optimize that down, shaving away a ton of parameters, and being able to run it on smaller infrastructure. So I think the evolution here, you know, all this stuff came out in the last six months in terms of being turned loose into the wild, but we're staying in the trenches with folks so that we can help optimize those as well and not require, again, the heavy compute, the heavy cost, the heavy power consumption as those models evolve as well. So we're staying right in with everybody while they're being built, but trying to get folks into production today with things that help with business value today. >> Jay, I really appreciate you coming on theCUBE, and before we came on camera, you said you just were on a customer call. I know you got a lot of activity. What specific things are you helping enterprises solve? What kind of problems? Take us through the spectrum from the beginning, people jumping in the deep end of the pool, some people kind of coming in, starting out slow. What are the scale? Can you scope the kind of use cases and problems that are emerging that people are calling you for? >> Absolutely, so I think if I break it down to kind of, like, your startup, or I maybe call 'em AI native to kind of steal from cloud native years ago, that group, it's pretty much, you know, part and parcel for how that group already runs. So if you have a data science team and an ML engineering team, you're building models, you're training models, you're deploying models. You're seeing firsthand the expense of starting to try to do that at scale. So it's really just a pure operational efficiency play. They kind of speak natively to our tools, which we're doing in the open source. So it's really helping, again, with the optimization of the models they've built, and then, again, giving them an alternative to expensive proprietary hardware accelerators to have to run them. Now, on the enterprise side, it varies, right? You have some kind of AI native folks there that already have these teams, but you also have kind of, like, AI curious, right? Like, they want to do it, but they don't really know where to start, and so for there, we actually have an open source toolkit that can help you get into this optimization, and then again, that runtime, that inferencing runtime, purpose-built for CPUs. It allows you to not have to worry, again, about do I have a hardware accelerator available? How do I integrate that into my application stack? If I don't already know how to build this into my infrastructure, does my ITOps teams, do they know how to do this, and what does that runway look like? How do I cost for this? How do I plan for this? When it's just x86 compute, we've been doing that for a while, right? So it obviously still requires more, but at least it's a little bit more predictable. >> It's funny you mentioned AI native. You know, born in the cloud was a phrase that was out there. Now, you have startups that are born in AI companies. So I think you have this kind of cloud kind of vibe going on. You have lift and shift was a big discussion. Then you had cloud native, kind of in the cloud, kind of making it all work. Is there a existing set of things? People will throw on this hat, and then what's the difference between AI native and kind of providing it to existing stuff? 'Cause we're a lot of people take some of these tools and apply it to either existing stuff almost, and it's not really a lift and shift, but it's kind of like bolting on AI to something else, and then starting with AI first or native AI. >> Absolutely. It's a- >> How would you- >> It's a great question. I think that probably, where I'd probably pull back to kind of allow kind of retail-type scenarios where, you know, for five, seven, nine years or more even, a lot of these folks already have data science teams, you know? I mean, they've been doing this for quite some time. The difference is the introduction of these neural networks and deep learning, right? Those kinds of models are just a little bit of a paradigm shift. So, you know, I obviously was trying to be fun with the term AI native, but I think it's more folks that kind of came up in that neural network world, so it's a little bit more second nature, whereas I think for maybe some traditional data scientists starting to get into neural networks, you have the complexity there and the training overhead, and a lot of the aspects of getting a model finely tuned and hyperparameterization and all of these aspects of it. It just adds a layer of complexity that they're just not as used to dealing with. And so our goal is to help make that easy, and then of course, make it easier to run anywhere that you have just kind of standard infrastructure. >> Well, the other point I'd bring out, and I'd love to get your reaction to, is not only is that a neural network team, people who have been focused on that, but also, if you look at some of the DataOps lately, AIOps markets, a lot of data engineering, a lot of scale, folks who have been kind of, like, in that data tsunami cloud world are seeing, they kind of been in this, right? They're, like, been experiencing that. >> No doubt. I think it's funny the data lake concept, right? And you got data oceans now. Like, the metaphors just keep growing on us, but where it is valuable in terms of trying to shift the mindset, I've always kind of been a fan of some of the naming shift. I know with AWS, they always talk about purpose-built databases. And I always liked that because, you know, you don't have one database that can do everything. Even ones that say they can, like, you still have to do implementation detail differences. So sitting back and saying, "What is my use case, and then which database will I use it for?" I think it's kind of similar here. And when you're building those data teams, if you don't have folks that are doing data engineering, kind of that data harvesting, free processing, you got to do all that before a model's even going to care about it. So yeah, it's definitely a central piece of this as well, and again, whether or not you're going to be AI negative as you're making your way to kind of, you know, on that journey, you know, data's definitely a huge component of it. >> Yeah, you would have loved our Supercloud event we had. Talk about naming and, you know, around data meshes was talked about a lot. You're starting to see the control plane layers of data. I think that was the beginning of what I saw as that data infrastructure shift, to be horizontally scalable. So I have to ask you, with Neural Magic, when your customers and the people that are prospects for you guys, they're probably asking a lot of questions because I think the general thing that we see is, "How do I get started? Which GPU do I use?" I mean, there's a lot of things that are kind of, I won't say technical or targeted towards people who are living in that world, but, like, as the mainstream enterprises come in, they're going to need a playbook. What do you guys see, what do you guys offer your clients when they come in, and what do you recommend? >> Absolutely, and I think where we hook in specifically tends to be on the training side. So again, I've built a model. Now, I want to really optimize that model. And then on the runtime side when you want to deploy it, you know, we run that optimized model. And so that's where we're able to provide. We even have a labs offering in terms of being able to pair up our engineering teams with a customer's engineering teams, and we can actually help with most of that pipeline. So even if it is something where you have a dataset and you want some help in picking a model, you want some help training it, you want some help deploying that, we can actually help there as well. You know, there's also a great partner ecosystem out there, like a lot of folks even in the "Startup Showcase" here, that extend beyond into kind of your earlier comment around data engineering or downstream ITOps or the all-up MLOps umbrella. So we can absolutely engage with our labs, and then, of course, you know, again, partners, which are always kind of key to this. So you are spot on. I think what's happened with the kind of this, they talk about a hockey stick. This is almost like a flat wall now with the rate of innovation right now in this space. And so we do have a lot of folks wanting to go straight from curious to native. And so that's definitely where the partner ecosystem comes in so hard 'cause there just isn't anybody or any teams out there that, I literally do from, "Here's my blank database, and I want an API that does all the stuff," right? Like, that's a big chunk, but we can definitely help with the model to delivery piece. >> Well, you guys are obviously a featured company in this space. Talk about the expertise. A lot of companies are like, I won't say faking it till they make it. You can't really fake security. You can't really fake AI, right? So there's going to be a learning curve. They'll be a few startups who'll come out of the gate early. You guys are one of 'em. Talk about what you guys have as expertise as a company, why you're successful, and what problems do you solve for customers? >> No, appreciate that. Yeah, we actually, we love to tell the story of our founder, Nir Shavit. So he's a 20-year professor at MIT. Actually, he was doing a lot of work on kind of multicore processing before there were even physical multicores, and actually even did a stint in computational neurobiology in the 2010s, and the impetus for this whole technology, has a great talk on YouTube about it, where he talks about the fact that his work there, he kind of realized that the way neural networks encode and how they're executed by kind of ramming data layer by layer through these kind of HPC-style platforms, actually was not analogous to how the human brain actually works. So we're on one side, we're building neural networks, and we're trying to emulate neurons. We're not really executing them that way. So our team, which one of the co-founders, also an ex-MIT, that was kind of the birth of why can't we leverage this super-performance CPU platform, which has those really fat, fast caches attached to each core, and actually start to find a way to break that model down in a way that I can execute things in parallel, not having to do them sequentially? So it is a lot of amazing, like, talks and stuff that show kind of the magic, if you will, a part of the pun of Neural Magic, but that's kind of the foundational layer of all the engineering that we do here. And in terms of how we're able to bring it to reality for customers, I'll give one customer quote where it's a large retailer, and it's a people-counting application. So a very common application. And that customer's actually been able to show literally double the amount of cameras being run with the same amount of compute. So for a one-to-one perspective, two-to-one, business leaders usually like that math, right? So we're able to show pure cost savings, but even performance-wise, you know, we have some of the common models like your ResNets and your YOLOs, where we can actually even perform better than hardware-accelerated solutions. So we're trying to do, I need to just dumb it down to better, faster, cheaper, but from a commodity perspective, that's where we're accelerating. >> That's not a bad business model. Make things easier to use, faster, and reduce the steps it takes to do stuff. So, you know, that's always going to be a good market. Now, you guys have DeepSparse, which we've talked about on our CUBE conversation prior to this interview, delivers ML models through the software so the hardware allows for a decoupling, right? >> Yep. >> Which is going to drive probably a cost advantage. Also, it's also probably from a deployment standpoint it must be easier. Can you share the benefits? Is it a cost side? Is it more of a deployment? What are the benefits of the DeepSparse when you guys decouple the software from the hardware on the ML models? >> No you actually, you hit 'em both 'cause that really is primarily the value. Because ultimately, again, we're so early. And I came from this world in a prior life where I'm doing Java development, WebSphere, WebLogic, Tomcat open source, right? When we were trying to do innovation, we had innovation buckets, 'cause everybody wanted to be on the web and have their app and a browser, right? We got all the money we needed to build something and show, hey, look at the thing on the web, right? But when you had to get in production, that was the challenge. So to what you're speaking to here, in this situation, we're able to show we're just a Python package. So whether you just install it on the operating system itself, or we also have a containerized version you can drop on any container orchestration platform, so ECS or EKS on AWS. And so you get all the auto-scaling features. So when you think about that kind of a world where you have everything from real-time inferencing to kind of after hours batch processing inferencing, the fact that you can auto scale that hardware up and down and it's CPU based, so you're paying by the minute instead of maybe paying by the hour at a lower cost shelf, it does everything from pure cost to, again, I can have my standard IT team say, "Hey, here's the Kubernetes in the container," and it just runs on the infrastructure we're already managing. So yeah, operational, cost and again, and many times even performance. (audio warbles) CPUs if I want to. >> Yeah, so that's easier on the deployment too. And you don't have this kind of, you know, blank check kind of situation where you don't know what's on the backend on the cost side. >> Exactly. >> And you control the actual hardware and you can manage that supply chain. >> And keep in mind, exactly. Because the other thing that sometimes gets lost in the conversation, depending on where a customer is, some of these workloads, like, you know, you and I remember a world where even like the roundtrip to the cloud and back was a problem for folks, right? We're used to extremely low latency. And some of these workloads absolutely also adhere to that. But there's some workloads where the latency isn't as important. And we actually even provide the tuning. Now, if we're giving you five milliseconds of latency and you don't need that, you can tune that back. So less CPU, lower cost. Now, throughput and other things come into play. But that's the kind of configurability and flexibility we give for operations. >> All right, so why should I call you if I'm a customer or prospect Neural Magic, what problem do I have or when do I know I need you guys? When do I call you in and what does my environment look like? When do I know? What are some of the signals that would tell me that I need Neural Magic? >> No, absolutely. So I think in general, any neural network, you know, the process I mentioned before called sparcification, it's, you know, an optimization process that we specialize in. Any neural network, you know, can be sparcified. So I think if it's a deep-learning neural network type model. If you're trying to get AI into production, you have cost concerns even performance-wise. I certainly hate to be too generic and say, "Hey, we'll talk to everybody." But really in this world right now, if it's a neural network, it's something where you're trying to get into production, you know, we are definitely offering, you know, kind of an at-scale performant deployable solution for deep learning models. >> So neural network you would define as what? Just devices that are connected that need to know about each other? What's the state-of-the-art current definition of neural network for customers that may think they have a neural network or might not know they have a neural network architecture? What is that definition for neural network? >> That's a great question. So basically, machine learning models that fall under this kind of category, you hear about transformers a lot, or I mentioned about YOLO, the YOLO family of computer vision models, or natural language processing models like BERT. If you have a data science team or even developers, some even regular, I used to call myself a nine to five developer 'cause I worked in the enterprise, right? So like, hey, we found a new open source framework, you know, I used to use Spring back in the day and I had to go figure it out. There's developers that are pulling these models down and they're figuring out how to get 'em into production, okay? So I think all of those kinds of situations, you know, if it's a machine learning model of the deep learning variety that's, you know, really specifically where we shine. >> Okay, so let me pretend I'm a customer for a minute. I have all these videos, like all these transcripts, I have all these people that we've interviewed, CUBE alumnis, and I say to my team, "Let's AI-ify, sparcify theCUBE." >> Yep. >> What do I do? I mean, do I just like, my developers got to get involved and they're going to be like, "Well, how do I upload it to the cloud? Do I use a GPU?" So there's a thought process. And I think a lot of companies are going through that example of let's get on this AI, how can it help our business? >> Absolutely. >> What does that progression look like? Take me through that example. I mean, I made up theCUBE example up, but we do have a lot of data. We have large data models and we have people and connect to the internet and so we kind of seem like there's a neural network. I think every company might have a neural network in place. >> Well, and I was going to say, I think in general, you all probably do represent even the standard enterprise more than most. 'Cause even the enterprise is going to have a ton of video content, a ton of text content. So I think it's a great example. So I think that that kind of sea or I'll even go ahead and use that term data lake again, of data that you have, you're probably going to want to be setting up kind of machine learning pipelines that are going to be doing all of the pre-processing from kind of the raw data to kind of prepare it into the format that say a YOLO would actually use or let's say BERT for natural language processing. So you have all these transcripts, right? So we would do a pre-processing path where we would create that into the file format that BERT, the machine learning model would know how to train off of. So that's kind of all the pre-processing steps. And then for training itself, we actually enable what's called sparse transfer learning. So that's transfer learning is a very popular method of doing training with existing models. So we would be able to retrain that BERT model with your transcript data that we have now done the pre-processing with to get it into the proper format. And now we have a BERT natural language processing model that's been trained on your data. And now we can deploy that onto DeepSparse runtime so that now you can ask that model whatever questions, or I should say pass, you're not going to ask it those kinds of questions ChatGPT, although we can do that too. But you're going to pass text through the BERT model and it's going to give you answers back. It could be things like sentiment analysis or text classification. You just call the model, and now when you pass text through it, you get the answers better, faster or cheaper. I'll use that reference again. >> Okay, we can create a CUBE bot to give us questions on the fly from the the AI bot, you know, from our previous guests. >> Well, and I will tell you using that as an example. So I had mentioned OPT before, kind of the open source version of ChatGPT. So, you know, typically that requires multiple GPUs to run. So our research team, I may have mentioned earlier, we've been able to sparcify that over 50% already and run it on only a single GPU. And so in that situation, you could train OPT with that corpus of data and do exactly what you say. Actually we could use Alexa, we could use Alexa to actually respond back with voice. How about that? We'll do an API call and we'll actually have an interactive Alexa-enabled bot. >> Okay, we're going to be a customer, let's put it on the list. But this is a great example of what you guys call software delivered AI, a topic we chatted about on theCUBE conversation. This really means this is a developer opportunity. This really is the convergence of the data growth, the restructuring, how data is going to be horizontally scalable, meets developers. So this is an AI developer model going on right now, which is kind of unique. >> It is, John, I will tell you what's interesting. And again, folks don't always think of it this way, you know, the AI magical goodness is now getting pushed in the middle where the developers and IT are operating. And so it again, that paradigm, although for some folks seem obvious, again, if you've been around for 20 years, that whole all that plumbing is a thing, right? And so what we basically help with is when you deploy the DeepSparse runtime, we have a very rich API footprint. And so the developers can call the API, ITOps can run it, or to your point, it's developer friendly enough that you could actually deploy our off-the-shelf models. We have something called the SparseZoo where we actually publish pre-optimized or pre-sparcified models. And so developers could literally grab those right off the shelf with the training they've already had and just put 'em right into their applications and deploy them as containers. So yeah, we enable that for sure as well. >> It's interesting, DevOps was infrastructure as code and we had a last season, a series on data as code, which we kind of coined. This is data as code. This is a whole nother level of opportunity where developers just want to have programmable data and apps with AI. This is a whole new- >> Absolutely. >> Well, absolutely great, great stuff. Our news team at SiliconANGLE and theCUBE said you guys had a little bit of a launch announcement you wanted to make here on the "AWS Startup Showcase." So Jay, you have something that you want to launch here? >> Yes, and thank you John for teeing me up. So I'm going to try to put this in like, you know, the vein of like an AWS, like main stage keynote launch, okay? So we're going to try this out. So, you know, a lot of our product has obviously been built on top of x86. I've been sharing that the past 15 minutes or so. And with that, you know, we're seeing a lot of acceleration for folks wanting to run on commodity infrastructure. But we've had customers and prospects and partners tell us that, you know, ARM and all of its kind of variance are very compelling, both cost performance-wise and also obviously with Edge. And wanted to know if there was anything we could do from a runtime perspective with ARM. And so we got the work and, you know, it's a hard problem to solve 'cause the instructions set for ARM is very different than the instruction set for x86, and our deep tensor column technology has to be able to work with that lower level instruction spec. But working really hard, the engineering team's been at it and we are happy to announce here at the "AWS Startup Showcase," that DeepSparse inference now has, or inference runtime now has support for AWS Graviton instances. So it's no longer just x86, it is also ARM and that obviously also opens up the door to Edge and further out the stack so that optimize once run anywhere, we're not going to open up. So it is an early access. So if you go to neuralmagic.com/graviton, you can sign up for early access, but we're excited to now get into the ARM side of the fence as well on top of Graviton. >> That's awesome. Our news team is going to jump on that news. We'll get it right up. We get a little scoop here on the "Startup Showcase." Jay Marshall, great job. That really highlights the flexibility that you guys have when you decouple the software from the hardware. And again, we're seeing open source driving a lot more in AI ops now with with machine learning and AI. So to me, that makes a lot of sense. And congratulations on that announcement. Final minute or so we have left, give a summary of what you guys are all about. Put a plug in for the company, what you guys are looking to do. I'm sure you're probably hiring like crazy. Take the last few minutes to give a plug for the company and give a summary. >> No, I appreciate that so much. So yeah, joining us out neuralmagic.com, you know, part of what we didn't spend a lot of time here, our optimization tools, we are doing all of that in the open source. It's called SparseML and I mentioned SparseZoo briefly. So we really want the data scientists community and ML engineering community to join us out there. And again, the DeepSparse runtime, it's actually free to use for trial purposes and for personal use. So you can actually run all this on your own laptop or on an AWS instance of your choice. We are now live in the AWS marketplace. So push button, deploy, come try us out and reach out to us on neuralmagic.com. And again, sign up for the Graviton early access. >> All right, Jay Marshall, Vice President of Business Development Neural Magic here, talking about performant, cost effective machine learning at scale. This is season three, episode one, focusing on foundational models as far as building data infrastructure and AI, AI native. I'm John Furrier with theCUBE. Thanks for watching. (bright upbeat music)

(soft music) >> Hello, everyone. Welcome to theCUBE's presentation of the AWS Startup Showcase. AI and Machine Learning: Top Startups Building Foundational Model Infrastructure. This is season 3, episode 1 of the ongoing series covering the exciting stuff from the AWS ecosystem, talking about machine learning and AI. I'm your host, John Furrier and today we are excited to be joined by Luis Ceze who's the CEO of OctoML and Anna Connolly, VP of customer success and experience OctoML. Great to have you on again, Luis. Anna, thanks for coming on. Appreciate it. >> Thank you, John. It's great to be here. >> Thanks for having us. >> I love the company. We had a CUBE conversation about this. You guys are really addressing how to run foundational models faster for less. And this is like the key theme. But before we get into it, this is a hot trend, but let's explain what you guys do. Can you set the narrative of what the company's about, why it was founded, what's your North Star and your mission? >> Yeah, so John, our mission is to make AI sustainable and accessible for everyone. And what we offer customers is, you know, a way of taking their models into production in the most efficient way possible by automating the process of getting a model and optimizing it for a variety of hardware and making cost-effective. So better, faster, cheaper model deployment. >> You know, the big trend here is AI. Everyone's seeing the ChatGPT, kind of the shot heard around the world. The BingAI and this fiasco and the ongoing experimentation. People are into it, and I think the business impact is clear. I haven't seen this in all of my career in the technology industry of this kind of inflection point. And every senior leader I talk to is rethinking about how to rebuild their business with AI because now the large language models have come in, these foundational models are here, they can see value in their data. This is a 10 year journey in the big data world. Now it's impacting that, and everyone's rebuilding their company around this idea of being AI first 'cause they see ways to eliminate things and make things more efficient. And so now they telling 'em to go do it. And they're like, what do we do? So what do you guys think? Can you explain what is this wave of AI and why is it happening, why now, and what should people pay attention to? What does it mean to them? >> Yeah, I mean, it's pretty clear by now that AI can do amazing things that captures people's imaginations. And also now can show things that are really impactful in businesses, right? So what people have the opportunity to do today is to either train their own model that adds value to their business or find open models out there that can do very valuable things to them. So the next step really is how do you take that model and put it into production in a cost-effective way so that the business can actually get value out of it, right? >> Anna, what's your take? Because customers are there, you're there to make 'em successful, you got the new secret weapon for their business. >> Yeah, I think we just see a lot of companies struggle to get from a trained model into a model that is deployed in a cost-effective way that actually makes sense for the application they're building. I think that's a huge challenge we see today, kind of across the board across all of our customers. >> Well, I see this, everyone asking the same question. I have data, I want to get value out of it. I got to get these big models, I got to train it. What's it going to cost? So I think there's a reality of, okay, I got to do it. Then no one has any visibility on what it costs. When they get into it, this is going to break the bank. So I have to ask you guys, the cost of training these models is on everyone's mind. OctoML, your company's focus on the cost side of it as well as the efficiency side of running these models in production. Why are the production costs such a concern and where specifically are people looking at it and why did it get here? >> Yeah, so training costs get a lot of attention because normally a large number, but we shouldn't forget that it's a large, typically one time upfront cost that customers pay. But, you know, when the model is put into production, the cost grows directly with model usage and you actually want your model to be used because it's adding value, right? So, you know, the question that a customer faces is, you know, they have a model, they have a trained model and now what? So how much would it cost to run in production, right? And now without the big wave in generative AI, which rightfully is getting a lot of attention because of the amazing things that it can do. It's important for us to keep in mind that generative AI models like ChatGPT are huge, expensive energy hogs. They cost a lot to run, right? And given that model usage growth directly, model cost grows directly with usage, what you want to do is make sure that once you put a model into production, you have the best cost structure possible so that you're not surprised when it's gets popular, right? So let me give you an example. So if you have a model that costs, say 1 to $2 million to train, but then it costs about one to two cents per session to use it, right? So if you have a million active users, even if they use just once a day, it's 10 to $20,000 a day to operate that model in production. And that very, very quickly, you know, get beyond what you paid to train it. >> Anna, these aren't small numbers, and it's cost to train and cost to operate, it kind of reminds me of when the cloud came around and the data center versus cloud options. Like, wait a minute, one, it costs a ton of cash to deploy, and then running it. This is kind of a similar dynamic. What are you seeing? >> Yeah, absolutely. I think we are going to see increasingly the cost and production outpacing the costs and training by a lot. I mean, people talk about training costs now because that's what they're confronting now because people are so focused on getting models performant enough to even use in an application. And now that we have them and they're that capable, we're really going to start to see production costs go up a lot. >> Yeah, Luis, if you don't mind, I know this might be a little bit of a tangent, but, you know, training's super important. I get that. That's what people are doing now, but then there's the deployment side of production. Where do people get caught up and miss the boat or misconfigure? What's the gotcha? Where's the trip wire or so to speak? Where do people mess up on the cost side? What do they do? Is it they don't think about it, they tie it to proprietary hardware? What's the issue? >> Yeah, several things, right? So without getting really technical, which, you know, I might get into, you know, you have to understand relationship between performance, you know, both in terms of latency and throughput and cost, right? So reducing latency is important because you improve responsiveness of the model. But it's really important to keep in mind that it often leads diminishing returns. Below a certain latency, making it faster won't make a measurable difference in experience, but it's going to cost a lot more. So understanding that is important. Now, if you care more about throughputs, which is the time it takes for you to, you know, units per period of time, you care about time to solution, we should think about this throughput per dollar. And understand what you want is the highest throughput per dollar, which may come at the cost of higher latency, which you're not going to care about, right? So, and the reality here, John, is that, you know, humans and especially folks in this space want to have the latest and greatest hardware. And often they commit a lot of money to get access to them and have to commit upfront before they understand the needs that their models have, right? So common mistake here, one is not spending time to understand what you really need, and then two, over-committing and using more hardware than you actually need. And not giving yourself enough freedom to get your workload to move around to the more cost-effective choice, right? So this is just a metaphoric choice. And then another thing that's important here too is making a model run faster on the hardware directly translates to lower cost, right? So, but it takes a lot of engineers, you need to think of ways of producing very efficient versions of your model for the target hardware that you're going to use. >> Anna, what's the customer angle here? Because price performance has been around for a long time, people get that, but now latency and throughput, that's key because we're starting to see this in apps. I mean, there's an end user piece. I even seeing it on the infrastructure side where they're taking a heavy lifting away from operational costs. So you got, you know, application specific to the user and/or top of the stack, and then you got actually being used in operations where they want both. >> Yeah, absolutely. Maybe I can illustrate this with a quick story with the customer that we had recently been working with. So this customer is planning to run kind of a transformer based model for tech generation at super high scale on Nvidia T4 GPU, so kind of a commodity GPU. And the scale was so high that they would've been paying hundreds of thousands of dollars in cloud costs per year just to serve this model alone. You know, one of many models in their application stack. So we worked with this team to optimize our model and then benchmark across several possible targets. So that matching the hardware that Luis was just talking about, including the newer kind of Nvidia A10 GPUs. And what they found during this process was pretty interesting. First, the team was able to shave a quarter of their spend just by using better optimization techniques on the T4, the older hardware. But actually moving to a newer GPU would allow them to serve this model in a sub two milliseconds latency, so super fast, which was able to unlock an entirely new kind of user experience. So they were able to kind of change the value they're delivering in their application just because they were able to move to this new hardware easily. So they ultimately decided to plan their deployment on the more expensive A10 because of this, but because of the hardware specific optimizations that we helped them with, they managed to even, you know, bring costs down from what they had originally planned. And so if you extend this kind of example to everything that's happening with generative AI, I think the story we just talked about was super relevant, but the scale can be even higher, you know, it can be tenfold that. We were recently conducting kind of this internal study using GPT-J as a proxy to illustrate the experience of just a company trying to use one of these large language models with an example scenario of creating a chatbot to help job seekers prepare for interviews. So if you imagine kind of a conservative usage scenario where the model generates just 3000 words per user per day, which is, you know, pretty conservative for how people are interacting with these models. It costs 5 cents a session and if you're a company and your app goes viral, so from, you know, beginning of the year there's nobody, at the end of the year there's a million daily active active users in that year alone, going from zero to a million. You'll be spending about $6 million a year, which is pretty unmanageable. That's crazy, right? >> Yeah. >> For a company or a product that's just launching. So I think, you know, for us we see the real way to make these kind of advancements accessible and sustainable, as we said is to bring down cost to serve using these techniques. >> That's a great story and I think that illustrates this idea that deployment cost can vary from situation to situation, from model to model and that the efficiency is so strong with this new wave, it eliminates heavy lifting, creates more efficiency, automates intellect. I mean, this is the trend, this is radical, this is going to increase. So the cost could go from nominal to millions, literally, potentially. So, this is what customers are doing. Yeah, that's a great story. What makes sense on a financial, is there a cost of ownership? Is there a pattern for best practice for training? What do you guys advise cuz this is a lot of time and money involved in all potential, you know, good scenarios of upside. But you can get over your skis as they say, and be successful and be out of business if you don't manage it. I mean, that's what people are talking about, right? >> Yeah, absolutely. I think, you know, we see kind of three main vectors to reduce cost. I think one is make your deployment process easier overall, so that your engineering effort to even get your app running goes down. Two, would be get more from the compute you're already paying for, you're already paying, you know, for your instances in the cloud, but can you do more with that? And then three would be shop around for lower cost hardware to match your use case. So on the first one, I think making the deployment easier overall, there's a lot of manual work that goes into benchmarking, optimizing and packaging models for deployment. And because the performance of machine learning models can be really hardware dependent, you have to go through this process for each target you want to consider running your model on. And this is hard, you know, we see that every day. But for teams who want to incorporate some of these large language models into their applications, it might be desirable because licensing a model from a large vendor like OpenAI can leave you, you know, over provision, kind of paying for capabilities you don't need in your application or can lock you into them and you lose flexibility. So we have a customer whose team actually prepares models for deployment in a SaaS application that many of us use every day. And they told us recently that without kind of an automated benchmarking and experimentation platform, they were spending several days each to benchmark a single model on a single hardware type. So this is really, you know, manually intensive and then getting more from the compute you're already paying for. We do see customers who leave money on the table by running models that haven't been optimized specifically for the hardware target they're using, like Luis was mentioning. And for some teams they just don't have the time to go through an optimization process and for others they might lack kind of specialized expertise and this is something we can bring. And then on shopping around for different hardware types, we really see a huge variation in model performance across hardware, not just CPU vs. GPU, which is, you know, what people normally think of. But across CPU vendors themselves, high memory instances and across cloud providers even. So the best strategy here is for teams to really be able to, we say, look before you leap by running real world benchmarking and not just simulations or predictions to find the best software, hardware combination for their workload. >> Yeah. You guys sound like you have a very impressive customer base deploying large language models. Where would you categorize your current customer base? And as you look out, as you guys are growing, you have new customers coming in, take me through the progression. Take me through the profile of some of your customers you have now, size, are they hyperscalers, are they big app folks, are they kicking the tires? And then as people are out there scratching heads, I got to get in this game, what's their psychology like? Are they coming in with specific problems or do they have specific orientation point of view about what they want to do? Can you share some data around what you're seeing? >> Yeah, I think, you know, we have customers that kind of range across the spectrum of sophistication from teams that basically don't have MLOps expertise in their company at all. And so they're really looking for us to kind of give a full service, how should I do everything from, you know, optimization, find the hardware, prepare for deployment. And then we have teams that, you know, maybe already have their serving and hosting infrastructure up and ready and they already have models in production and they're really just looking to, you know, take the extra juice out of the hardware and just do really specific on that optimization piece. I think one place where we're doing a lot more work now is kind of in the developer tooling, you know, model selection space. And that's kind of an area that we're creating more tools for, particularly within the PyTorch ecosystem to bring kind of this power earlier in the development cycle so that as people are grabbing a model off the shelf, they can, you know, see how it might perform and use that to inform their development process. >> Luis, what's the big, I like this idea of picking the models because isn't that like going to the market and picking the best model for your data? It's like, you know, it's like, isn't there a certain approaches? What's your view on this? 'Cause this is where everyone, I think it's going to be a land rush for this and I want to get your thoughts. >> For sure, yeah. So, you know, I guess I'll start with saying the one main takeaway that we got from the GPT-J study is that, you know, having a different understanding of what your model's compute and memory requirements are, very quickly, early on helps with the much smarter AI model deployments, right? So, and in fact, you know, Anna just touched on this, but I want to, you know, make sure that it's clear that OctoML is putting that power into user's hands right now. So in partnership with AWS, we are launching this new PyTorch native profiler that allows you with a single, you know, one line, you know, code decorator allows you to see how your code runs on a variety of different hardware after accelerations. So it gives you very clear, you know, data on how you should think about your model deployments. And this ties back to choices of models. So like, if you have a set of choices that are equally good of models in terms of functionality and you want to understand after acceleration how are you going to deploy, how much they're going to cost or what are the options using a automated process of making a decision is really, really useful. And in fact, so I think these events can get early access to this by signing up for the Octopods, you know, this is exclusive group for insiders here, so you can go to OctoML.ai/pods to sign up. >> So that Octopod, is that a program? What is that, is that access to code? Is that a beta, what is that? Explain, take a minute and explain Octopod. >> I think the Octopod would be a group of people who is interested in experiencing this functionality. So it is the friends and users of OctoML that would be the Octopod. And then yes, after you sign up, we would provide you essentially the tool in code form for you to try out in your own. I mean, part of the benefit of this is that it happens in your own local environment and you're in control of everything kind of within the workflow that developers are already using to create and begin putting these models into their applications. So it would all be within your control. >> Got it. I think the big question I have for you is when do you, when does that one of your customers know they need to call you? What's their environment look like? What are they struggling with? What are the conversations they might be having on their side of the fence? If anyone's watching this, they're like, "Hey, you know what, I've got my team, we have a lot of data. Do we have our own language model or do I use someone else's?" There's a lot of this, I will say discovery going on around what to do, what path to take, what does that customer look like, if someone's listening, when do they know to call you guys, OctoML? >> Well, I mean the most obvious one is that you have a significant spend on AI/ML, come and talk to us, you know, putting AIML into production. So that's the clear one. In fact, just this morning I was talking to someone who is in life sciences space and is having, you know, 15 to $20 million a year cloud related to AI/ML deployment is a clear, it's a pretty clear match right there, right? So that's on the cost side. But I also want to emphasize something that Anna said earlier that, you know, the hardware and software complexity involved in putting model into production is really high. So we've been able to abstract that away, offering a clean automation flow enables one, to experiment early on, you know, how models would run and get them to production. And then two, once they are into production, gives you an automated flow to continuously updating your model and taking advantage of all this acceleration and ability to run the model on the right hardware. So anyways, let's say one then is cost, you know, you have significant cost and then two, you have an automation needs. And Anna please compliment that. >> Yeah, Anna you can please- >> Yeah, I think that's exactly right. Maybe the other time is when you are expecting a big scale up in serving your application, right? You're launching a new feature, you expect to get a lot of usage or, and you want to kind of anticipate maybe your CTO, your CIO, whoever pays your cloud bills is going to come after you, right? And so they want to know, you know, what's the return on putting this model essentially into my application stack? Am I going to, is the usage going to match what I'm paying for it? And then you can understand that. >> So you guys have a lot of the early adopters, they got big data teams, they're pushed in the production, they want to get a little QA, test the waters, understand, use your technology to figure it out. Is there any cases where people have gone into production, they have to pull it out? It's like the old lemon laws with your car, you buy a car and oh my god, it's not the way I wanted it. I mean, I can imagine the early people through the wall, so to speak, in the wave here are going to be bloody in the sense that they've gone in and tried stuff and get stuck with huge bills. Are you seeing that? Are people pulling stuff out of production and redeploying? Or I can imagine that if I had a bad deployment, I'd want to refactor that or actually replatform that. Do you see that too? >> Definitely after a sticker shock, yes, your customers will come and make sure that, you know, the sticker shock won't happen again. >> Yeah. >> But then there's another more thorough aspect here that I think we likely touched on, be worth elaborating a bit more is just how are you going to scale in a way that's feasible depending on the allocation that you get, right? So as we mentioned several times here, you know, model deployment is so hardware dependent and so complex that you tend to get a model for a hardware choice and then you want to scale that specific type of instance. But what if, when you want to scale because suddenly luckily got popular and, you know, you want to scale it up and then you don't have that instance anymore. So how do you live with whatever you have at that moment is something that we see customers needing as well. You know, so in fact, ideally what we want is customers to not think about what kind of specific instances they want. What they want is to know what their models need. Say, they know the SLA and then find a set of hybrid targets and instances that hit the SLA whenever they're also scaling, they're going to scale with more freedom, right? Instead of having to wait for AWS to give them more specific allocation for a specific instance. What if you could live with other types of hardware and scale up in a more free way, right? So that's another thing that we see customers, you know, like they need more freedom to be able to scale with whatever is available. >> Anna, you touched on this with the business model impact to that 6 million cost, if that goes out of control, there's a business model aspect and there's a technical operation aspect to the cost side too. You want to be mindful of riding the wave in a good way, but not getting over your skis. So that brings up the point around, you know, confidence, right? And teamwork. Because if you're in production, there's probably a team behind it. Talk about the team aspect of your customers. I mean, they're dedicated, they go put stuff into production, they're developers, there're data. What's in it for them? Are they getting better, are they in the beach, you know, reading the book. Are they, you know, are there easy street for them? What's the customer benefit to the teams? >> Yeah, absolutely. With just a few clicks of a button, you're in production, right? That's the dream. So yeah, I mean I think that, you know, we illustrated it before a little bit. I think the automated kind of benchmarking and optimization process, like when you think about the effort it takes to get that data by hand, which is what people are doing today, they just don't do it. So they're making decisions without the best information because it's, you know, there just isn't the bandwidth to get the information that they need to make the best decision and then know exactly how to deploy it. So I think it's actually bringing kind of a new insight and capability to these teams that they didn't have before. And then maybe another aspect on the team side is that it's making the hand-off of the models from the data science teams to the model deployment teams more seamless. So we have, you know, we have seen in the past that this kind of transition point is the place where there are a lot of hiccups, right? The data science team will give a model to the production team and it'll be too slow for the application or it'll be too expensive to run and it has to go back and be changed and kind of this loop. And so, you know, with the PyTorch profiler that Luis was talking about, and then also, you know, the other ways we do optimization that kind of prevents that hand-off problem from happening. >> Luis and Anna, you guys have a great company. Final couple minutes left. Talk about the company, the people there, what's the culture like, you know, if Intel has Moore's law, which is, you know, doubling the performance in few years, what's the culture like there? Is it, you know, more throughput, better pricing? Explain what's going on with the company and put a plug in. Luis, we'll start with you. >> Yeah, absolutely. I'm extremely proud of the team that we built here. You know, we have a people first culture, you know, very, very collaborative and folks, we all have a shared mission here of making AI more accessible and sustainable. We have a very diverse team in terms of backgrounds and life stories, you know, to do what we do here, we need a team that has expertise in software engineering, in machine learning, in computer architecture. Even though we don't build chips, we need to understand how they work, right? So, and then, you know, the fact that we have this, this very really, really varied set of backgrounds makes the environment, you know, it's say very exciting to learn more about, you know, assistance end-to-end. But also makes it for a very interesting, you know, work environment, right? So people have different backgrounds, different stories. Some of them went to grad school, others, you know, were in intelligence agencies and now are working here, you know. So we have a really interesting set of people and, you know, life is too short not to work with interesting humans. You know, that's something that I like to think about, you know. >> I'm sure your off-site meetings are a lot of fun, people talking about computer architectures, silicon advances, the next GPU, the big data models coming in. Anna, what's your take? What's the culture like? What's the company vibe and what are you guys looking to do? What's the customer success pattern? What's up? >> Yeah, absolutely. I mean, I, you know, second all of the great things that Luis just said about the team. I think one that I, an additional one that I'd really like to underscore is kind of this customer obsession, to use a term you all know well. And focus on the end users and really making the experiences that we're bringing to our user who are developers really, you know, useful and valuable for them. And so I think, you know, all of these tools that we're trying to put in the hands of users, the industry and the market is changing so rapidly that our products across the board, you know, all of the companies that, you know, are part of the showcase today, we're all evolving them so quickly and we can only do that kind of really hand in glove with our users. So that would be another thing I'd emphasize. >> I think the change dynamic, the power dynamics of this industry is just the beginning. I'm very bullish that this is going to be probably one of the biggest inflection points in history of the computer industry because of all the dynamics of the confluence of all the forces, which you mentioned some of them, I mean PC, you know, interoperability within internetworking and you got, you know, the web and then mobile. Now we have this, I mean, I wouldn't even put social media even in the close to this. Like, this is like, changes user experience, changes infrastructure. There's going to be massive accelerations in performance on the hardware side from AWS's of the world and cloud and you got the edge and more data. This is really what big data was going to look like. This is the beginning. Final question, what do you guys see going forward in the future? >> Well, it's undeniable that machine learning and AI models are becoming an integral part of an interesting application today, right? So, and the clear trends here are, you know, more and more competitional needs for these models because they're only getting more and more powerful. And then two, you know, seeing the complexity of the infrastructure where they run, you know, just considering the cloud, there's like a wide variety of choices there, right? So being able to live with that and making the most out of it in a way that does not require, you know, an impossible to find team is something that's pretty clear. So the need for automation, abstracting with the complexity is definitely here. And we are seeing this, you know, trends are that you also see models starting to move to the edge as well. So it's clear that we're seeing, we are going to live in a world where there's no large models living in the cloud. And then, you know, edge models that talk to these models in the cloud to form, you know, an end-to-end truly intelligent application. >> Anna? >> Yeah, I think, you know, our, Luis said it at the beginning. Our vision is to make AI sustainable and accessible. And I think as this technology just expands in every company and every team, that's going to happen kind of on its own. And we're here to help support that. And I think you can't do that without tools like those like OctoML. >> I think it's going to be an error of massive invention, creativity, a lot of the format heavy lifting is going to allow the talented people to automate their intellect. I mean, this is really kind of what we see going on. And Luis, thank you so much. Anna, thanks for coming on this segment. Thanks for coming on theCUBE and being part of the AWS Startup Showcase. I'm John Furrier, your host. Thanks for watching. (upbeat music)

(upbeat music) >> Hello everyone. Welcome to theCube's presentation of the "AWS Startup Showcase." The topic this episode is AI and machine learning, top startups building foundational model infrastructure. This is season three, episode one of the ongoing series covering exciting startups from the AWS ecosystem. And this time we're talking about AI and machine learning. I'm your host, John Furrier. I'm excited I'm joined today by Robert Nishihara, who's the co-founder and CEO of a hot startup called Anyscale. He's here to talk about Ray, the open source project, Anyscale's infrastructure for foundation as well. Robert, thank you for joining us today. >> Yeah, thanks so much as well. >> I've been following your company since the founding pre pandemic and you guys really had a great vision scaled up and in a perfect position for this big wave that we all see with ChatGPT and OpenAI that's gone mainstream. Finally, AI has broken out through the ropes and now gone mainstream, so I think you guys are really well positioned. I'm looking forward to to talking with you today. But before we get into it, introduce the core mission for Anyscale. Why do you guys exist? What is the North Star for Anyscale? >> Yeah, like you mentioned, there's a tremendous amount of excitement about AI right now. You know, I think a lot of us believe that AI can transform just every different industry. So one of the things that was clear to us when we started this company was that the amount of compute needed to do AI was just exploding. Like to actually succeed with AI, companies like OpenAI or Google or you know, these companies getting a lot of value from AI, were not just running these machine learning models on their laptops or on a single machine. They were scaling these applications across hundreds or thousands or more machines and GPUs and other resources in the Cloud. And so to actually succeed with AI, and this has been one of the biggest trends in computing, maybe the biggest trend in computing in, you know, in recent history, the amount of compute has been exploding. And so to actually succeed with that AI, to actually build these scalable applications and scale the AI applications, there's a tremendous software engineering lift to build the infrastructure to actually run these scalable applications. And that's very hard to do. So one of the reasons many AI projects and initiatives fail is that, or don't make it to production, is the need for this scale, the infrastructure lift, to actually make it happen. So our goal here with Anyscale and Ray, is to make that easy, is to make scalable computing easy. So that as a developer or as a business, if you want to do AI, if you want to get value out of AI, all you need to know is how to program on your laptop. Like, all you need to know is how to program in Python. And if you can do that, then you're good to go. Then you can do what companies like OpenAI or Google do and get value out of machine learning. >> That programming example of how easy it is with Python reminds me of the early days of Cloud, when infrastructure as code was talked about was, it was just code the infrastructure programmable. That's super important. That's what AI people wanted, first program AI. That's the new trend. And I want to understand, if you don't mind explaining, the relationship that Anyscale has to these foundational models and particular the large language models, also called LLMs, was seen with like OpenAI and ChatGPT. Before you get into the relationship that you have with them, can you explain why the hype around foundational models? Why are people going crazy over foundational models? What is it and why is it so important? >> Yeah, so foundational models and foundation models are incredibly important because they enable businesses and developers to get value out of machine learning, to use machine learning off the shelf with these large models that have been trained on tons of data and that are useful out of the box. And then, of course, you know, as a business or as a developer, you can take those foundational models and repurpose them or fine tune them or adapt them to your specific use case and what you want to achieve. But it's much easier to do that than to train them from scratch. And I think there are three, for people to actually use foundation models, there are three main types of workloads or problems that need to be solved. One is training these foundation models in the first place, like actually creating them. The second is fine tuning them and adapting them to your use case. And the third is serving them and actually deploying them. Okay, so Ray and Anyscale are used for all of these three different workloads. Companies like OpenAI or Cohere that train large language models. Or open source versions like GPTJ are done on top of Ray. There are many startups and other businesses that fine tune, that, you know, don't want to train the large underlying foundation models, but that do want to fine tune them, do want to adapt them to their purposes, and build products around them and serve them, those are also using Ray and Anyscale for that fine tuning and that serving. And so the reason that Ray and Anyscale are important here is that, you know, building and using foundation models requires a huge scale. It requires a lot of data. It requires a lot of compute, GPUs, TPUs, other resources. And to actually take advantage of that and actually build these scalable applications, there's a lot of infrastructure that needs to happen under the hood. And so you can either use Ray and Anyscale to take care of that and manage the infrastructure and solve those infrastructure problems. Or you can build the infrastructure and manage the infrastructure yourself, which you can do, but it's going to slow your team down. It's going to, you know, many of the businesses we work with simply don't want to be in the business of managing infrastructure and building infrastructure. They want to focus on product development and move faster. >> I know you got a keynote presentation we're going to go to in a second, but I think you hit on something I think is the real tipping point, doing it yourself, hard to do. These are things where opportunities are and the Cloud did that with data centers. Turned a data center and made it an API. The heavy lifting went away and went to the Cloud so people could be more creative and build their product. In this case, build their creativity. Is that kind of what's the big deal? Is that kind of a big deal happening that you guys are taking the learnings and making that available so people don't have to do that? >> That's exactly right. So today, if you want to succeed with AI, if you want to use AI in your business, infrastructure work is on the critical path for doing that. To do AI, you have to build infrastructure. You have to figure out how to scale your applications. That's going to change. We're going to get to the point, and you know, with Ray and Anyscale, we're going to remove the infrastructure from the critical path so that as a developer or as a business, all you need to focus on is your application logic, what you want the the program to do, what you want your application to do, how you want the AI to actually interface with the rest of your product. Now the way that will happen is that Ray and Anyscale will still, the infrastructure work will still happen. It'll just be under the hood and taken care of by Ray in Anyscale. And so I think something like this is really necessary for AI to reach its potential, for AI to have the impact and the reach that we think it will, you have to make it easier to do. >> And just for clarification to point out, if you don't mind explaining the relationship of Ray and Anyscale real quick just before we get into the presentation. >> So Ray is an open source project. We created it. We were at Berkeley doing machine learning. We started Ray so that, in order to provide an easy, a simple open source tool for building and running scalable applications. And Anyscale is the managed version of Ray, basically we will run Ray for you in the Cloud, provide a lot of tools around the developer experience and managing the infrastructure and providing more performance and superior infrastructure. >> Awesome. I know you got a presentation on Ray and Anyscale and you guys are positioning as the infrastructure for foundational models. So I'll let you take it away and then when you're done presenting, we'll come back, I'll probably grill you with a few questions and then we'll close it out so take it away. >> Robert: Sounds great. So I'll say a little bit about how companies are using Ray and Anyscale for foundation models. The first thing I want to mention is just why we're doing this in the first place. And the underlying observation, the underlying trend here, and this is a plot from OpenAI, is that the amount of compute needed to do machine learning has been exploding. It's been growing at something like 35 times every 18 months. This is absolutely enormous. And other people have written papers measuring this trend and you get different numbers. But the point is, no matter how you slice and dice it, it' a astronomical rate. Now if you compare that to something we're all familiar with, like Moore's Law, which says that, you know, the processor performance doubles every roughly 18 months, you can see that there's just a tremendous gap between the needs, the compute needs of machine learning applications, and what you can do with a single chip, right. So even if Moore's Law were continuing strong and you know, doing what it used to be doing, even if that were the case, there would still be a tremendous gap between what you can do with the chip and what you need in order to do machine learning. And so given this graph, what we've seen, and what has been clear to us since we started this company, is that doing AI requires scaling. There's no way around it. It's not a nice to have, it's really a requirement. And so that led us to start Ray, which is the open source project that we started to make it easy to build these scalable Python applications and scalable machine learning applications. And since we started the project, it's been adopted by a tremendous number of companies. Companies like OpenAI, which use Ray to train their large models like ChatGPT, companies like Uber, which run all of their deep learning and classical machine learning on top of Ray, companies like Shopify or Spotify or Instacart or Lyft or Netflix, ByteDance, which use Ray for their machine learning infrastructure. Companies like Ant Group, which makes Alipay, you know, they use Ray across the board for fraud detection, for online learning, for detecting money laundering, you know, for graph processing, stream processing. Companies like Amazon, you know, run Ray at a tremendous scale and just petabytes of data every single day. And so the project has seen just enormous adoption since, over the past few years. And one of the most exciting use cases is really providing the infrastructure for building training, fine tuning, and serving foundation models. So I'll say a little bit about, you know, here are some examples of companies using Ray for foundation models. Cohere trains large language models. OpenAI also trains large language models. You can think about the workloads required there are things like supervised pre-training, also reinforcement learning from human feedback. So this is not only the regular supervised learning, but actually more complex reinforcement learning workloads that take human input about what response to a particular question, you know is better than a certain other response. And incorporating that into the learning. There's open source versions as well, like GPTJ also built on top of Ray as well as projects like Alpa coming out of UC Berkeley. So these are some of the examples of exciting projects in organizations, training and creating these large language models and serving them using Ray. Okay, so what actually is Ray? Well, there are two layers to Ray. At the lowest level, there's the core Ray system. This is essentially low level primitives for building scalable Python applications. Things like taking a Python function or a Python class and executing them in the cluster setting. So Ray core is extremely flexible and you can build arbitrary scalable applications on top of Ray. So on top of Ray, on top of the core system, what really gives Ray a lot of its power is this ecosystem of scalable libraries. So on top of the core system you have libraries, scalable libraries for ingesting and pre-processing data, for training your models, for fine tuning those models, for hyper parameter tuning, for doing batch processing and batch inference, for doing model serving and deployment, right. And a lot of the Ray users, the reason they like Ray is that they want to run multiple workloads. They want to train and serve their models, right. They want to load their data and feed that into training. And Ray provides common infrastructure for all of these different workloads. So this is a little overview of what Ray, the different components of Ray. So why do people choose to go with Ray? I think there are three main reasons. The first is the unified nature. The fact that it is common infrastructure for scaling arbitrary workloads, from data ingest to pre-processing to training to inference and serving, right. This also includes the fact that it's future proof. AI is incredibly fast moving. And so many people, many companies that have built their own machine learning infrastructure and standardized on particular workflows for doing machine learning have found that their workflows are too rigid to enable new capabilities. If they want to do reinforcement learning, if they want to use graph neural networks, they don't have a way of doing that with their standard tooling. And so Ray, being future proof and being flexible and general gives them that ability. Another reason people choose Ray in Anyscale is the scalability. This is really our bread and butter. This is the reason, the whole point of Ray, you know, making it easy to go from your laptop to running on thousands of GPUs, making it easy to scale your development workloads and run them in production, making it easy to scale, you know, training to scale data ingest, pre-processing and so on. So scalability and performance, you know, are critical for doing machine learning and that is something that Ray provides out of the box. And lastly, Ray is an open ecosystem. You can run it anywhere. You can run it on any Cloud provider. Google, you know, Google Cloud, AWS, Asure. You can run it on your Kubernetes cluster. You can run it on your laptop. It's extremely portable. And not only that, it's framework agnostic. You can use Ray to scale arbitrary Python workloads. You can use it to scale and it integrates with libraries like TensorFlow or PyTorch or JAX or XG Boost or Hugging Face or PyTorch Lightning, right, or Scikit-learn or just your own arbitrary Python code. It's open source. And in addition to integrating with the rest of the machine learning ecosystem and these machine learning frameworks, you can use Ray along with all of the other tooling in the machine learning ecosystem. That's things like weights and biases or ML flow, right. Or you know, different data platforms like Databricks, you know, Delta Lake or Snowflake or tools for model monitoring for feature stores, all of these integrate with Ray. And that's, you know, Ray provides that kind of flexibility so that you can integrate it into the rest of your workflow. And then Anyscale is the scalable compute platform that's built on top, you know, that provides Ray. So Anyscale is a managed Ray service that runs in the Cloud. And what Anyscale does is it offers the best way to run Ray. And if you think about what you get with Anyscale, there are fundamentally two things. One is about moving faster, accelerating the time to market. And you get that by having the managed service so that as a developer you don't have to worry about managing infrastructure, you don't have to worry about configuring infrastructure. You also, it provides, you know, optimized developer workflows. Things like easily moving from development to production, things like having the observability tooling, the debug ability to actually easily diagnose what's going wrong in a distributed application. So things like the dashboards and the other other kinds of tooling for collaboration, for monitoring and so on. And then on top of that, so that's the first bucket, developer productivity, moving faster, faster experimentation and iteration. The second reason that people choose Anyscale is superior infrastructure. So this is things like, you know, cost deficiency, being able to easily take advantage of spot instances, being able to get higher GPU utilization, things like faster cluster startup times and auto scaling. Things like just overall better performance and faster scheduling. And so these are the kinds of things that Anyscale provides on top of Ray. It's the managed infrastructure. It's fast, it's like the developer productivity and velocity as well as performance. So this is what I wanted to share about Ray in Anyscale. >> John: Awesome. >> Provide that context. But John, I'm curious what you think. >> I love it. I love the, so first of all, it's a platform because that's the platform architecture right there. So just to clarify, this is an Anyscale platform, not- >> That's right. >> Tools. So you got tools in the platform. Okay, that's key. Love that managed service. Just curious, you mentioned Python multiple times, is that because of PyTorch and TensorFlow or Python's the most friendly with machine learning or it's because it's very common amongst all developers? >> That's a great question. Python is the language that people are using to do machine learning. So it's the natural starting point. Now, of course, Ray is actually designed in a language agnostic way and there are companies out there that use Ray to build scalable Java applications. But for the most part right now we're focused on Python and being the best way to build these scalable Python and machine learning applications. But, of course, down the road there always is that potential. >> So if you're slinging Python code out there and you're watching that, you're watching this video, get on Anyscale bus quickly. Also, I just, while you were giving the presentation, I couldn't help, since you mentioned OpenAI, which by the way, congratulations 'cause they've had great scale, I've noticed in their rapid growth 'cause they were the fastest company to the number of users than anyone in the history of the computer industry, so major successor, OpenAI and ChatGPT, huge fan. I'm not a skeptic at all. I think it's just the beginning, so congratulations. But I actually typed into ChatGPT, what are the top three benefits of Anyscale and came up with scalability, flexibility, and ease of use. Obviously, scalability is what you guys are called. >> That's pretty good. >> So that's what they came up with. So they nailed it. Did you have an inside prompt training, buy it there? Only kidding. (Robert laughs) >> Yeah, we hard coded that one. >> But that's the kind of thing that came up really, really quickly if I asked it to write a sales document, it probably will, but this is the future interface. This is why people are getting excited about the foundational models and the large language models because it's allowing the interface with the user, the consumer, to be more human, more natural. And this is clearly will be in every application in the future. >> Absolutely. This is how people are going to interface with software, how they're going to interface with products in the future. It's not just something, you know, not just a chat bot that you talk to. This is going to be how you get things done, right. How you use your web browser or how you use, you know, how you use Photoshop or how you use other products. Like you're not going to spend hours learning all the APIs and how to use them. You're going to talk to it and tell it what you want it to do. And of course, you know, if it doesn't understand it, it's going to ask clarifying questions. You're going to have a conversation and then it'll figure it out. >> This is going to be one of those things, we're going to look back at this time Robert and saying, "Yeah, from that company, that was the beginning of that wave." And just like AWS and Cloud Computing, the folks who got in early really were in position when say the pandemic came. So getting in early is a good thing and that's what everyone's talking about is getting in early and playing around, maybe replatforming or even picking one or few apps to refactor with some staff and managed services. So people are definitely jumping in. So I have to ask you the ROI cost question. You mentioned some of those, Moore's Law versus what's going on in the industry. When you look at that kind of scale, the first thing that jumps out at people is, "Okay, I love it. Let's go play around." But what's it going to cost me? Am I going to be tied to certain GPUs? What's the landscape look like from an operational standpoint, from the customer? Are they locked in and the benefit was flexibility, are you flexible to handle any Cloud? What is the customers, what are they looking at? Basically, that's my question. What's the customer looking at? >> Cost is super important here and many of the companies, I mean, companies are spending a huge amount on their Cloud computing, on AWS, and on doing AI, right. And I think a lot of the advantage of Anyscale, what we can provide here is not only better performance, but cost efficiency. Because if we can run something faster and more efficiently, it can also use less resources and you can lower your Cloud spending, right. We've seen companies go from, you know, 20% GPU utilization with their current setup and the current tools they're using to running on Anyscale and getting more like 95, you know, 100% GPU utilization. That's something like a five x improvement right there. So depending on the kind of application you're running, you know, it's a significant cost savings. We've seen companies that have, you know, processing petabytes of data every single day with Ray going from, you know, getting order of magnitude cost savings by switching from what they were previously doing to running their application on Ray. And when you have applications that are spending, you know, potentially $100 million a year and getting a 10 X cost savings is just absolutely enormous. So these are some of the kinds of- >> Data infrastructure is super important. Again, if the customer, if you're a prospect to this and thinking about going in here, just like the Cloud, you got infrastructure, you got the platform, you got SaaS, same kind of thing's going to go on in AI. So I want to get into that, you know, ROI discussion and some of the impact with your customers that are leveraging the platform. But first I hear you got a demo. >> Robert: Yeah, so let me show you, let me give you a quick run through here. So what I have open here is the Anyscale UI. I've started a little Anyscale Workspace. So Workspaces are the Anyscale concept for interactive developments, right. So here, imagine I'm just, you want to have a familiar experience like you're developing on your laptop. And here I have a terminal. It's not on my laptop. It's actually in the cloud running on Anyscale. And I'm just going to kick this off. This is going to train a large language model, so OPT. And it's doing this on 32 GPUs. We've got a cluster here with a bunch of CPU cores, bunch of memory. And as that's running, and by the way, if I wanted to run this on instead of 32 GPUs, 64, 128, this is just a one line change when I launch the Workspace. And what I can do is I can pull up VS code, right. Remember this is the interactive development experience. I can look at the actual code. Here it's using Ray train to train the torch model. We've got the training loop and we're saying that each worker gets access to one GPU and four CPU cores. And, of course, as I make the model larger, this is using deep speed, as I make the model larger, I could increase the number of GPUs that each worker gets access to, right. And how that is distributed across the cluster. And if I wanted to run on CPUs instead of GPUs or a different, you know, accelerator type, again, this is just a one line change. And here we're using Ray train to train the models, just taking my vanilla PyTorch model using Hugging Face and then scaling that across a bunch of GPUs. And, of course, if I want to look at the dashboard, I can go to the Ray dashboard. There are a bunch of different visualizations I can look at. I can look at the GPU utilization. I can look at, you know, the CPU utilization here where I think we're currently loading the model and running that actual application to start the training. And some of the things that are really convenient here about Anyscale, both I can get that interactive development experience with VS code. You know, I can look at the dashboards. I can monitor what's going on. It feels, I have a terminal, it feels like my laptop, but it's actually running on a large cluster. And I can, with however many GPUs or other resources that I want. And so it's really trying to combine the best of having the familiar experience of programming on your laptop, but with the benefits, you know, being able to take advantage of all the resources in the Cloud to scale. And it's like when, you know, you're talking about cost efficiency. One of the biggest reasons that people waste money, one of the silly reasons for wasting money is just forgetting to turn off your GPUs. And what you can do here is, of course, things will auto terminate if they're idle. But imagine you go to sleep, I have this big cluster. You can turn it off, shut off the cluster, come back tomorrow, restart the Workspace, and you know, your big cluster is back up and all of your code changes are still there. All of your local file edits. It's like you just closed your laptop and came back and opened it up again. And so this is the kind of experience we want to provide for our users. So that's what I wanted to share with you. >> Well, I think that whole, couple of things, lines of code change, single line of code change, that's game changing. And then the cost thing, I mean human error is a big deal. People pass out at their computer. They've been coding all night or they just forget about it. I mean, and then it's just like leaving the lights on or your water running in your house. It's just, at the scale that it is, the numbers will add up. That's a huge deal. So I think, you know, compute back in the old days, there's no compute. Okay, it's just compute sitting there idle. But you know, data cranking the models is doing, that's a big point. >> Another thing I want to add there about cost efficiency is that we make it really easy to use, if you're running on Anyscale, to use spot instances and these preemptable instances that can just be significantly cheaper than the on-demand instances. And so when we see our customers go from what they're doing before to using Anyscale and they go from not using these spot instances 'cause they don't have the infrastructure around it, the fault tolerance to handle the preemption and things like that, to being able to just check a box and use spot instances and save a bunch of money. >> You know, this was my whole, my feature article at Reinvent last year when I met with Adam Selipsky, this next gen Cloud is here. I mean, it's not auto scale, it's infrastructure scale. It's agility. It's flexibility. I think this is where the world needs to go. Almost what DevOps did for Cloud and what you were showing me that demo had this whole SRE vibe. And remember Google had site reliability engines to manage all those servers. This is kind of like an SRE vibe for data at scale. I mean, a similar kind of order of magnitude. I mean, I might be a little bit off base there, but how would you explain it? >> It's a nice analogy. I mean, what we are trying to do here is get to the point where developers don't think about infrastructure. Where developers only think about their application logic. And where businesses can do AI, can succeed with AI, and build these scalable applications, but they don't have to build, you know, an infrastructure team. They don't have to develop that expertise. They don't have to invest years in building their internal machine learning infrastructure. They can just focus on the Python code, on their application logic, and run the stuff out of the box. >> Awesome. Well, I appreciate the time. Before we wrap up here, give a plug for the company. I know you got a couple websites. Again, go, Ray's got its own website. You got Anyscale. You got an event coming up. Give a plug for the company looking to hire. Put a plug in for the company. >> Yeah, absolutely. Thank you. So first of all, you know, we think AI is really going to transform every industry and the opportunity is there, right. We can be the infrastructure that enables all of that to happen, that makes it easy for companies to succeed with AI, and get value out of AI. Now we have, if you're interested in learning more about Ray, Ray has been emerging as the standard way to build scalable applications. Our adoption has been exploding. I mentioned companies like OpenAI using Ray to train their models. But really across the board companies like Netflix and Cruise and Instacart and Lyft and Uber, you know, just among tech companies. It's across every industry. You know, gaming companies, agriculture, you know, farming, robotics, drug discovery, you know, FinTech, we see it across the board. And all of these companies can get value out of AI, can really use AI to improve their businesses. So if you're interested in learning more about Ray and Anyscale, we have our Ray Summit coming up in September. This is going to highlight a lot of the most impressive use cases and stories across the industry. And if your business, if you want to use LLMs, you want to train these LLMs, these large language models, you want to fine tune them with your data, you want to deploy them, serve them, and build applications and products around them, give us a call, talk to us. You know, we can really take the infrastructure piece, you know, off the critical path and make that easy for you. So that's what I would say. And, you know, like you mentioned, we're hiring across the board, you know, engineering, product, go-to-market, and it's an exciting time. >> Robert Nishihara, co-founder and CEO of Anyscale, congratulations on a great company you've built and continuing to iterate on and you got growth ahead of you, you got a tailwind. I mean, the AI wave is here. I think OpenAI and ChatGPT, a customer of yours, have really opened up the mainstream visibility into this new generation of applications, user interface, roll of data, large scale, how to make that programmable so we're going to need that infrastructure. So thanks for coming on this season three, episode one of the ongoing series of the hot startups. In this case, this episode is the top startups building foundational model infrastructure for AI and ML. I'm John Furrier, your host. Thanks for watching. (upbeat music)

(gentle upbeat music) >> Hey, friends of theCUBE. It's Lisa Martin live at Stanford University covering the Eighth Annual Women In Data Science. But you've been a Cube fan for a long time. So you know that we've been here since the beginning of WiDS, which is 2015. We always loved to come and cover this event. We learned great things about data science, about women leaders, underrepresented minorities. And this year we have a special component. We've got two grad students from Stanford's Master's program and Data Journalism joining. One of my them is here with me, Hannah Freitag, my co-host. Great to have you. And we are pleased to welcome from Intuit for the first time, Shir Meir Lador Group Manager at Data Science. Shir, it's great to have you. Thank you for joining us. >> Thank you for having me. >> And I was just secrets girl talking with my boss of theCUBE who informed me that you're in great company. Intuit's Chief Technology Officer, Marianna Tessel is an alumni of theCUBE. She was on at our Supercloud event in January. So welcome back into it. >> Thank you very much. We're happy to be with you. >> Tell us a little bit about what you're doing. You're a data science group manager as I mentioned, but also you've had you've done some cool things I want to share with the audience. You're the co-founder of the PyData Tel Aviv Meetups the co-host of the unsupervised podcast about data science in Israel. You give talks, about machine learning, about data science. Tell us a little bit about your background. Were you always interested in STEM studies from the time you were small? >> So I was always interested in mathematics when I was small, I went to this special program for youth going to university. So I did my test in mathematics earlier and studied in university some courses. And that's when I understood I want to do something in that field. And then when I got to go to university, I went to electrical engineering when I found out about algorithms and how interested it is to be able to find solutions to problems, to difficult problems with math. And this is how I found my way into machine learning. >> Very cool. There's so much, we love talking about machine learning and AI on theCUBE. There's so much potential. Of course, we have to have data. One of the things that I love about WiDS and Hannah and I and our co-host Tracy, have been talking about this all day is the impact of data in everyone's life. If you break it down, I was at Mobile World Congress last week, all about connectivity telecom, and of course we have these expectation that we're going to be connected 24/7 from wherever we are in the world and we can do whatever we want. I can do an Uber transaction, I can watch Netflix, I can do a bank transaction. It all is powered by data. And data science is, some of the great applications of it is what it's being applied to. Things like climate change or police violence or health inequities. Talk about some of the data science projects that you're working on at Intuit. I'm an intuit user myself, but talk to me about some of those things. Give the audience really a feel for what you're doing. >> So if you are a Intuit product user, you probably use TurboTax. >> I do >> In the past. So for those who are not familiar, TurboTax help customers submit their taxes. Basically my group is in charge of getting all the information automatically from your documents, the documents that you upload to TurboTax. We extract that information to accelerate your tax submission to make it less work for our customers. So- >> Thank you. >> Yeah, and this is why I'm so proud to be working at this team because our focus is really to help our customers to simplify all the you know, financial heavy lifting with taxes and also with small businesses. We also do a lot of work in extracting information from small business documents like bill, receipts, different bank statements. Yeah, so this is really exciting for me, the opportunity to work to apply data science and machine learning to solution that actually help people. Yeah >> Yeah, in the past years there have been more and more digital products emerging that needs some sort of data security. And how did your team, or has your team developed in the past years with more and more products or companies offering digital services? >> Yeah, so can you clarify the question again? Sorry. >> Yeah, have you seen that you have more customers? Like has your team expanded in the past years with more digital companies starting that need kind of data security? >> Well, definitely. I think, you know, since I joined Intuit, I joined like five and a half years ago back when I was in Tel Aviv. I recently moved to the Bay Area. So when I joined, there were like a dozens of data scientists and machine learning engineers on Intuit. And now there are a few hundreds. So we've definitely grown with the year and there are so many new places we can apply machine learning to help our customers. So this is amazing, so much we can do with machine learning to get more money in the pocket of our customers and make them do less work. >> I like both of those. More money in my pocket and less work. That's awesome. >> Exactly. >> So keep going Intuit. But one of the things that is so cool is just the the abstraction of the complexity that Intuit's doing. I upload documents or it scans my receipts. I was just in Barcelona last week all these receipts and conversion euros to dollars and it takes that complexity away from the end user who doesn't know all that's going on in the background, but you're making people's lives simpler. Unfortunately, we all have to pay taxes, most of us should. And of course we're in tax season right now. And so it's really cool what you're doing with ML and data science to make fundamental processes to people's lives easier and just a little bit less complicated. >> Definitely. And I think that's what's also really amazing about Intuit it, is how it combines human in the loop as well as AI. Because in some of the tax situation it's very complicated maybe to do it yourself. And then there's an option to work with an expert online that goes on a video with you and helps you do your taxes. And the expert's work is also accelerated by AI because we build tools for those experts to do the work more efficiently. >> And that's what it's all about is you know, using data to be more efficient, to be faster, to be smarter, but also to make complicated processes in our daily lives, in our business lives just a little bit easier. One of the things I've been geeking out about recently is ChatGPT. I was using it yesterday. I was telling everyone I was asking it what's hot in data science and I didn't know would it know what hot is and it did, it gave me trends. But one of the things that I was so, and Hannah knows I've been telling this all day, I was so excited to learn over the weekend that the the CTO of OpenAI is a female. I didn't know that. And I thought why are we not putting her on a pedestal? Because people are likening ChatGPT to like the launch of the iPhone. I mean revolutionary. And here we have what I think is exciting for all of us females, whether you're in tech or not, is another role model. Because really ultimately what WiDS is great at doing is showcasing women in technical roles. Because I always say you can't be what you can't see. We need to be able to see more role models, female role role models, underrepresented minorities of course men, because a lot of my sponsors and mentors are men, but we need more women that we can look up to and see ah, she's doing this, why can't I? Talk to me about how you stay the course in data science. What excites you about the potential, the opportunities based on what you've already accomplished what inspires you to continue and be one of those females that we say oh my God, I could be like Shir. >> I think that what inspires me the most is the endless opportunities that we have. I think we haven't even started tapping into everything that we can do with generative AI, for example. There's so much that can be done to further help you know, people make more money and do less work because there's still so much work that we do that we don't need to. You know, this is with Intuit, but also there are so many other use cases like I heard today you know, with the talk about the police. So that was really exciting how you can apply machine learning and data to actually help people, to help people that been through wrongful things. So I was really moved by that. And I'm also really excited about all the medical applications that we can have with data. >> Yeah, yeah. It's true that data science is so diverse in terms of what fields it can cover but it's equally important to have diverse teams and have like equity and inclusion in your teams. Where is Intuit at promoting women, non-binary minorities in your teams to progress data science? >> Yeah, so I have so much to say on this. >> Good. >> But in my work in Tel Aviv, I had the opportunity to start with Intuit women in data science branch in Tel Aviv. So that's why I'm super excited to be here today for that because basically this is the original conference, but as you know, there are branches all over the world and I got the opportunity to lead the Tel Aviv branch with Israel since 2018. And we've been through already this year it's going to be it's next week, it's going to be the sixth conference. And every year our number of submission to make talk in the conference doubled itself. >> Nice. >> We started with 20 submission, then 50, then 100. This year we have over 200 submissions of females to give talk at the conference. >> Ah, that's fantastic. >> And beyond the fact that there's so much traction, I also feel the great impact it has on the community in Israel because one of the reason we started WiDS was that when I was going to conferences I was seeing so little women on stage in all the technical conferences. You know, kind of the reason why I guess you know, Margaret and team started the WiDS conference. So I saw the same thing in Israel and I was always frustrated. I was organizing PyData Meetups as you mentioned and I was always having such a hard time to get female speakers to talk. I was trying to role model, but that's not enough, you know. We need more. So once we started WiDS and people saw you know, so many examples on the stage and also you know females got opportunity to talk in a place for that. Then it also started spreading and you can see more and more female speakers across other conferences, which are not women in data science. So I think just the fact that Intuits started this conference back in Israel and also in Bangalore and also the support Intuit does for WiDS in Stanford here, it shows how much WiDS values are aligned with our values. Yeah, and I think that to chauffeur that I think we have over 35% females in the data science and machine learning engineering roles, which is pretty amazing I think compared to the industry. >> Way above average. Yeah, absolutely. I was just, we've been talking about some of the AnitaB.org stats from 2022 showing that 'cause usually if we look at the industry to you point, over the last, I don't know, probably five, 10 years we're seeing the number of female technologists around like a quarter, 25% or so. 2022 data from AnitaB.org showed that that number is now 27.6%. So it's very slowly- >> It's very slowly increasing. >> Going in the right direction. >> Too slow. >> And that representation of women technologists increase at every level, except intern, which I thought was really interesting. And I wonder is there a covid relation there? >> I don't know. >> What do we need to do to start opening up the the top of the pipeline, the funnel to go downstream to find kids like you when you were younger and always interested in engineering and things like that. But the good news is that the hiring we've seen improvements, but it sounds like Intuit is way ahead of the curve there with 35% women in data science or technical roles. And what's always nice and refreshing that we've talked, Hannah about this too is seeing companies actually put action into initiatives. It's one thing for a company to say we're going to have you know, 50% females in our organization by 2030. It's a whole other ball game to actually create a strategy, execute on it, and share progress. So kudos to Intuit for what it's doing because that is more companies need to adopt that same sort of philosophy. And that's really cultural. >> Yeah. >> At an organization and culture can be hard to change, but it sounds like you guys kind of have it dialed in. >> I think we definitely do. That's why I really like working and Intuit. And I think that a lot of it is with the role modeling, diversity and inclusion, and by having women leaders. When you see a woman in leadership position, as a woman it makes you want to come work at this place. And as an evidence, when I build the team I started in Israel at Intuit, I have over 50% women in my team. >> Nice. >> Yeah, because when you have a woman in the interviewers panel, it's much easier, it's more inclusive. That's why we always try to have at least you know, one woman and also other minorities represented in our interviews panel. Yeah, and I think that in general it's very important as a leader to kind of know your own biases and trying to have defined standard and rubrics in how you evaluate people to avoid for those biases. So all of that inclusiveness and leadership really helps to get more diversity in your teams. >> It's critical. That thought diversity is so critical, especially if we talk about AI and we're almost out of time, I just wanted to bring up, you brought up a great point about the diversity and equity. With respect to data science and AI, we know in AI there's biases in data. We need to have more inclusivity, more representation to help start shifting that so the biases start to be dialed down and I think a conference like WiDS and it sounds like someone like you and what you've already done so far in the work that you're doing having so many females raise their hands to want to do talks at events is a good situation. It's a good scenario and hopefully it will continue to move the needle on the percentage of females in technical roles. So we thank you Shir for your time sharing with us your story, what you're doing, how Intuit and WiDS are working together. It sounds like there's great alignment there and I think we're at the tip of the iceberg with what we can do with data science and inclusion and equity. So we appreciate all of your insights and your time. >> Thank you very much. >> All right. >> I enjoyed very, very much >> Good. We hope, we aim to please. Thank you for our guests and for Hannah Freitag. This is Lisa Martin coming to you live from Stanford University. This is our coverage of the eighth Annual Women in Data Science Conference. Stick around, next guest will be here in just a minute.

(gentle music) >> Hey! Welcome back to theCUBE's coverage of WiDS 2023, the eighth Annual Women In Data Science Conference. I'm your host, Lisa Martin. We are at Stanford University, as you know we are every year, having some wonderful conversations with some very inspiring women and men in data science and technical roles. I'm very pleased to introduce Tracy Zhang, my co-host, who is in the Data Journalism program at Stanford. And Tracy and I are pleased to welcome our next guest, Rhonda Crate, Principal Data Scientist at Boeing. Great to have you on the program, Rhonda. >> Tracy: Welcome. >> Hey, thanks for having me. >> Were you always interested in data science or STEM from the time you were young? >> No, actually. I was always interested in archeology and anthropology. >> That's right, we were talking about that, anthropology. Interesting. >> We saw the anthropology background, not even a bachelor's degree, but also a master's degree in anthropology. >> So you were committed for a while. >> I was, I was. I actually started college as a fine arts major, but I always wanted to be an archeologist. So at the last minute, 11 credits in, left to switch to anthropology. And then when I did my master's, I focused a little bit more on quantitative research methods and then I got my Stat Degree. >> Interesting. Talk about some of the data science projects that you're working on. When I think of Boeing, I always think of aircraft. But you are doing a lot of really cool things in IT, data analytics. Talk about some of those intriguing data science projects that you're working on. >> Yeah. So when I first started at Boeing, I worked in information technology and data analytics. And Boeing, at the time, had cored up data science in there. And so we worked as a function across the enterprise working on anything from shared services to user experience in IT products, to airplane programs. So, it has a wide range. I worked on environment health and safety projects for a long time as well. So looking at ergonomics and how people actually put parts onto airplanes, along with things like scheduling and production line, part failures, software testing. Yeah, there's a wide spectrum of things. >> But I think that's so fantastic. We've been talking, Tracy, today about just what we often see at WiDS, which is this breadth of diversity in people's background. You talked about anthropology, archeology, you're doing data science. But also all of the different opportunities that you've had at Boeing. To see so many facets of that organization. I always think that breadth of thought diversity can be hugely impactful. >> Yeah. So I will say my anthropology degree has actually worked to my benefit. I'm a huge proponent of integrating liberal arts and sciences together. And it actually helps me. I'm in the Technical Fellowship program at Boeing, so we have different career paths. So you can go into management, you can be a regular employee, or you can go into the Fellowship program. So right now I'm an Associate Technical Fellow. And part of how I got into the Fellowship program was that diversity in my background, what made me different, what made me stand out on projects. Even applying a human aspect to things like ergonomics, as silly as that sounds, but how does a person actually interact in the space along with, here are the actual measurements coming off of whatever system it is that you're working on. So, I think there's a lot of opportunities, especially in safety as well, which is a big initiative for Boeing right now, as you can imagine. >> Tracy: Yeah, definitely. >> I can't go into too specifics. >> No, 'cause we were like, I think a theme for today that kind of we brought up in in all of our talk is how data is about people, how data is about how people understand the world and how these data can make impact on people's lives. So yeah, I think it's great that you brought this up, and I'm very happy that your anthropology background can tap into that and help in your day-to-day data work too. >> Yeah. And currently, right now, I actually switched over to Strategic Workforce Planning. So it's more how we understand our workforce, how we work towards retaining the talent, how do we get the right talent in our space, and making sure overall that we offer a culture and work environment that is great for our employees to come to. >> That culture is so important. You know, I was looking at some anitab.org stats from 2022 and you know, we always talk about the number of women in technical roles. For a long time it's been hovering around that 25% range. The data from anitab.org showed from '22, it's now 27.6%. So, a little increase. But one of the biggest challenges still, and Tracy and I and our other co-host, Hannah, have been talking about this, is attrition. Attrition more than doubled last year. What are some of the things that Boeing is doing on the retention side, because that is so important especially as, you know, there's this pipeline leakage of women leaving technical roles. Tell us about what Boeing's, how they're invested. >> Yeah, sure. We actually have a publicly available Global Diversity Report that anybody can go and look at and see our statistics for our organization. Right now, off the top of my head, I think we're hovering at about 24% in the US for women in our company. It has been a male majority company for many years. We've invested heavily in increasing the number of women in roles. One interesting thing about this year that came out is that even though with the great resignation and those types of things, the attrition level between men and women were actually pretty close to being equal, which is like the first time in our history. Usually it tends on more women leaving. >> Lisa: That's a good sign. >> Right. >> Yes, that's a good sign. >> And we've actually focused on hiring and bringing in more women and diversity in our company. >> Yeah, some of the stats too from anitab.org talked about the increase, and I have to scroll back and find my notes, the increase in 51% more women being hired in 2022 than 2021 for technical roles. So the data, pun intended, is showing us. I mean, the data is there to show the impact that having females in executive leadership positions make from a revenue perspective. >> Tracy: Definitely. >> Companies are more profitable when there's women at the head, or at least in senior leadership roles. But we're seeing some positive trends, especially in terms of representation of women technologists. One of the things though that I found interesting, and I'm curious to get your thoughts on this, Rhonda, is that the representation of women technologists is growing in all areas, except interns. >> Rhonda: Hmm. >> So I think, we've got to go downstream. You teach, I have to go back to my notes on you, did my due diligence, R programming classes through Boeings Ed Wells program, this is for WSU College of Arts and Sciences, talk about what you teach and how do you think that intern kind of glut could be solved? >> Yeah. So, they're actually two separate programs. So I teach a data analytics course at Washington State University as an Adjunct Professor. And then the Ed Wells program is a SPEEA, which is an Aerospace Union, focused on bringing up more technology and skills to the actual workforce itself. So it's kind of a couple different audiences. One is more seasoned employees, right? The other one is our undergraduates. I teach a Capstone class, so it's a great way to introduce students to what it's actually like to work on an industry project. We partner with Google and Microsoft and Boeing on those. The idea is also that maybe those companies have openings for the students when they're done. Since it's Senior Capstone, there's not a lot of opportunities for internships. But the opportunities to actually get hired increase a little bit. In regards to Boeing, we've actually invested a lot in hiring more women interns. I think the number was 40%, but you'd have to double check. >> Lisa: That's great, that's fantastic. >> Tracy: That's way above average, I think. >> That's a good point. Yeah, it is above average. >> Double check on that. That's all from my memory. >> Is this your first WiDS, or have you been before? >> I did virtually last year. >> Okay. One of the things that I love, I love covering this event every year. theCUBE's been covering it since it's inception in 2015. But it's just the inspiration, the vibe here at Stanford is so positive. WiDS is a movement. It's not an initiative, an organization. There are going to be, I think annually this year, there will be 200 different events. Obviously today we're live on International Women's Day. 60 plus countries, 100,000 plus people involved. So, this is such a positive environment for women and men, because we need everybody, underrepresented minorities, to be able to understand the implication that data has across our lives. If we think about stripping away titles in industries, everybody is a consumer, not everybody, most of mobile devices. And we have this expectation, I was in Barcelona last week at a Mobile World Congress, we have this expectation that we're going to be connected 24/7. I can get whatever I want wherever I am in the world, and that's all data driven. And the average person that isn't involved in data science wouldn't understand that. At the same time, they have expectations that depend on organizations like Boeing being data driven so that they can get that experience that they expect in their consumer lives in any aspect of their lives. And that's one of the things I find so interesting and inspiring about data science. What are some of the things that keep you motivated to continue pursuing this? >> Yeah I will say along those lines, I think it's great to invest in K-12 programs for Data Literacy. I know one of my mentors and directors of the Data Analytics program, Dr. Nairanjana Dasgupta, we're really familiar with each other. So, she runs a WSU program for K-12 Data Literacy. It's also something that we strive for at Boeing, and we have an internal Data Literacy program because, believe it or not, most people are in business. And there's a lot of disconnect between interpreting and understanding data. For me, what kind of drives me to continue data science is that connection between people and data and how we use it to improve our world, which is partly why I work at Boeing too 'cause I feel that they produce products that people need like satellites and airplanes, >> Absolutely. >> and everything. >> Well, it's tangible, it's relatable. We can understand it. Can you do me a quick favor and define data literacy for anyone that might not understand what that means? >> Yeah, so it's just being able to understand elements of data, whether that's a bar chart or even in a sentence, like how to read a statistic and interpret a statistic in a sentence, for example. >> Very cool. >> Yeah. And sounds like Boeing's doing a great job in these programs, and also trying to hire more women. So yeah, I wanted to ask, do you think there's something that Boeing needs to work on? Or where do you see yourself working on say the next five years? >> Yeah, I think as a company, we always think that there's always room for improvement. >> It never, never stops. >> Tracy: Definitely. (laughs) >> I know workforce strategy is an area that they're currently really heavily investing in, along with safety. How do we build safer products for people? How do we help inform the public about things like Covid transmission in airports? For example, we had the Confident Traveler Initiative which was a big push that we had, and we had to be able to inform people about data models around Covid, right? So yeah, I would say our future is more about an investment in our people and in our culture from my perspective >> That's so important. One of the hardest things to change especially for a legacy organization like Boeing, is culture. You know, when I talk with CEO's or CIO's or COO's about what's your company's vision, what's your strategy? Especially those companies that are on that digital journey that have no choice these days. Everybody expects to have a digital experience, whether you're transacting an an Uber ride, you're buying groceries, or you're traveling by air. That culture sounds like Boeing is really focused on that. And that's impressive because that's one of the hardest things to morph and mold, but it's so essential. You know, as we look around the room here at WiDS it's obviously mostly females, but we're talking about women, underrepresented minorities. We're talking about men as well who are mentors and sponsors to us. I'd love to get your advice to your younger self. What would you tell yourself in terms of where you are now to become a leader in the technology field? >> Yeah, I mean, it's kind of an interesting question because I always try to think, live with no regrets to an extent. >> Lisa: I like that. >> But, there's lots of failures along the way. (Tracy laughing) I don't know if I would tell myself anything different because honestly, if I did, I wouldn't be where I am. >> Lisa: Good for you. >> I started out in fine arts, and I didn't end up there. >> That's good. >> Such a good point, yeah. >> We've been talking about that and I find that a lot at events like WiDS, is women have these zigzaggy patterns. I studied biology, I have a master's in molecular biology, I'm in media and marketing. We talked about transportable skills. There's a case I made many years ago when I got into tech about, well in science you learn the art of interpreting esoteric data and creating a story from it. And that's a transportable skill. But I always say, you mentioned failure, I always say failure is not a bad F word. It allows us to kind of zig and zag and learn along the way. And I think that really fosters thought diversity. And in data science, that is one of the things we absolutely need to have is that diversity and thought. You know, we talk about AI models being biased, we need the data and we need the diverse brains to help ensure that the biases are identified, extracted, and removed. Speaking of AI, I've been geeking out with ChatGPT. So, I'm on it yesterday and I ask it, "What's hot in data science?" And I was like, is it going to get that? What's hot? And it did it, it came back with trends. I think if I ask anything, "What's hot?", I should be to Paris Hilton, but I didn't. And so I was geeking out. One of the things I learned recently that I thought was so super cool is the CTO of OpenAI is a woman, Mira Murati, which I didn't know until over the weekend. Because I always think if I had to name top females in tech, who would they be? And I always default to Sheryl Sandberg, Carly Fiorina, Susan Wojcicki running YouTube. Who are some of the people in your history, in your current, that are really inspiring to you? Men, women, indifferent. >> Sure. I think Boeing is one of the companies where you actually do see a lot of women in leadership roles. I think we're one of the top companies with a number of women executives, actually. Susan Doniz, who's our Chief Information Officer, I believe she's actually slotted to speak at a WiDS event come fall. >> Lisa: Cool. >> So that will be exciting. Susan's actually relatively newer to Boeing in some ways. A Boeing time skill is like three years is still kind of new. (laughs) But she's been around for a while and she's done a lot of inspiring things, I think, for women in the organization. She does a lot with Latino communities and things like that as well. For me personally, you know, when I started at Boeing Ahmad Yaghoobi was one of my mentors and my Technical Lead. He came from Iran during a lot of hard times in the 1980s. His brother actually wrote a memoir, (laughs) which is just a fun, interesting fact. >> Tracy: Oh my God! >> Lisa: Wow! >> And so, I kind of gravitate to people that I can learn from that's not in my sphere, that might make me uncomfortable. >> And you probably don't even think about how many people you're influencing along the way. >> No. >> We just keep going and learning from our mentors and probably lose sight of, "I wonder how many people actually admire me?" And I'm sure there are many that admire you, Rhonda, for what you've done, going from anthropology to archeology. You mentioned before we went live you were really interested in photography. Keep going and really gathering all that breadth 'cause it's only making you more inspiring to people like us. >> Exactly. >> We thank you so much for joining us on the program and sharing a little bit about you and what brought you to WiDS. Thank you so much, Rhonda. >> Yeah, thank you. >> Tracy: Thank you so much for being here. >> Lisa: Yeah. >> Alright. >> For our guests, and for Tracy Zhang, this is Lisa Martin live at Stanford University covering the eighth Annual Women In Data Science Conference. Stick around. Next guest will be here in just a second. (gentle music)

(gentle music) >> Hello everyone. Welcome to this CUBE conversation here in Palo Alto, California. I'm John Furrier, host of theCUBE. We got a great remote guest coming in. Joseph Nelson, co-founder and CEO of RoboFlow hot startup in AI, computer vision. Really interesting topic in this wave of AI next gen hitting. Joseph, thanks for coming on this CUBE conversation. >> Thanks for having me. >> Yeah, I love the startup tsunami that's happening here in this wave. RoboFlow, you're in the middle of it. Exciting opportunities, you guys are in the cutting edge. I think computer vision's been talked about more as just as much as the large language models and these foundational models are merging. You're in the middle of it. What's it like right now as a startup and growing in this new wave hitting? >> It's kind of funny, it's, you know, I kind of describe it like sometimes you're in a garden of gnomes. It's like we feel like we've got this giant headstart with hundreds of thousands of people building with computer vision, training their own models, but that's a fraction of what it's going to be in six months, 12 months, 24 months. So, as you described it, a wave is a good way to think about it. And the wave is still building before it gets to its full size. So it's a ton of fun. >> Yeah, I think it's one of the most exciting areas in computer science. I wish I was in my twenties again, because I would be all over this. It's the intersection, there's so many disciplines, right? It's not just tech computer science, it's computer science, it's systems, it's software, it's data. There's so much aperture of things going on around your world. So, I mean, you got to be batting all the students away kind of trying to get hired in there, probably. I can only imagine you're hiring regiment. I'll ask that later, but first talk about what the company is that you're doing. How it's positioned, what's the market you're going after, and what's the origination story? How did you guys get here? How did you just say, hey, want to do this? What was the origination story? What do you do and how did you start the company? >> Yeah, yeah. I'll give you the what we do today and then I'll shift into the origin. RoboFlow builds tools for making the world programmable. Like anything that you see should be read write access if you think about it with a programmer's mind or legible. And computer vision is a technology that enables software to be added to these real world objects that we see. And so any sort of interface, any sort of object, any sort of scene, we can interact with it, we can make it more efficient, we can make it more entertaining by adding the ability for the tools that we use and the software that we write to understand those objects. And at RoboFlow, we've empowered a little over a hundred thousand developers, including those in half the Fortune 100 so far in that mission. Whether that's Walmart understanding the retail in their stores, Cardinal Health understanding the ways that they're helping their patients, or even electric vehicle manufacturers ensuring that they're making the right stuff at the right time. As you mentioned, it's early. Like I think maybe computer vision has touched one, maybe 2% of the whole economy and it'll be like everything in a very short period of time. And so we're focused on enabling that transformation. I think it's it, as far as I think about it, I've been fortunate to start companies before, start, sell these sorts of things. This is the last company I ever wanted to start and I think it will be, should we do it right, the world's largest in riding the wave of bringing together the disparate pieces of that technology. >> What was the motivating point of the formation? Was it, you know, you guys were hanging around? Was there some catalyst? What was the moment where it all kind of came together for you? >> You know what's funny is my co-founder, Brad and I, we were making computer vision apps for making board games more fun to play. So in 2017, Apple released AR kit, augmented reality kit for building augmented reality applications. And Brad and I are both sort of like hacker persona types. We feel like we don't really understand the technology until we build something with it and so we decided that we should make an app that if you point your phone at a Sudoku puzzle, it understands the state of the board and then it kind of magically fills in that experience with all the digits in real time, which totally ruins the game of Sudoku to be clear. But it also just creates this like aha moment of like, oh wow, like the ability for our pocket devices to understand and see the world as good or better than we can is possible. And so, you know, we actually did that as I mentioned in 2017, and the app went viral. It was, you know, top of some subreddits, top of Injure, Reddit, the hacker community as well as Product Hunt really liked it. So it actually won Product Hunt AR app of the year, which was the same year that the Tesla model three won the product of the year. So we joked that we share an award with Elon our shared (indistinct) But frankly, so that was 2017. RoboFlow wasn't incorporated as a business until 2019. And so, you know, when we made Magic Sudoku, I was running a different company at the time, Brad was running a different company at the time, and we kind of just put it out there and were excited by how many people liked it. And we assumed that other curious developers would see this inevitable future of, oh wow, you know. This is much more than just a pedestrian point your phone at a board game. This is everything can be seen and understood and rewritten in a different way. Things like, you know, maybe your fridge. Knowing what ingredients you have and suggesting recipes or auto ordering for you, or we were talking about some retail use cases of automated checkout. Like anything can be seen and observed and we presume that that would kick off a Cambrian explosion of applications. It didn't. So you fast forward to 2019, we said, well we might as well be the guys to start to tackle this sort of problem. And because of our success with board games before, we returned to making more board game solving applications. So we made one that solves Boggle, you know, the four by four word game, we made one that solves chess, you point your phone at a chess board and it understands the state of the board and then can make move recommendations. And each additional board game that we added, we realized that the tooling was really immature. The process of collecting images, knowing which images are actually going to be useful for improving model performance, training those models, deploying those models. And if we really wanted to make the world programmable, developers waiting for us to make an app for their thing of interest is a lot less efficient, less impactful than taking our tool chain and releasing that externally. And so, that's what RoboFlow became. RoboFlow became the internal tools that we used to make these game changing applications readily available. And as you know, when you give developers new tools, they create new billion dollar industries, let alone all sorts of fun hobbyist projects along the way. >> I love that story. Curious, inventive, little radical. Let's break the rules, see how we can push the envelope on the board games. That's how companies get started. It's a great story. I got to ask you, okay, what happens next? Now, okay, you realize this new tooling, but this is like how companies get built. Like they solve their own problem that they had 'cause they realized there's one, but then there has to be a market for it. So you actually guys knew that this was coming around the corner. So okay, you got your hacker mentality, you did that thing, you got the award and now you're like, okay, wow. Were you guys conscious of the wave coming? Was it one of those things where you said, look, if we do this, we solve our own problem, this will be big for everybody. Did you have that moment? Was that in 2019 or was that more of like, it kind of was obvious to you guys? >> Absolutely. I mean Brad puts this pretty effectively where he describes how we lived through the initial internet revolution, but we were kind of too young to really recognize and comprehend what was happening at the time. And then mobile happened and we were working on different companies that were not in the mobile space. And computer vision feels like the wave that we've caught. Like, this is a technology and capability that rewrites how we interact with the world, how everyone will interact with the world. And so we feel we've been kind of lucky this time, right place, right time of every enterprise will have the ability to improve their operations with computer vision. And so we've been very cognizant of the fact that computer vision is one of those groundbreaking technologies that every company will have as a part of their products and services and offerings, and we can provide the tooling to accelerate that future. >> Yeah, and the developer angle, by the way, I love that because I think, you know, as we've been saying in theCUBE all the time, developer's the new defacto standard bodies because what they adopt is pure, you know, meritocracy. And they pick the best. If it's sell service and it's good and it's got open source community around it, its all in. And they'll vote. They'll vote with their code and that is clear. Now I got to ask you, as you look at the market, we were just having this conversation on theCUBE in Barcelona at recent Mobile World Congress, now called MWC, around 5G versus wifi. And the debate was specifically computer vision, like facial recognition. We were talking about how the Cleveland Browns were using facial recognition for people coming into the stadium they were using it for ships in international ports. So the question was 5G versus wifi. My question is what infrastructure or what are the areas that need to be in place to make computer vision work? If you have developers building apps, apps got to run on stuff. So how do you sort that out in your mind? What's your reaction to that? >> A lot of the times when we see applications that need to run in real time and on video, they'll actually run at the edge without internet. And so a lot of our users will actually take their models and run it in a fully offline environment. Now to act on that information, you'll often need to have internet signal at some point 'cause you'll need to know how many people were in the stadium or what shipping crates are in my port at this point in time. You'll need to relay that information somewhere else, which will require connectivity. But actually using the model and creating the insights at the edge does not require internet. I mean we have users that deploy models on underwater submarines just as much as in outer space actually. And those are not very friendly environments to internet, let alone 5g. And so what you do is you use an edge device, like an Nvidia Jetson is common, mobile devices are common. Intel has some strong edge devices, the Movidius family of chips for example. And you use that compute that runs completely offline in real time to process those signals. Now again, what you do with those signals may require connectivity and that becomes a question of the problem you're solving of how soon you need to relay that information to another place. >> So, that's an architectural issue on the infrastructure. If you're a tactical edge war fighter for instance, you might want to have highly available and maybe high availability. I mean, these are words that mean something. You got storage, but it's not at the edge in real time. But you can trickle it back and pull it down. That's management. So that's more of a business by business decision or environment, right? >> That's right, that's right. Yeah. So I mean we can talk through some specifics. So for example, the RoboFlow actually powers the broadcaster that does the tennis ball tracking at Wimbledon. That runs completely at the edge in real time in, you know, technically to track the tennis ball and point the camera, you actually don't need internet. Now they do have internet of course to do the broadcasting and relay the signal and feeds and these sorts of things. And so that's a case where you have both edge deployment of running the model and high availability act on that model. We have other instances where customers will run their models on drones and the drone will go and do a flight and it'll say, you know, this many residential homes are in this given area, or this many cargo containers are in this given shipping yard. Or maybe we saw these environmental considerations of soil erosion along this riverbank. The model in that case can run on the drone during flight without internet, but then you only need internet once the drone lands and you're going to act on that information because for example, if you're doing like a study of soil erosion, you don't need to be real time. You just need to be able to process and make use of that information once the drone finishes its flight. >> Well I can imagine a zillion use cases. I heard of a use case interview at a company that does computer vision to help people see if anyone's jumping the fence on their company. Like, they know what a body looks like climbing a fence and they can spot it. Pretty easy use case compared to probably some of the other things, but this is the horizontal use cases, its so many use cases. So how do you guys talk to the marketplace when you say, hey, we have generative AI for commuter vision. You might know language models that's completely different animal because vision's like the world, right? So you got a lot more to do. What's the difference? How do you explain that to customers? What can I build and what's their reaction? >> Because we're such a developer centric company, developers are usually creative and show you the ways that they want to take advantage of new technologies. I mean, we've had people use things for identifying conveyor belt debris, doing gas leak detection, measuring the size of fish, airplane maintenance. We even had someone that like a hobby use case where they did like a specific sushi identifier. I dunno if you know this, but there's a specific type of whitefish that if you grew up in the western hemisphere and you eat it in the eastern hemisphere, you get very sick. And so there was someone that made an app that tells you if you happen to have that fish in the sushi that you're eating. But security camera analysis, transportation flows, plant disease detection, really, you know, smarter cities. We have people that are doing curb management identifying, and a lot of these use cases, the fantastic thing about building tools for developers is they're a creative bunch and they have these ideas that if you and I sat down for 15 minutes and said, let's guess every way computer vision can be used, we would need weeks to list all the example use cases. >> We'd miss everything. >> And we'd miss. And so having the community show us the ways that they're using computer vision is impactful. Now that said, there are of course commercial industries that have discovered the value and been able to be out of the gate. And that's where we have the Fortune 100 customers, like we do. Like the retail customers in the Walmart sector, healthcare providers like Medtronic, or vehicle manufacturers like Rivian who all have very difficult either supply chain, quality assurance, in stock, out of stock, anti-theft protection considerations that require successfully making sense of the real world. >> Let me ask you a question. This is maybe a little bit in the weeds, but it's more developer focused. What are some of the developer profiles that you're seeing right now in terms of low-hanging fruit applications? And can you talk about the academic impact? Because I imagine if I was in school right now, I'd be all over it. Are you seeing Master's thesis' being worked on with some of your stuff? Is the uptake in both areas of younger pre-graduates? And then inside the workforce, What are some of the devs like? Can you share just either what their makeup is, what they work on, give a little insight into the devs you're working with. >> Leading developers that want to be on state-of-the-art technology build with RoboFlow because they know they can use the best in class open source. They know that they can get the most out of their data. They know that they can deploy extremely quickly. That's true among students as you mentioned, just as much as as industries. So we welcome students and I mean, we have research grants that will regularly support for people to publish. I mean we actually have a channel inside our internal slack where every day, more student publications that cite building with RoboFlow pop up. And so, that helps inspire some of the use cases. Now what's interesting is that the use case is relatively, you know, useful or applicable for the business or the student. In other words, if a student does a thesis on how to do, we'll say like shingle damage detection from satellite imagery and they're just doing that as a master's thesis, in fact most insurance businesses would be interested in that sort of application. So, that's kind of how we see uptick and adoption both among researchers who want to be on the cutting edge and publish, both with RoboFlow and making use of open source tools in tandem with the tool that we provide, just as much as industry. And you know, I'm a big believer in the philosophy that kind of like what the hackers are doing nights and weekends, the Fortune 500 are doing in a pretty short order period of time and we're experiencing that transition. Computer vision used to be, you know, kind of like a PhD, multi-year investment endeavor. And now with some of the tooling that we're working on in open source technologies and the compute that's available, these science fiction ideas are possible in an afternoon. And so you have this idea of maybe doing asset management or the aerial observation of your shingles or things like this. You have a few hundred images and you can de-risk whether that's possible for your business today. So there's pretty broad-based adoption among both researchers that want to be on the state of the art, as much as companies that want to reduce the time to value. >> You know, Joseph, you guys and your partner have got a great front row seat, ground floor, presented creation wave here. I'm seeing a pattern emerging from all my conversations on theCUBE with founders that are successful, like yourselves, that there's two kind of real things going on. You got the enterprises grabbing the products and retrofitting into their legacy and rebuilding their business. And then you have startups coming out of the woodwork. Young, seeing greenfield or pick a specific niche or focus and making that the signature lever to move the market. >> That's right. >> So can you share your thoughts on the startup scene, other founders out there and talk about that? And then I have a couple questions for like the enterprises, the old school, the existing legacy. Little slower, but the startups are moving fast. What are some of the things you're seeing as startups are emerging in this field? >> I think you make a great point that independent of RoboFlow, very successful, especially developer focused businesses, kind of have three customer types. You have the startups and maybe like series A, series B startups that you're building a product as fast as you can to keep up with them, and they're really moving just as fast as as you are and pulling the product out at you for things that they need. The second segment that you have might be, call it SMB but not enterprise, who are able to purchase and aren't, you know, as fast of moving, but are stable and getting value and able to get to production. And then the third type is enterprise, and that's where you have typically larger contract value sizes, slower moving in terms of adoption and feedback for your product. And I think what you see is that successful companies balance having those three customer personas because you have the small startups, small fast moving upstarts that are discerning buyers who know the market and elect to build on tooling that is best in class. And so you basically kind of pass the smell test of companies who are quite discerning in their purchases, plus are moving so quick they're pulling their product out of you. Concurrently, you have a product that's enterprise ready to service the scalability, availability, and trust of enterprise buyers. And that's ultimately where a lot of companies will see tremendous commercial success. I mean I remember seeing the Twilio IPO, Uber being like a full 20% of their revenue, right? And so there's this very common pattern where you have the ability to find some of those upstarts that you make bets on, like the next Ubers of the world, the smaller companies that continue to get developed with the product and then the enterprise whom allows you to really fund the commercial success of the business, and validate the size of the opportunity in market that's being creative. >> It's interesting, there's so many things happening there. It's like, in a way it's a new category, but it's not a new category. It becomes a new category because of the capabilities, right? So, it's really interesting, 'cause that's what you're talking about is a category, creating. >> I think developer tools. So people often talk about B to B and B to C businesses. I think developer tools are in some ways a third way. I mean ultimately they're B to B, you're selling to other businesses and that's where your revenue's coming from. However, you look kind of like a B to C company in the ways that you measure product adoption and kind of go to market. In other words, you know, we're often tracking the leading indicators of commercial success in the form of usage, adoption, retention. Really consumer app, traditionally based metrics of how to know you're building the right stuff, and that's what product led growth companies do. And then you ultimately have commercial traction in a B to B way. And I think that that actually kind of looks like a third thing, right? Like you can do these sort of funny zany marketing examples that you might see historically from consumer businesses, but yet you ultimately make your money from the enterprise who has these de-risked high value problems you can solve for them. And I selfishly think that that's the best of both worlds because I don't have to be like Evan Spiegel, guessing the next consumer trend or maybe creating the next consumer trend and catching lightning in a bottle over and over again on the consumer side. But I still get to have fun in our marketing and make sort of fun, like we're launching the world's largest game of rock paper scissors being played with computer vision, right? Like that's sort of like a fun thing you can do, but then you can concurrently have the commercial validation and customers telling you the things that they need to be built for them next to solve commercial pain points for them. So I really do think that you're right by calling this a new category and it really is the best of both worlds. >> It's a great call out, it's a great call out. In fact, I always juggle with the VC. I'm like, it's so easy. Your job is so easy to pick the winners. What are you talking about its so easy? I go, just watch what the developers jump on. And it's not about who started, it could be someone in the dorm room to the boardroom person. You don't know because that B to C, the C, it's B to D you know? You know it's developer 'cause that's a human right? That's a consumer of the tool which influences the business that never was there before. So I think this direct business model evolution, whether it's media going direct or going direct to the developers rather than going to a gatekeeper, this is the reality. >> That's right. >> Well I got to ask you while we got some time left to describe, I want to get into this topic of multi-modality, okay? And can you describe what that means in computer vision? And what's the state of the growth of that portion of this piece? >> Multi modality refers to using multiple traditionally siloed problem types, meaning text, image, video, audio. So you could treat an audio problem as only processing audio signal. That is not multimodal, but you could use the audio signal at the same time as a video feed. Now you're talking about multi modality. In computer vision, multi modality is predominantly happening with images and text. And one of the biggest releases in this space is actually two years old now, was clip, contrastive language image pre-training, which took 400 million image text pairs and basically instead of previously when you do classification, you basically map every single image to a single class, right? Like here's a bunch of images of chairs, here's a bunch of images of dogs. What clip did is used, you can think about it like, the class for an image being the Instagram caption for the image. So it's not one single thing. And by training on understanding the corpora, you basically see which words, which concepts are associated with which pixels. And this opens up the aperture for the types of problems and generalizability of models. So what does this mean? This means that you can get to value more quickly from an existing trained model, or at least validate that what you want to tackle with a computer vision, you can get there more quickly. It also opens up the, I mean. Clip has been the bedrock of some of the generative image techniques that have come to bear, just as much as some of the LLMs. And increasingly we're going to see more and more of multi modality being a theme simply because at its core, you're including more context into what you're trying to understand about the world. I mean, in its most basic sense, you could ask yourself, if I have an image, can I know more about that image with just the pixels? Or if I have the image and the sound of when that image was captured or it had someone describe what they see in that image when the image was captured, which one's going to be able to get you more signal? And so multi modality helps expand the ability for us to understand signal processing. >> Awesome. And can you just real quick, define clip for the folks that don't know what that means? >> Yeah. Clip is a model architecture, it's an acronym for contrastive language image pre-training and like, you know, model architectures that have come before it captures the almost like, models are kind of like brands. So I guess it's a brand of a model where you've done these 400 million image text pairs to match up which visual concepts are associated with which text concepts. And there have been new releases of clip, just at bigger sizes of bigger encoding's, of longer strings of texture, or larger image windows. But it's been a really exciting advancement that OpenAI released in January, 2021. >> All right, well great stuff. We got a couple minutes left. Just I want to get into more of a company-specific question around culture. All startups have, you know, some sort of cultural vibe. You know, Intel has Moore's law doubles every whatever, six months. What's your culture like at RoboFlow? I mean, if you had to describe that culture, obviously love the hacking story, you and your partner with the games going number one on Product Hunt next to Elon and Tesla and then hey, we should start a company two years later. That's kind of like a curious, inventing, building, hard charging, but laid back. That's my take. How would you describe the culture? >> I think that you're right. The culture that we have is one of shipping, making things. So every week each team shares what they did for our customers on a weekly basis. And we have such a strong emphasis on being better week over week that those sorts of things compound. So one big emphasis in our culture is getting things done, shipping, doing things for our customers. The second is we're an incredibly transparent place to work. For example, how we think about giving decisions, where we're progressing against our goals, what problems are biggest and most important for the company is all open information for those that are inside the company to know and progress against. The third thing that I'd use to describe our culture is one that thrives with autonomy. So RoboFlow has a number of individuals who have founded companies before, some of which have sold their businesses for a hundred million plus upon exit. And the way that we've been able to attract talent like that is because the problems that we're tackling are so immense, yet individuals are able to charge at it with the way that they think is best. And this is what pairs well with transparency. If you have a strong sense of what the company's goals are, how we're progressing against it, and you have this ownership mentality of what can I do to change or drive progress against that given outcome, then you create a really healthy pairing of, okay cool, here's where the company's progressing. Here's where things are going really well, here's the places that we most need to improve and work on. And if you're inside that company as someone who has a preponderance to be a self-starter and even a history of building entire functions or companies yourself, then you're going to be a place where you can really thrive. You have the inputs of the things where we need to work on to progress the company's goals. And you have the background of someone that is just necessarily a fast moving and ambitious type of individual. So I think the best way to describe it is a transparent place with autonomy and an emphasis on getting things done. >> Getting shit done as they say. Getting stuff done. Great stuff. Hey, final question. Put a plug out there for the company. What are you going to hire? What's your pipeline look like for people? What jobs are open? I'm sure you got hiring all around. Give a quick plug for the company what you're looking for. >> I appreciate you asking. Basically you're either building the product or helping customers be successful with the product. So in the building product category, we have platform engineering roles, machine learning engineering roles, and we're solving some of the hardest and most impactful problems of bringing such a groundbreaking technology to the masses. And so it's a great place to be where you can kind of be your own user as an engineer. And then if you're enabling people to be successful with the products, I mean you're working in a place where there's already such a strong community around it and you can help shape, foster, cultivate, activate, and drive commercial success in that community. So those are roles that tend themselves to being those that build the product for developer advocacy, those that are account executives that are enabling our customers to realize commercial success, and even hybrid roles like we call it field engineering, where you are a technical resource to drive success within customer accounts. And so all this is listed on roboflow.com/careers. And one thing that I actually kind of want to mention John that's kind of novel about the thing that's working at RoboFlow. So there's been a lot of discussion around remote companies and there's been a lot of discussion around in-person companies and do you need to be in the office? And one thing that we've kind of recognized is you can actually chart a third way. You can create a third way which we call satellite, which basically means people can work from where they most like to work and there's clusters of people, regular onsite's. And at RoboFlow everyone gets, for example, $2,500 a year that they can use to spend on visiting coworkers. And so what's sort of organically happened is team numbers have started to pull together these resources and rent out like, lavish Airbnbs for like a week and then everyone kind of like descends in and works together for a week and makes and creates things. And we call this lighthouses because you know, a lighthouse kind of brings ships into harbor and we have an emphasis on shipping. >> Yeah, quality people that are creative and doers and builders. You give 'em some cash and let the self-governing begin, you know? And like, creativity goes through the roof. It's a great story. I think that sums up the culture right there, Joseph. Thanks for sharing that and thanks for this great conversation. I really appreciate it and it's very inspiring. Thanks for coming on. >> Yeah, thanks for having me, John. >> Joseph Nelson, co-founder and CEO of RoboFlow. Hot company, great culture in the right place in a hot area, computer vision. This is going to explode in value. The edge is exploding. More use cases, more development, and developers are driving the change. Check out RoboFlow. This is theCUBE. I'm John Furrier, your host. Thanks for watching. (gentle music)

(bright upbeat music) >> Hello and welcome to this Cube Conversation. I'm John Furrier, host of theCUBE. We've got a great conversation featuring Arthur AI. I'm your host. I'm excited to have Adam Wenchel who's the Co-Founder and CEO. Thanks for joining us today, appreciate it. >> Yeah, thanks for having me on, John, looking forward to the conversation. >> I got to say, it's been an exciting world in AI or artificial intelligence. Just an explosion of interest kind of in the mainstream with the language models, which people don't really get, but they're seeing the benefits of some of the hype around OpenAI. Which kind of wakes everyone up to, "Oh, I get it now." And then of course the pessimism comes in, all the skeptics are out there. But this breakthrough in generative AI field is just awesome, it's really a shift, it's a wave. We've been calling it probably the biggest inflection point, then the others combined of what this can do from a surge standpoint, applications. I mean, all aspects of what we used to know is the computing industry, software industry, hardware, is completely going to get turbo. So we're totally obviously bullish on this thing. So, this is really interesting. So my first question is, I got to ask you, what's you guys taking? 'Cause you've been doing this, you're in it, and now all of a sudden you're at the beach where the big waves are. What's the explosion of interest is there? What are you seeing right now? >> Yeah, I mean, it's amazing, so for starters, I've been in AI for over 20 years and just seeing this amount of excitement and the growth, and like you said, the inflection point we've hit in the last six months has just been amazing. And, you know, what we're seeing is like people are getting applications into production using LLMs. I mean, really all this excitement just started a few months ago, with ChatGPT and other breakthroughs and the amount of activity and the amount of new systems that we're seeing hitting production already so soon after that is just unlike anything we've ever seen. So it's pretty awesome. And, you know, these language models are just, they could be applied in so many different business contexts and that it's just the amount of value that's being created is again, like unprecedented compared to anything. >> Adam, you know, you've been in this for a while, so it's an interesting point you're bringing up, and this is a good point. I was talking with my friend John Markoff, former New York Times journalist and he was talking about, there's been a lot of work been done on ethics. So there's been, it's not like it's new. It's like been, there's a lot of stuff that's been baking over many, many years and, you know, decades. So now everyone wakes up in the season, so I think that is a key point I want to get into some of your observations. But before we get into it, I want you to explain for the folks watching, just so we can kind of get a definition on the record. What's an LLM, what's a foundational model and what's generative ai? Can you just quickly explain the three things there? >> Yeah, absolutely. So an LLM or a large language model, it's just a large, they would imply a large language model that's been trained on a huge amount of data typically pulled from the internet. And it's a general purpose language model that can be built on top for all sorts of different things, that includes traditional NLP tasks like document classification and sentiment understanding. But the thing that's gotten people really excited is it's used for generative tasks. So, you know, asking it to summarize documents or asking it to answer questions. And these aren't new techniques, they've been around for a while, but what's changed is just this new class of models that's based on new architectures. They're just so much more capable that they've gone from sort of science projects to something that's actually incredibly useful in the real world. And there's a number of companies that are making them accessible to everyone so that you can build on top of them. So that's the other big thing is, this kind of access to these models that can power generative tasks has been democratized in the last few months and it's just opening up all these new possibilities. And then the third one you mentioned foundation models is sort of a broader term for the category that includes LLMs, but it's not just language models that are included. So we've actually seen this for a while in the computer vision world. So people have been building on top of computer vision models, pre-trained computer vision models for a while for image classification, object detection, that's something we've had customers doing for three or four years already. And so, you know, like you said, there are antecedents to like, everything that's happened, it's not entirely new, but it does feel like a step change. >> Yeah, I did ask ChatGPT to give me a riveting introduction to you and it gave me an interesting read. If we have time, I'll read it. It's kind of, it's fun, you get a kick out of it. "Ladies and gentlemen, today we're a privileged "to have Adam Wenchel, Founder of Arthur who's going to talk "about the exciting world of artificial intelligence." And then it goes on with some really riveting sentences. So if we have time, I'll share that, it's kind of funny. It was good. >> Okay. >> So anyway, this is what people see and this is why I think it's exciting 'cause I think people are going to start refactoring what they do. And I've been saying this on theCUBE now for about a couple months is that, you know, there's a scene in "Moneyball" where Billy Beane sits down with the Red Sox owner and the Red Sox owner says, "If people aren't rebuilding their teams on your model, "they're going to be dinosaurs." And it reminds me of what's happening right now. And I think everyone that I talk to in the business sphere is looking at this and they're connecting the dots and just saying, if we don't rebuild our business with this new wave, they're going to be out of business because there's so much efficiency, there's so much automation, not like DevOps automation, but like the generative tasks that will free up the intellect of people. Like just the simple things like do an intro or do this for me, write some code, write a countermeasure to a hack. I mean, this is kind of what people are doing. And you mentioned computer vision, again, another huge field where 5G things are coming on, it's going to accelerate. What do you say to people when they kind of are leaning towards that, I need to rethink my business? >> Yeah, it's 100% accurate and what's been amazing to watch the last few months is the speed at which, and the urgency that companies like Microsoft and Google or others are actually racing to, to do that rethinking of their business. And you know, those teams, those companies which are large and haven't always been the fastest moving companies are working around the clock. And the pace at which they're rolling out LLMs across their suite of products is just phenomenal to watch. And it's not just the big, the large tech companies as well, I mean, we're seeing the number of startups, like we get, every week a couple of new startups get in touch with us for help with their LLMs and you know, there's just a huge amount of venture capital flowing into it right now because everyone realizes the opportunities for transforming like legal and healthcare and content creation in all these different areas is just wide open. And so there's a massive gold rush going on right now, which is amazing. >> And the cloud scale, obviously horizontal scalability of the cloud brings us to another level. We've been seeing data infrastructure since the Hadoop days where big data was coined. Now you're seeing this kind of take fruit, now you have vertical specialization where data shines, large language models all of a set up perfectly for kind of this piece. And you know, as you mentioned, you've been doing it for a long time. Let's take a step back and I want to get into how you started the company, what drove you to start it? Because you know, as an entrepreneur you're probably saw this opportunity before other people like, "Hey, this is finally it, it's here." Can you share the origination story of what you guys came up with, how you started it, what was the motivation and take us through that origination story. >> Yeah, absolutely. So as I mentioned, I've been doing AI for many years. I started my career at DARPA, but it wasn't really until 2015, 2016, my previous company was acquired by Capital One. Then I started working there and shortly after I joined, I was asked to start their AI team and scale it up. And for the first time I was actually doing it, had production models that we were working with, that was at scale, right? And so there was hundreds of millions of dollars of business revenue and certainly a big group of customers who were impacted by the way these models acted. And so it got me hyper-aware of these issues of when you get models into production, it, you know. So I think people who are earlier in the AI maturity look at that as a finish line, but it's really just the beginning and there's this constant drive to make them better, make sure they're not degrading, make sure you can explain what they're doing, if they're impacting people, making sure they're not biased. And so at that time, there really weren't any tools to exist to do this, there wasn't open source, there wasn't anything. And so after a few years there, I really started talking to other people in the industry and there was a really clear theme that this needed to be addressed. And so, I joined with my Co-Founder John Dickerson, who was on the faculty in University of Maryland and he'd been doing a lot of research in these areas. And so we ended up joining up together and starting Arthur. >> Awesome. Well, let's get into what you guys do. Can you explain the value proposition? What are people using you for now? Where's the action? What's the customers look like? What do prospects look like? Obviously you mentioned production, this has been the theme. It's not like people woke up one day and said, "Hey, I'm going to put stuff into production." This has kind of been happening. There's been companies that have been doing this at scale and then yet there's a whole follower model coming on mainstream enterprise and businesses. So there's kind of the early adopters are there now in production. What do you guys do? I mean, 'cause I think about just driving the car off the lot is not, you got to manage operations. I mean, that's a big thing. So what do you guys do? Talk about the value proposition and how you guys make money? >> Yeah, so what we do is, listen, when you go to validate ahead of deploying these models in production, starts at that point, right? So you want to make sure that if you're going to be upgrading a model, if you're going to replacing one that's currently in production, that you've proven that it's going to perform well, that it's going to be perform ethically and that you can explain what it's doing. And then when you launch it into production, traditionally data scientists would spend 25, 30% of their time just manually checking in on their model day-to-day babysitting as we call it, just to make sure that the data hasn't drifted, the model performance hasn't degraded, that a programmer did make a change in an upstream data system. You know, there's all sorts of reasons why the world changes and that can have a real adverse effect on these models. And so what we do is bring the same kind of automation that you have for other kinds of, let's say infrastructure monitoring, application monitoring, we bring that to your AI systems. And that way if there ever is an issue, it's not like weeks or months till you find it and you find it before it has an effect on your P&L and your balance sheet, which is too often before they had tools like Arthur, that was the way they were detected. >> You know, I was talking to Swami at Amazon who I've known for a long time for 13 years and been on theCUBE multiple times and you know, I watched Amazon try to pick up that sting with stage maker about six years ago and so much has happened since then. And he and I were talking about this wave, and I kind of brought up this analogy to how when cloud started, it was, Hey, I don't need a data center. 'Cause when I did my startup that time when Amazon, one of my startups at that time, my choice was put a box in the colo, get all the configuration before I could write over the line of code. So the cloud became the benefit for that and you can stand up stuff quickly and then it grew from there. Here it's kind of the same dynamic, you don't want to have to provision a large language model or do all this heavy lifting. So that seeing companies coming out there saying, you can get started faster, there's like a new way to get it going. So it's kind of like the same vibe of limiting that heavy lifting. >> Absolutely. >> How do you look at that because this seems to be a wave that's going to be coming in and how do you guys help companies who are going to move quickly and start developing? >> Yeah, so I think in the race to this kind of gold rush mentality, race to get these models into production, there's starting to see more sort of examples and evidence that there are a lot of risks that go along with it. Either your model says things, your system says things that are just wrong, you know, whether it's hallucination or just making things up, there's lots of examples. If you go on Twitter and the news, you can read about those, as well as sort of times when there could be toxic content coming out of things like that. And so there's a lot of risks there that you need to think about and be thoughtful about when you're deploying these systems. But you know, you need to balance that with the business imperative of getting these things into production and really transforming your business. And so that's where we help people, we say go ahead, put them in production, but just make sure you have the right guardrails in place so that you can do it in a smart way that's going to reflect well on you and your company. >> Let's frame the challenge for the companies now that you have, obviously there's the people who doing large scale production and then you have companies maybe like as small as us who have large linguistic databases or transcripts for example, right? So what are customers doing and why are they deploying AI right now? And is it a speed game, is it a cost game? Why have some companies been able to deploy AI at such faster rates than others? And what's a best practice to onboard new customers? >> Yeah, absolutely. So I mean, we're seeing across a bunch of different verticals, there are leaders who have really kind of started to solve this puzzle about getting AI models into production quickly and being able to iterate on them quickly. And I think those are the ones that realize that imperative that you mentioned earlier about how transformational this technology is. And you know, a lot of times, even like the CEOs or the boards are very personally kind of driving this sense of urgency around it. And so, you know, that creates a lot of movement, right? And so those companies have put in place really smart infrastructure and rails so that people can, data scientists aren't encumbered by having to like hunt down data, get access to it. They're not encumbered by having to stand up new platforms every time they want to deploy an AI system, but that stuff is already in place. There's a really nice ecosystem of products out there, including Arthur, that you can tap into. Compared to five or six years ago when I was building at a top 10 US bank, at that point you really had to build almost everything yourself and that's not the case now. And so it's really nice to have things like, you know, you mentioned AWS SageMaker and a whole host of other tools that can really accelerate things. >> What's your profile customer? Is it someone who already has a team or can people who are learning just dial into the service? What's the persona? What's the pitch, if you will, how do you align with that customer value proposition? Do people have to be built out with a team and in play or is it pre-production or can you start with people who are just getting going? >> Yeah, people do start using it pre-production for validation, but I think a lot of our customers do have a team going and they're starting to put, either close to putting something into production or about to, it's everything from large enterprises that have really sort of complicated, they have dozens of models running all over doing all sorts of use cases to tech startups that are very focused on a single problem, but that's like the lifeblood of the company and so they need to guarantee that it works well. And you know, we make it really easy to get started, especially if you're using one of the common model development platforms, you can just kind of turn key, get going and make sure that you have a nice feedback loop. So then when your models are out there, it's pointing out, areas where it's performing well, areas where it's performing less well, giving you that feedback so that you can make improvements, whether it's in training data or futurization work or algorithm selection. There's a number of, you know, depending on the symptoms, there's a number of things you can do to increase performance over time and we help guide people on that journey. >> So Adam, I have to ask, since you have such a great customer base and they're smart and they got teams and you're on the front end, I mean, early adopters is kind of an overused word, but they're killing it. They're putting stuff in the production's, not like it's a test, it's not like it's early. So as the next wave comes of fast followers, how do you see that coming online? What's your vision for that? How do you see companies that are like just waking up out of the frozen, you know, freeze of like old IT to like, okay, they got cloud, but they're not yet there. What do you see in the market? I see you're in the front end now with the top people really nailing AI and working hard. What's the- >> Yeah, I think a lot of these tools are becoming, or every year they get easier, more accessible, easier to use. And so, you know, even for that kind of like, as the market broadens, it takes less and less of a lift to put these systems in place. And the thing is, every business is unique, they have their own kind of data and so you can use these foundation models which have just been trained on generic data. They're a great starting point, a great accelerant, but then, in most cases you're either going to want to create a model or fine tune a model using data that's really kind of comes from your particular customers, the people you serve and so that it really reflects that and takes that into account. And so I do think that these, like the size of that market is expanding and its broadening as these tools just become easier to use and also the knowledge about how to build these systems becomes more widespread. >> Talk about your customer base you have now, what's the makeup, what size are they? Give a taste a little bit of a customer base you got there, what's they look like? I'll say Capital One, we know very well while you were at there, they were large scale, lot of data from fraud detection to all kinds of cool stuff. What do your customers now look like? >> Yeah, so we have a variety, but I would say one area we're really strong, we have several of the top 10 US banks, that's not surprising, that's a strength for us, but we also have Fortune 100 customers in healthcare, in manufacturing, in retail, in semiconductor and electronics. So what we find is like in any sort of these major verticals, there's typically, you know, one, two, three kind of companies that are really leading the charge and are the ones that, you know, in our opinion, those are the ones that for the next multiple decades are going to be the leaders, the ones that really kind of lead the charge on this AI transformation. And so we're very fortunate to be working with some of those. And then we have a number of startups as well who we love working with just because they're really pushing the boundaries technologically and so they provide great feedback and make sure that we're continuing to innovate and staying abreast of everything that's going on. >> You know, these early markups, even when the hyperscalers were coming online, they had to build everything themselves. That's the new, they're like the alphas out there building it. This is going to be a big wave again as that fast follower comes in. And so when you look at the scale, what advice would you give folks out there right now who want to tee it up and what's your secret sauce that will help them get there? >> Yeah, I think that the secret to teeing it up is just dive in and start like the, I think these are, there's not really a secret. I think it's amazing how accessible these are. I mean, there's all sorts of ways to access LLMs either via either API access or downloadable in some cases. And so, you know, go ahead and get started. And then our secret sauce really is the way that we provide that performance analysis of what's going on, right? So we can tell you in a very actionable way, like, hey, here's where your model is doing good things, here's where it's doing bad things. Here's something you want to take a look at, here's some potential remedies for it. We can help guide you through that. And that way when you're putting it out there, A, you're avoiding a lot of the common pitfalls that people see and B, you're able to really kind of make it better in a much faster way with that tight feedback loop. >> It's interesting, we've been kind of riffing on this supercloud idea because it was just different name than multicloud and you see apps like Snowflake built on top of AWS without even spending any CapEx, you just ride that cloud wave. This next AI, super AI wave is coming. I don't want to call AIOps because I think there's a different distinction. If you, MLOps and AIOps seem a little bit old, almost a few years back, how do you view that because everyone's is like, "Is this AIOps?" And like, "No, not kind of, but not really." How would you, you know, when someone says, just shoots off the hip, "Hey Adam, aren't you doing AIOps?" Do you say, yes we are, do you say, yes, but we do differently because it's doesn't seem like it's the same old AIOps. What's your- >> Yeah, it's a good question. AIOps has been a term that was co-opted for other things and MLOps also has people have used it for different meanings. So I like the term just AI infrastructure, I think it kind of like describes it really well and succinctly. >> But you guys are doing the ops. I mean that's the kind of ironic thing, it's like the next level, it's like NextGen ops, but it's not, you don't want to be put in that bucket. >> Yeah, no, it's very operationally focused platform that we have, I mean, it fires alerts, people can action off them. If you're familiar with like the way people run security operations centers or network operations centers, we do that for data science, right? So think of it as a DSOC, a Data Science Operations Center where all your models, you might have hundreds of models running across your organization, you may have five, but as problems are detected, alerts can be fired and you can actually work the case, make sure they're resolved, escalate them as necessary. And so there is a very strong operational aspect to it, you're right. >> You know, one of the things I think is interesting is, is that, if you don't mind commenting on it, is that the aspect of scale is huge and it feels like that was made up and now you have scale and production. What's your reaction to that when people say, how does scale impact this? >> Yeah, scale is huge for some of, you know, I think, I think look, the highest leverage business areas to apply these to, are generally going to be the ones at the biggest scale, right? And I think that's one of the advantages we have. Several of us come from enterprise backgrounds and we're used to doing things enterprise grade at scale and so, you know, we're seeing more and more companies, I think they started out deploying AI and sort of, you know, important but not necessarily like the crown jewel area of their business, but now they're deploying AI right in the heart of things and yeah, the scale that some of our companies are operating at is pretty impressive. >> John: Well, super exciting, great to have you on and congratulations. I got a final question for you, just random. What are you most excited about right now? Because I mean, you got to be pretty pumped right now with the way the world is going and again, I think this is just the beginning. What's your personal view? How do you feel right now? >> Yeah, the thing I'm really excited about for the next couple years now, you touched on it a little bit earlier, but is a sort of convergence of AI and AI systems with sort of turning into AI native businesses. And so, as you sort of do more, get good further along this transformation curve with AI, it turns out that like the better the performance of your AI systems, the better the performance of your business. Because these models are really starting to underpin all these key areas that cumulatively drive your P&L. And so one of the things that we work a lot with our customers is to do is just understand, you know, take these really esoteric data science notions and performance and tie them to all their business KPIs so that way you really are, it's kind of like the operating system for running your AI native business. And we're starting to see more and more companies get farther along that maturity curve and starting to think that way, which is really exciting. >> I love the AI native. I haven't heard any startup yet say AI first, although we kind of use the term, but I guarantee that's going to come in all the pitch decks, we're an AI first company, it's going to be great run. Adam, congratulations on your success to you and the team. Hey, if we do a few more interviews, we'll get the linguistics down. We can have bots just interact with you directly and ask you, have an interview directly. >> That sounds good, I'm going to go hang out on the beach, right? So, sounds good. >> Thanks for coming on, really appreciate the conversation. Super exciting, really important area and you guys doing great work. Thanks for coming on. >> Adam: Yeah, thanks John. >> Again, this is Cube Conversation. I'm John Furrier here in Palo Alto, AI going next gen. This is legit, this is going to a whole nother level that's going to open up huge opportunities for startups, that's going to use opportunities for investors and the value to the users and the experience will come in, in ways I think no one will ever see. So keep an eye out for more coverage on siliconangle.com and theCUBE.net, thanks for watching. (bright upbeat music)

(upbeat instrumental) >> Hello and welcome to this CUBE conversation. I'm John Furrier, host of theCUBE, here in Palo Alto, California. Got a great conversation with Robert Nishihara who's the co-founder and CEO of Anyscale. Robert, great to have you on this CUBE conversation. It's great to see you. We did your first Ray Summit a couple years ago and congratulations on your venture. Great to have you on. >> Thank you. Thanks for inviting me. >> So you're first time CEO out of Berkeley in Data. You got the Databricks is coming out of there. You got a bunch of activity coming from Berkeley. It's like a, it really is kind of like where a lot of innovations going on data. Anyscale has been one of those startups that has risen out of that scene. Right? You look at the success of what the Data lakes are now. Now you've got the generative AI. This has been a really interesting innovation market. This new wave is coming. Tell us what's going on with Anyscale right now, as you guys are gearing up and getting some growth. What's happening with the company? >> Yeah, well one of the most exciting things that's been happening in computing recently, is the rise of AI and the excitement about AI, and the potential for AI to really transform every industry. Now of course, one of the of the biggest challenges to actually making that happen is that doing AI, that AI is incredibly computationally intensive, right? To actually succeed with AI to actually get value out of AI. You're typically not just running it on your laptop, you're often running it and scaling it across thousands of machines, or hundreds of machines or GPUs, and to, so organizations and companies and businesses that do AI often end up building a large infrastructure team to manage the distributed systems, the computing to actually scale these applications. And that's a, that's a, a huge software engineering lift, right? And so, one of the goals for Anyscale is really to make that easy. To get to the point where, developers and teams and companies can succeed with AI. Can build these scalable AI applications, without really you know, without a huge investment in infrastructure with a lot of, without a lot of expertise in infrastructure, where really all they need to know is how to program on their laptop, how to program in Python. And if you have that, then that's really all you need to succeed with AI. So that's what we've been focused on. We're building Ray, which is an open source project that's been starting to get adopted by tons of companies, to actually train these models, to deploy these models, to do inference with these models, you know, to ingest and pre-process their data. And our goals, you know, here with the company are really to make Ray successful. To grow the Ray community, and then to build a great product around it and simplify the development and deployment, and productionization of machine learning for, for all these businesses. >> It's a great trend. Everyone wants developer productivity seeing that, clearly right now. And plus, developers are voting literally on what standards become. As you look at how the market is open source driven, a lot of that I love the model, love the Ray project love the, love the Anyscale value proposition. How big are you guys now, and how is that value proposition of Ray and Anyscale and foundational models coming together? Because it seems like you guys are in a perfect storm situation where you guys could get a real tailwind and draft off the the mega trend that everyone's getting excited. The new toy is ChatGPT. So you got to look at that and say, hey, I mean, come on, you guys did all the heavy lifting. >> Absolutely. >> You know how many people you are, and what's the what's the proposition for you guys these days? >> You know our company's about a hundred people, that a bit larger than that. Ray's been going really quickly. It's been, you know, companies using, like OpenAI uses Ray to train their models, like ChatGPT. Companies like Uber run all their deep learning you know, and classical machine learning on top of Ray. Companies like Shopify, Spotify, Netflix, Cruise, Lyft, Instacart, you know, Bike Dance. A lot of these companies are investing heavily in Ray for their machine learning infrastructure. And I think it's gotten to the point where, if you're one of these, you know type of businesses, and you're looking to revamp your machine learning infrastructure. If you're looking to enable new capabilities, you know make your teams more productive, increase, speed up the experimentation cycle, you know make it more performance, like build, you know, run applications that are more scalable, run them faster, run them in a more cost efficient way. All of these types of companies are at least evaluating Ray and Ray is an increasingly common choice there. I think if they're not using Ray, if many of these companies that end up not using Ray, they often end up building their own infrastructure. So Ray has been, the growth there has been incredibly exciting over the, you know we had our first in-person Ray Summit just back in August, and planning the next one for, for coming September. And so when you asked about the value proposition, I think there's there's really two main things, when people choose to go with Ray and Anyscale. One reason is about moving faster, right? It's about developer productivity, it's about speeding up the experimentation cycle, easily getting their models in production. You know, we hear many companies say that they, you know they, once they prototype a model, once they develop a model, it's another eight weeks, or 12 weeks to actually get that model in production. And that's a reason they talk to us. We hear companies say that, you know they've been training their models and, and doing inference on a single machine, and they've been sort of scaling vertically, like using bigger and bigger machines. But they, you know, you can only do that for so long, and at some point you need to go beyond a single machine and that's when they start talking to us. Right? So one of the main value propositions is around moving faster. I think probably the phrase I hear the most is, companies saying that they don't want their machine learning people to have to spend all their time configuring infrastructure. All this is about productivity. >> Yeah. >> The other. >> It's the big brains in the company. That are being used to do remedial tasks that should be automated right? I mean that's. >> Yeah, and I mean, it's hard stuff, right? It's also not these people's area of expertise, and or where they're adding the most value. So all of this is around developer productivity, moving faster, getting to market faster. The other big value prop and the reason people choose Ray and choose Anyscale, is around just providing superior infrastructure. This is really, can we scale more? You know, can we run it faster, right? Can we run it in a more cost effective way? We hear people saying that they're not getting good GPU utilization with the existing tools they're using, or they can't scale beyond a certain point, or you know they don't have a way to efficiently use spot instances to save costs, right? Or their clusters, you know can't auto scale up and down fast enough, right? These are all the kinds of things that Ray and Anyscale, where Ray and Anyscale add value and solve these kinds of problems. >> You know, you bring up great points. Auto scaling concept, early days, it was easy getting more compute. Now it's complicated. They're built into more integrated apps in the cloud. And you mentioned those companies that you're working with, that's impressive. Those are like the big hardcore, I call them hardcore. They have a good technical teams. And as the wave starts to move from these companies that were hyper scaling up all the time, the mainstream are just developers, right? So you need an interface in, so I see the dots connecting with you guys and I want to get your reaction. Is that how you see it? That you got the alphas out there kind of kicking butt, building their own stuff, alpha developers and infrastructure. But mainstream just wants programmability. They want that heavy lifting taken care of for them. Is that kind of how you guys see it? I mean, take us through that. Because to get crossover to be democratized, the automation's got to be there. And for developer productivity to be in, it's got to be coding and programmability. >> That's right. Ultimately for AI to really be successful, and really you know, transform every industry in the way we think it has the potential to. It has to be easier to use, right? And that is, and being easier to use, there's many dimensions to that. But an important one is that as a developer to do AI, you shouldn't have to be an expert in distributed systems. You shouldn't have to be an expert in infrastructure. If you do have to be, that's going to really limit the number of people who can do this, right? And I think there are so many, all of the companies we talk to, they don't want to be in the business of building and managing infrastructure. It's not that they can't do it. But it's going to slow them down, right? They want to allocate their time and their energy toward building their product, right? To building a better product, getting their product to market faster. And if we can take the infrastructure work off of the critical path for them, that's going to speed them up, it's going to simplify their lives. And I think that is critical for really enabling all of these companies to succeed with AI. >> Talk about the customers you guys are talking to right now, and how that translates over. Because I think you hit a good thread there. Data infrastructure is critical. Managed services are coming online, open sources continuing to grow. You have these people building their own, and then if they abandon it or don't scale it properly, there's kind of consequences. 'Cause it's a system you mentioned, it's a distributed system architecture. It's not as easy as standing up a monolithic app these days. So when you guys go to the marketplace and talk to customers, put the customers in buckets. So you got the ones that are kind of leaning in, that are pretty peaked, probably working with you now, open source. And then what's the customer profile look like as you go mainstream? Are they looking to manage service, looking for more architectural system, architecture approach? What's the, Anyscale progression? How do you engage with your customers? What are they telling you? >> Yeah, so many of these companies, yes, they're looking for managed infrastructure 'cause they want to move faster, right? Now the kind of these profiles of these different customers, they're three main workloads that companies run on Anyscale, run with Ray. It's training related workloads, and it is serving and deployment related workloads, like actually deploying your models, and it's batch processing, batch inference related workloads. Like imagine you want to do computer vision on tons and tons of, of images or videos, or you want to do natural language processing on millions of documents or audio, or speech or things like that, right? So the, I would say the, there's a pretty large variety of use cases, but the most common you know, we see tons of people working with computer vision data, you know, computer vision problems, natural language processing problems. And it's across many different industries. We work with companies doing drug discovery, companies doing you know, gaming or e-commerce, right? Companies doing robotics or agriculture. So there's a huge variety of the types of industries that can benefit from AI, and can really get a lot of value out of AI. And, but the, but the problems are the same problems that they all want to solve. It's like how do you make your team move faster, you know succeed with AI, be more productive, speed up the experimentation, and also how do you do this in a more performant way, in a faster, cheaper, in a more cost efficient, more scalable way. >> It's almost like the cloud game is coming back to AI and these foundational models, because I was just on a podcast, we recorded our weekly podcast, and I was just riffing with Dave Vellante, my co-host on this, were like, hey, in the early days of Amazon, if you want to build an app, you just, you have to build a data center, and then you go to now you go to the cloud, cloud's easier, pay a little money, penny's on the dollar, you get your app up and running. Cloud computing is born. With foundation models in generative AI. The old model was hard, heavy lifting, expensive, build out, before you get to do anything, as you mentioned time. So I got to think that you're pretty much in a good position with this foundational model trend in generative AI because I just looked at the foundation map, foundation models, map of the ecosystem. You're starting to see layers of, you got the tooling, you got platform, you got cloud. It's filling out really quickly. So why is Anyscale important to this new trend? How do you talk to people when they ask you, you know what does ChatGPT mean for Anyscale? And how does the financial foundational model growth, fit into your plan? >> Well, foundational models are hugely important for the industry broadly. Because you're going to have these really powerful models that are trained that you know, have been trained on tremendous amounts of data. tremendous amounts of computes, and that are useful out of the box, right? That people can start to use, and query, and get value out of, without necessarily training these huge models themselves. Now Ray fits in and Anyscale fit in, in a number of places. First of all, they're useful for creating these foundation models. Companies like OpenAI, you know, use Ray for this purpose. Companies like Cohere use Ray for these purposes. You know, IBM. If you look at, there's of course also open source versions like GPTJ, you know, created using Ray. So a lot of these large language models, large foundation models benefit from training on top of Ray. And, but of course for every company training and creating these huge foundation models, you're going to have many more that are fine tuning these models with their own data. That are deploying and serving these models for their own applications, that are building other application and business logic around these models. And that's where Ray also really shines, because Ray you know, is, can provide common infrastructure for all of these workloads. The training, the fine tuning, the serving, the data ingest and pre-processing, right? The hyper parameter tuning, the and and so on. And so where the reason Ray and Anyscale are important here, is that, again, foundation models are large, foundation models are compute intensive, doing you know, using both creating and using these foundation models requires tremendous amounts of compute. And there there's a big infrastructure lift to make that happen. So either you are using Ray and Anyscale to do this, or you are building the infrastructure and managing the infrastructure yourself. Which you can do, but it's, it's hard. >> Good luck with that. I always say good luck with that. I mean, I think if you really need to do, build that hardened foundation, you got to go all the way. And I think this, this idea of composability is interesting. How is Ray working with OpenAI for instance? Take, take us through that. Because I think you're going to see a lot of people talking about, okay I got trained models, but I'm going to have not one, I'm going to have many. There's big debate that OpenAI is going to be the mother of all LLMs, but now, but really people are also saying that to be many more, either purpose-built or specific. The fusion and these things come together there's like a blending of data, and that seems to be a value proposition. How does Ray help these guys get their models up? Can you take, take us through what Ray's doing for say OpenAI and others, and how do you see the models interacting with each other? >> Yeah, great question. So where, where OpenAI uses Ray right now, is for the training workloads. Training both to create ChatGPT and models like that. There's both a supervised learning component, where you're pre-training this model on doing supervised pre-training with example data. There's also a reinforcement learning component, where you are fine-tuning the model and continuing to train the model, but based on human feedback, based on input from humans saying that, you know this response to this question is better than this other response to this question, right? And so Ray provides the infrastructure for scaling the training across many, many GPUs, many many machines, and really running that in an efficient you know, performance fault tolerant way, right? And so, you know, open, this is not the first version of OpenAI's infrastructure, right? They've gone through iterations where they did start with building the infrastructure themselves. They were using tools like MPI. But at some point, you know, given the complexity, given the scale of what they're trying to do, you hit a wall with MPI and that's going to happen with a lot of other companies in this space. And at that point you don't have many other options other than to use Ray or to build your own infrastructure. >> That's awesome. And then your vision on this data interaction, because the old days monolithic models were very rigid. You couldn't really interface with them. But we're kind of seeing this future of data fusion, data interaction, data blending at large scale. What's your vision? How do you, what's your vision of where this goes? Because if this goes the way people think. You can have this data chemistry kind of thing going on where people are integrating all kinds of data with each other at large scale. So you need infrastructure, intelligence, reasoning, a lot of code. Is this something that you see? What's your vision in all this? Take us through. >> AI is going to be used everywhere right? It's, we see this as a technology that's going to be ubiquitous, and is going to transform every business. I mean, imagine you make a product, maybe you were making a tool like Photoshop or, or whatever the, you know, tool is. The way that people are going to use your tool, is not by investing, you know, hundreds of hours into learning all of the different, you know specific buttons they need to press and workflows they need to go through it. They're going to talk to it, right? They're going to say, ask it to do the thing they want it to do right? And it's going to do it. And if it, if it doesn't know what it's want, what it's, what's being asked of it. It's going to ask clarifying questions, right? And then you're going to clarify, and you're going to have a conversation. And this is going to make many many many kinds of tools and technology and products easier to use, and lower the barrier to entry. And so, and this, you know, many companies fit into this category of trying to build products that, and trying to make them easier to use, this is just one kind of way it can, one kind of way that AI will will be used. But I think it's, it's something that's pretty ubiquitous. >> Yeah. It'll be efficient, it'll be efficiency up and down the stack, and will change the productivity equation completely. You just highlighted one, I don't want to fill out forms, just stand up my environment for me. And then start coding away. Okay well this is great stuff. Final word for the folks out there watching, obviously new kind of skill set for hiring. You guys got engineers, give a plug for the company, for Anyscale. What are you looking for? What are you guys working on? Give a, take the last minute to put a plug in for the company. >> Yeah well if you're interested in AI and if you think AI is really going to be transformative, and really be useful for all these different industries. We are trying to provide the infrastructure to enable that to happen, right? So I think there's the potential here, to really solve an important problem, to get to the point where developers don't need to think about infrastructure, don't need to think about distributed systems. All they think about is their application logic, and what they want their application to do. And I think if we can achieve that, you know we can be the foundation or the platform that enables all of these other companies to succeed with AI. So that's where we're going. I think something like this has to happen if AI is going to achieve its potential, we're looking for, we're hiring across the board, you know, great engineers, on the go-to-market side, product managers, you know people who want to really, you know, make this happen. >> Awesome well congratulations. I know you got some good funding behind you. You're in a good spot. I think this is happening. I think generative AI and foundation models is going to be the next big inflection point, as big as the pc inter-networking, internet and smartphones. This is a whole nother application framework, a whole nother set of things. So this is the ground floor. Robert, you're, you and your team are right there. Well done. >> Thank you so much. >> All right. Thanks for coming on this CUBE conversation. I'm John Furrier with theCUBE. Breaking down a conversation around AI and scaling up in this new next major inflection point. This next wave is foundational models, generative AI. And thanks to ChatGPT, the whole world's now knowing about it. So it really is changing the game and Anyscale is right there, one of the hot startups, that is in good position to ride this next wave. Thanks for watching. (upbeat instrumental)

(upbeat music) >> Hello, everyone, welcome to theCUBE. I'm John Furrier, the host of theCUBE in Palo Alto, California. Also it's SiliconANGLE News. Got two great guests here to talk about AI, the impact of the future of the internet, the applications, the people. Amr Awadallah, the founder and CEO, Ed Alban is the CEO of Vectara, a new startup that emerged out of the original Cloudera, I would say, 'cause Amr's known, famous for the Cloudera founding, which was really the beginning of the big data movement. And now as AI goes mainstream, there's so much to talk about, so much to go on. And plus the new company is one of the, now what I call the wave, this next big wave, I call it the fifth wave in the industry. You know, you had PCs, you had the internet, you had mobile. This generative AI thing is real. And you're starting to see startups come out in droves. Amr obviously was founder of Cloudera, Big Data, and now Vectara. And Ed Albanese, you guys have a new company. Welcome to the show. >> Thank you. It's great to be here. >> So great to see you. Now the story is theCUBE started in the Cloudera office. Thanks to you, and your friendly entrepreneurship views that you have. We got to know each other over the years. But Cloudera had Hadoop, which was the beginning of what I call the big data wave, which then became what we now call data lakes, data oceans, and data infrastructure that's developed from that. It's almost interesting to look back 12 plus years, and see that what AI is doing now, right now, is opening up the eyes to the mainstream, and the application's almost mind blowing. You know, Sati Natel called it the Mosaic Moment, didn't say Netscape, he built Netscape (laughing) but called it the Mosaic Moment. You're seeing companies in startups, kind of the alpha geeks running here, because this is the new frontier, and there's real meat on the bone, in terms of like things to do. Why? Why is this happening now? What's is the confluence of the forces happening, that are making this happen? >> Yeah, I mean if you go back to the Cloudera days, with big data, and so on, that was more about data processing. Like how can we process data, so we can extract numbers from it, and do reporting, and maybe take some actions, like this is a fraud transaction, or this is not. And in the meanwhile, many of the researchers working in the neural network, and deep neural network space, were trying to focus on data understanding, like how can I understand the data, and learn from it, so I can take actual actions, based on the data directly, just like a human does. And we were only good at doing that at the level of somebody who was five years old, or seven years old, all the way until about 2013. And starting in 2013, which is only 10 years ago, a number of key innovations started taking place, and each one added on. It was no major innovation that just took place. It was a couple of really incremental ones, but they added on top of each other, in a very exponentially additive way, that led to, by the end of 2019, we now have models, deep neural network models, that can read and understand human text just like we do. Right? And they can reason about it, and argue with you, and explain it to you. And I think that's what is unlocking this whole new wave of innovation that we're seeing right now. So data understanding would be the essence of it. >> So it's not a Big Bang kind of theory, it's been evolving over time, and I think that the tipping point has been the advancements and other things. I mean look at cloud computing, and look how fast it just crept up on AWS. I mean AWS you back three, five years ago, I was talking to Swami yesterday, and their big news about AI, expanding the Hugging Face's relationship with AWS. And just three, five years ago, there wasn't a model training models out there. But as compute comes out, and you got more horsepower,, these large language models, these foundational models, they're flexible, they're not monolithic silos, they're interacting. There's a whole new, almost fusion of data happening. Do you see that? I mean is that part of this? >> Of course, of course. I mean this wave is building on all the previous waves. We wouldn't be at this point if we did not have hardware that can scale, in a very efficient way. We wouldn't be at this point, if we don't have data that we're collecting about everything we do, that we're able to process in this way. So this, this movement, this motion, this phase we're in, absolutely builds on the shoulders of all the previous phases. For some of the observers from the outside, when they see chatGPT for the first time, for them was like, "Oh my god, this just happened overnight." Like it didn't happen overnight. (laughing) GPT itself, like GPT3, which is what chatGPT is based on, was released a year ahead of chatGPT, and many of us were seeing the power it can provide, and what it can do. I don't know if Ed agrees with that. >> Yeah, Ed? >> I do. Although I would acknowledge that the possibilities now, because of what we've hit from a maturity standpoint, have just opened up in an incredible way, that just wasn't tenable even three years ago. And that's what makes it, it's true that it developed incrementally, in the same way that, you know, the possibilities of a mobile handheld device, you know, in 2006 were there, but when the iPhone came out, the possibilities just exploded. And that's the moment we're in. >> Well, I've had many conversations over the past couple months around this area with chatGPT. John Markoff told me the other day, that he calls it, "The five dollar toy," because it's not that big of a deal, in context to what AI's doing behind the scenes, and all the work that's done on ethics, that's happened over the years, but it has woken up the mainstream, so everyone immediately jumps to ethics. "Does it work? "It's not factual," And everyone who's inside the industry is like, "This is amazing." 'Cause you have two schools of thought there. One's like, people that think this is now the beginning of next gen, this is now we're here, this ain't your grandfather's chatbot, okay?" With NLP, it's got reasoning, it's got other things. >> I'm in that camp for sure. >> Yeah. Well I mean, everyone who knows what's going on is in that camp. And as the naysayers start to get through this, and they go, "Wow, it's not just plagiarizing homework, "it's helping me be better. "Like it could rewrite my memo, "bring the lead to the top." It's so the format of the user interface is interesting, but it's still a data-driven app. >> Absolutely. >> So where does it go from here? 'Cause I'm not even calling this the first ending. This is like pregame, in my opinion. What do you guys see this going, in terms of scratching the surface to what happens next? >> I mean, I'll start with, I just don't see how an application is going to look the same in the next three years. Who's going to want to input data manually, in a form field? Who is going to want, or expect, to have to put in some text in a search box, and then read through 15 different possibilities, and try to figure out which one of them actually most closely resembles the question they asked? You know, I don't see that happening. Who's going to start with an absolute blank sheet of paper, and expect no help? That is not how an application will work in the next three years, and it's going to fundamentally change how people interact and spend time with opening any element on their mobile phone, or on their computer, to get something done. >> Yes. I agree with that. Like every single application, over the next five years, will be rewritten, to fit within this model. So imagine an HR application, I don't want to name companies, but imagine an HR application, and you go into application and you clicking on buttons, because you want to take two weeks of vacation, and menus, and clicking here and there, reasons and managers, versus just telling the system, "I'm taking two weeks of vacation, going to Las Vegas," book it, done. >> Yeah. >> And the system just does it for you. If you weren't completing in your input, in your description, for what you want, then the system asks you back, "Did you mean this? "Did you mean that? "Were you trying to also do this as well?" >> Yeah. >> "What was the reason?" And that will fit it for you, and just do it for you. So I think the user interface that we have with apps, is going to change to be very similar to the user interface that we have with each other. And that's why all these apps will need to evolve. >> I know we don't have a lot of time, 'cause you guys are very busy, but I want to definitely have multiple segments with you guys, on this topic, because there's so much to talk about. There's a lot of parallels going on here. I was talking again with Swami who runs all the AI database at AWS, and I asked him, I go, "This feels a lot like the original AWS. "You don't have to provision a data center." A lot of this heavy lifting on the back end, is these large language models, with these foundational models. So the bottleneck in the past, was the energy, and cost to actually do it. Now you're seeing it being stood up faster. So there's definitely going to be a tsunami of apps. I would see that clearly. What is it? We don't know yet. But also people who are going to leverage the fact that I can get started building value. So I see a startup boom coming, and I see an application tsunami of refactoring things. >> Yes. >> So the replatforming is already kind of happening. >> Yes, >> OpenAI, chatGPT, whatever. So that's going to be a developer environment. I mean if Amazon turns this into an API, or a Microsoft, what you guys are doing. >> We're turning it into API as well. That's part of what we're doing as well, yes. >> This is why this is exciting. Amr, you've lived the big data dream, and and we used to talk, if you didn't have a big data problem, if you weren't full of data, you weren't really getting it. Now people have all the data, and they got to stand this up. >> Yeah. >> So the analogy is again, the mobile, I like the mobile movement, and using mobile as an analogy, most companies were not building for a mobile environment, right? They were just building for the web, and legacy way of doing apps. And as soon as the user expectations shifted, that my expectation now, I need to be able to do my job on this small screen, on the mobile device with a touchscreen. Everybody had to invest in re-architecting, and re-implementing every single app, to fit within that model, and that model of interaction. And we are seeing the exact same thing happen now. And one of the core things we're focused on at Vectara, is how to simplify that for organizations, because a lot of them are overwhelmed by large language models, and ML. >> They don't have the staff. >> Yeah, yeah, yeah. They're understaffed, they don't have the skills. >> But they got developers, they've got DevOps, right? >> Yes. >> So they have the DevSecOps going on. >> Exactly, yes. >> So our goal is to simplify it enough for them that they can start leveraging this technology effectively, within their applications. >> Ed, you're the COO of the company, obviously a startup. You guys are growing. You got great backup, and good team. You've also done a lot of business development, and technical business development in this area. If you look at the landscape right now, and I agree the apps are coming, every company I talk to, that has that jet chatGPT of, you know, epiphany, "Oh my God, look how cool this is. "Like magic." Like okay, it's code, settle down. >> Mm hmm. >> But everyone I talk to is using it in a very horizontal way. I talk to a very senior person, very tech alpha geek, very senior person in the industry, technically. they're using it for log data, they're using it for configuration of routers. And in other areas, they're using it for, every vertical has a use case. So this is horizontally scalable from a use case standpoint. When you hear horizontally scalable, first thing I chose in my mind is cloud, right? >> Mm hmm. >> So cloud, and scalability that way. And the data is very specialized. So now you have this vertical specialization, horizontally scalable, everyone will be refactoring. What do you see, and what are you seeing from customers, that you talk to, and prospects? >> Yeah, I mean put yourself in the shoes of an application developer, who is actually trying to make their application a bit more like magic. And to have that soon-to-be, honestly, expected experience. They've got to think about things like performance, and how efficiently that they can actually execute a query, or a question. They've got to think about cost. Generative isn't cheap, like the inference of it. And so you've got to be thoughtful about how and when you take advantage of it, you can't use it as a, you know, everything looks like a nail, and I've got a hammer, and I'm going to hit everything with it, because that will be wasteful. Developers also need to think about how they're going to take advantage of, but not lose their own data. So there has to be some controls around what they feed into the large language model, if anything. Like, should they fine tune a large language model with their own data? Can they keep it logically separated, but still take advantage of the powers of a large language model? And they've also got to take advantage, and be aware of the fact that when data is generated, that it is a different class of data. It might not fully be their own. >> Yeah. >> And it may not even be fully verified. And so when the logical cycle starts, of someone making a request, the relationship between that request, and the output, those things have to be stored safely, logically, and identified as such. >> Yeah. >> And taken advantage of in an ongoing fashion. So these are mega problems, each one of them independently, that, you know, you can think of it as middleware companies need to take advantage of, and think about, to help the next wave of application development be logical, sensible, and effective. It's not just calling some raw API on the cloud, like openAI, and then just, you know, you get your answer and you're done, because that is a very brute force approach. >> Well also I will point, first of all, I agree with your statement about the apps experience, that's going to be expected, form filling. Great point. The interesting about chatGPT. >> Sorry, it's not just form filling, it's any action you would like to take. >> Yeah. >> Instead of clicking, and dragging, and dropping, and doing it on a menu, or on a touch screen, you just say it, and it's and it happens perfectly. >> Yeah. It's a different interface. And that's why I love that UIUX experiences, that's the people falling out of their chair moment with chatGPT, right? But a lot of the things with chatGPT, if you feed it right, it works great. If you feed it wrong and it goes off the rails, it goes off the rails big. >> Yes, yes. >> So the the Bing catastrophes. >> Yeah. >> And that's an example of garbage in, garbage out, classic old school kind of comp-side phrase that we all use. >> Yep. >> Yes. >> This is about data in injection, right? It reminds me the old SQL days, if you had to, if you can sling some SQL, you were a magician, you know, to get the right answer, it's pretty much there. So you got to feed the AI. >> You do, Some people call this, the early word to describe this as prompt engineering. You know, old school, you know, search, or, you know, engagement with data would be, I'm going to, I have a question or I have a query. New school is, I have, I have to issue it a prompt, because I'm trying to get, you know, an action or a reaction, from the system. And the active engineering, there are a lot of different ways you could do it, all the way from, you know, raw, just I'm going to send you whatever I'm thinking. >> Yeah. >> And you get the unintended outcomes, to more constrained, where I'm going to just use my own data, and I'm going to constrain the initial inputs, the data I already know that's first party, and I trust, to, you know, hyper constrain, where the application is actually, it's looking for certain elements to respond to. >> It's interesting Amr, this is why I love this, because one we are in the media, we're recording this video now, we'll stream it. But we got all your linguistics, we're talking. >> Yes. >> This is data. >> Yep. >> So the data quality becomes now the new intellectual property, because, if you have that prompt source data, it makes data or content, in our case, the original content, intellectual property. >> Absolutely. >> Because that's the value. And that's where you see chatGPT fall down, is because they're trying to scroll the web, and people think it's search. It's not necessarily search, it's giving you something that you wanted. It is a lot of that, I remember in Cloudera, you said, "Ask the right questions." Remember that phrase you guys had, that slogan? >> Mm hmm. And that's prompt engineering. So that's exactly, that's the reinvention of "Ask the right question," is prompt engineering is, if you don't give these models the question in the right way, and very few people know how to frame it in the right way with the right context, then you will get garbage out. Right? That is the garbage in, garbage out. But if you specify the question correctly, and you provide with it the metadata that constrain what that question is going to be acted upon or answered upon, then you'll get much better answers. And that's exactly what we solved Vectara. >> Okay. So before we get into the last couple minutes we have left, I want to make sure we get a plug in for the opportunity, and the profile of Vectara, your new company. Can you guys both share with me what you think the current situation is? So for the folks who are now having those moments of, "Ah, AI's bullshit," or, "It's not real, it's a lot of stuff," from, "Oh my god, this is magic," to, "Okay, this is the future." >> Yes. >> What would you say to that person, if you're at a cocktail party, or in the elevator say, "Calm down, this is the first inning." How do you explain the dynamics going on right now, to someone who's either in the industry, but not in the ropes? How would you explain like, what this wave's about? How would you describe it, and how would you prepare them for how to change their life around this? >> Yeah, so I'll go first and then I'll let Ed go. Efficiency, efficiency is the description. So we figured that a way to be a lot more efficient, a way where you can write a lot more emails, create way more content, create way more presentations. Developers can develop 10 times faster than they normally would. And that is very similar to what happened during the Industrial Revolution. I always like to look at examples from the past, to read what will happen now, and what will happen in the future. So during the Industrial Revolution, it was about efficiency with our hands, right? So I had to make a piece of cloth, like this piece of cloth for this shirt I'm wearing. Our ancestors, they had to spend month taking the cotton, making it into threads, taking the threads, making them into pieces of cloth, and then cutting it. And now a machine makes it just like that, right? And the ancestors now turned from the people that do the thing, to manage the machines that do the thing. And I think the same thing is going to happen now, is our efficiency will be multiplied extremely, as human beings, and we'll be able to do a lot more. And many of us will be able to do things they couldn't do before. So another great example I always like to use is the example of Google Maps, and GPS. Very few of us knew how to drive a car from one location to another, and read a map, and get there correctly. But once that efficiency of an AI, by the way, behind these things is very, very complex AI, that figures out how to do that for us. All of us now became amazing navigators that can go from any point to any point. So that's kind of how I look at the future. >> And that's a great real example of impact. Ed, your take on how you would talk to a friend, or colleague, or anyone who asks like, "How do I make sense of the current situation? "Is it real? "What's in it for me, and what do I do?" I mean every company's rethinking their business right now, around this. What would you say to them? >> You know, I usually like to show, rather than describe. And so, you know, the other day I just got access, I've been using an application for a long time, called Notion, and it's super popular. There's like 30 or 40 million users. And the new version of Notion came out, which has AI embedded within it. And it's AI that allows you primarily to create. So if you could break down the world of AI into find and create, for a minute, just kind of logically separate those two things, find is certainly going to be massively impacted in our experiences as consumers on, you know, Google and Bing, and I can't believe I just said the word Bing in the same sentence as Google, but that's what's happening now (all laughing), because it's a good example of change. >> Yes. >> But also inside the business. But on the crate side, you know, Notion is a wiki product, where you try to, you know, note down things that you are thinking about, or you want to share and memorialize. But sometimes you do need help to get it down fast. And just in the first day of using this new product, like my experience has really fundamentally changed. And I think that anybody who would, you know, anybody say for example, that is using an existing app, I would show them, open up the app. Now imagine the possibility of getting a starting point right off the bat, in five seconds of, instead of having to whole cloth draft this thing, imagine getting a starting point then you can modify and edit, or just dispose of and retry again. And that's the potential for me. I can't imagine a scenario where, in a few years from now, I'm going to be satisfied if I don't have a little bit of help, in the same way that I don't manually spell check every email that I send. I automatically spell check it. I love when I'm getting type ahead support inside of Google, or anything. Doesn't mean I always take it, or when texting. >> That's efficiency too. I mean the cloud was about developers getting stuff up quick. >> Exactly. >> All that heavy lifting is there for you, so you don't have to do it. >> Right? >> And you get to the value faster. >> Exactly. I mean, if history taught us one thing, it's, you have to always embrace efficiency, and if you don't fast enough, you will fall behind. Again, looking at the industrial revolution, the companies that embraced the industrial revolution, they became the leaders in the world, and the ones who did not, they all like. >> Well the AI thing that we got to watch out for, is watching how it goes off the rails. If it doesn't have the right prompt engineering, or data architecture, infrastructure. >> Yes. >> It's a big part. So this comes back down to your startup, real quick, I know we got a couple minutes left. Talk about the company, the motivation, and we'll do a deeper dive on on the company. But what's the motivation? What are you targeting for the market, business model? The tech, let's go. >> Actually, I would like Ed to go first. Go ahead. >> Sure, I mean, we're a developer-first, API-first platform. So the product is oriented around allowing developers who may not be superstars, in being able to either leverage, or choose, or select their own large language models for appropriate use cases. But they that want to be able to instantly add the power of large language models into their application set. We started with search, because we think it's going to be one of the first places that people try to take advantage of large language models, to help find information within an application context. And we've built our own large language models, focused on making it very efficient, and elegant, to find information more quickly. So what a developer can do is, within minutes, go up, register for an account, and get access to a set of APIs, that allow them to send data, to be converted into a format that's easy to understand for large language models, vectors. And then secondarily, they can issue queries, ask questions. And they can ask them very, the questions that can be asked, are very natural language questions. So we're talking about long form sentences, you know, drill down types of questions, and they can get answers that either come back in depending upon the form factor of the user interface, in list form, or summarized form, where summarized equals the opportunity to kind of see a condensed, singular answer. >> All right. I have a. >> Oh okay, go ahead, you go. >> I was just going to say, I'm going to be a customer for you, because I want, my dream was to have a hologram of theCUBE host, me and Dave, and have questions be generated in the metaverse. So you know. (all laughing) >> There'll be no longer any guests here. They'll all be talking to you guys. >> Give a couple bullets, I'll spit out 10 good questions. Publish a story. This brings the automation, I'm sorry to interrupt you. >> No, no. No, no, I was just going to follow on on the same. So another way to look at exactly what Ed described is, we want to offer you chatGPT for your own data, right? So imagine taking all of the recordings of all of the interviews you have done, and having all of the content of that being ingested by a system, where you can now have a conversation with your own data and say, "Oh, last time when I met Amr, "which video games did we talk about? "Which movie or book did we use as an analogy "for how we should be embracing data science, "and big data, which is moneyball," I know you use moneyball all the time. And you start having that conversation. So, now the data doesn't become a passive asset that you just have in your organization. No. It's an active participant that's sitting with you, on the table, helping you make decisions. >> One of my favorite things to do with customers, is to go to their site or application, and show them me using it. So for example, one of the customers I talked to was one of the biggest property management companies in the world, that lets people go and rent homes, and houses, and things like that. And you know, I went and I showed them me searching through reviews, looking for information, and trying different words, and trying to find out like, you know, is this place quiet? Is it comfortable? And then I put all the same data into our platform, and I showed them the world of difference you can have when you start asking that question wholeheartedly, and getting real information that doesn't have anything to do with the words you asked, but is really focused on the meaning. You know, when I asked like, "Is it quiet?" You know, answers would come back like, "The wind whispered through the trees peacefully," and you know, it's like nothing to do with quiet in the literal word sense, but in the meaning sense, everything to do with it. And that that was magical even for them, to see that. >> Well you guys are the front end of this big wave. Congratulations on the startup, Amr. I know you guys got great pedigree in big data, and you've got a great team, and congratulations. Vectara is the name of the company, check 'em out. Again, the startup boom is coming. This will be one of the major waves, generative AI is here. I think we'll look back, and it will be pointed out as a major inflection point in the industry. >> Absolutely. >> There's not a lot of hype behind that. People are are seeing it, experts are. So it's going to be fun, thanks for watching. >> Thanks John. (soft music)

(bright upbeat music) >> Hello, everyone. Welcome to SiliconANGLE News breaking story here. Amazon Web Services expanding their relationship with Hugging Face, breaking news here on SiliconANGLE. I'm John Furrier, SiliconANGLE reporter, founder, and also co-host of theCUBE. And I have with me, Swami, from Amazon Web Services, vice president of database, analytics, machine learning with AWS. Swami, great to have you on for this breaking news segment on AWS's big news. Thanks for coming on and taking the time. >> Hey, John, pleasure to be here. >> You know- >> Looking forward to it. >> We've had many conversations on theCUBE over the years, we've watched Amazon really move fast into the large data modeling, SageMaker became a very smashing success, obviously you've been on this for a while. Now with ChatGPT OpenAI, a lot of buzz going mainstream, takes it from behind the curtain inside the ropes, if you will, in the industry to a mainstream. And so this is a big moment, I think, in the industry, I want to get your perspective, because your news with Hugging Face, I think is another tell sign that we're about to tip over into a new accelerated growth around making AI now application aware, application centric, more programmable, more API access. What's the big news about, with AWS Hugging Face, you know, what's going on with this announcement? >> Yeah. First of all, they're very excited to announce our expanded collaboration with Hugging Face, because with this partnership, our goal, as you all know, I mean, Hugging Face, I consider them like the GitHub for machine learning. And with this partnership, Hugging Face and AWS, we'll be able to democratize AI for a broad range of developers, not just specific deep AI startups. And now with this, we can accelerate the training, fine tuning and deployment of these large language models, and vision models from Hugging Face in the cloud. And the broader context, when you step back and see what customer problem we are trying to solve with this announcement, essentially if you see these foundational models, are used to now create like a huge number of applications, suggest like tech summarization, question answering, or search image generation, creative, other things. And these are all stuff we are seeing in the likes of these ChatGPT style applications. But there is a broad range of enterprise use cases that we don't even talk about. And it's because these kind of transformative, generative AI capabilities and models are not available to, I mean, millions of developers. And because either training these elements from scratch can be very expensive or time consuming and need deep expertise, or more importantly, they don't need these generic models, they need them to be fine tuned for the specific use cases. And one of the biggest complaints we hear is that these models, when they try to use it for real production use cases, they are incredibly expensive to train and incredibly expensive to run inference on, to use it at a production scale. So, and unlike web search style applications, where the margins can be really huge, here in production use cases and enterprises, you want efficiency at scale. That's where Hugging Face and AWS share our mission. And by integrating with Trainium and Inferentia, we're able to handle the cost efficient training and inference at scale, I'll deep dive on it. And by teaming up on the SageMaker front, now the time it takes to build these models and fine tune them is also coming down. So that's what makes this partnership very unique as well. So I'm very excited. >> I want to get into the time savings and the cost savings as well on the training and inference, it's a huge issue, but before we get into that, just how long have you guys been working with Hugging Face? I know there's a previous relationship, this is an expansion of that relationship, can you comment on what's different about what's happened before and then now? >> Yeah. So, Hugging Face, we have had a great relationship in the past few years as well, where they have actually made their models available to run on AWS, you know, fashion. Even in fact, their Bloom Project was something many of our customers even used. Bloom Project, for context, is their open source project which builds a GPT-3 style model. And now with this expanded collaboration, now Hugging Face selected AWS for that next generation office generative AI model, building on their highly successful Bloom Project as well. And the nice thing is, now, by direct integration with Trainium and Inferentia, where you get cost savings in a really significant way, now, for instance, Trn1 can provide up to 50% cost to train savings, and Inferentia can deliver up to 60% better costs, and four x more higher throughput than (indistinct). Now, these models, especially as they train that next generation generative AI models, it is going to be, not only more accessible to all the developers, who use it in open, so it'll be a lot cheaper as well. And that's what makes this moment really exciting, because we can't democratize AI unless we make it broadly accessible and cost efficient and easy to program and use as well. >> Yeah. >> So very exciting. >> I'll get into the SageMaker and CodeWhisperer angle in a second, but you hit on some good points there. One, accessibility, which is, I call the democratization, which is getting this in the hands of developers, and/or AI to develop, we'll get into that in a second. So, access to coding and Git reasoning is a whole nother wave. But the three things I know you've been working on, I want to put in the buckets here and comment, one, I know you've, over the years, been working on saving time to train, that's a big point, you mentioned some of those stats, also cost, 'cause now cost is an equation on, you know, bundling whether you're uncoupling with hardware and software, that's a big issue. Where do I find the GPUs? Where's the horsepower cost? And then also sustainability. You've mentioned that in the past, is there a sustainability angle here? Can you talk about those three things, time, cost, and sustainability? >> Certainly. So if you look at it from the AWS perspective, we have been supporting customers doing machine learning for the past years. Just for broader context, Amazon has been doing ML the past two decades right from the early days of ML powered recommendation to actually also supporting all kinds of generative AI applications. If you look at even generative AI application within Amazon, Amazon search, when you go search for a product and so forth, we have a team called MFi within Amazon search that helps bring these large language models into creating highly accurate search results. And these are created with models, really large models with tens of billions of parameters, scales to thousands of training jobs every month and trained on large model of hardware. And this is an example of a really good large language foundation model application running at production scale, and also, of course, Alexa, which uses a large generator model as well. And they actually even had a research paper that showed that they are more, and do better in accuracy than other systems like GPT-3 and whatnot. So, and we also touched on things like CodeWhisperer, which uses generative AI to improve developer productivity, but in a responsible manner, because 40% of some of the studies show 40% of this generated code had serious security flaws in it. This is where we didn't just do generative AI, we combined with automated reasoning capabilities, which is a very, very useful technique to identify these issues and couple them so that it produces highly secure code as well. Now, all these learnings taught us few things, and which is what you put in these three buckets. And yeah, like more than 100,000 customers using ML and AI services, including leading startups in the generative AI space, like stability AI, AI21 Labs, or Hugging Face, or even Alexa, for that matter. They care about, I put them in three dimension, one is around cost, which we touched on with Trainium and Inferentia, where we actually, the Trainium, you provide to 50% better cost savings, but the other aspect is, Trainium is a lot more power efficient as well compared to traditional one. And Inferentia is also better in terms of throughput, when it comes to what it is capable of. Like it is able to deliver up to three x higher compute performance and four x higher throughput, compared to it's previous generation, and it is extremely cost efficient and power efficient as well. >> Well. >> Now, the second element that really is important is in a day, developers deeply value the time it takes to build these models, and they don't want to build models from scratch. And this is where SageMaker, which is, even going to Kaggle uses, this is what it is, number one, enterprise ML platform. What it did to traditional machine learning, where tens of thousands of customers use StageMaker today, including the ones I mentioned, is that what used to take like months to build these models have dropped down to now a matter of days, if not less. Now, a generative AI, the cost of building these models, if you look at the landscape, the model parameter size had jumped by more than thousand X in the past three years, thousand x. And that means the training is like a really big distributed systems problem. How do you actually scale these model training? How do you actually ensure that you utilize these efficiently? Because these machines are very expensive, let alone they consume a lot of power. So, this is where SageMaker capability to build, automatically train, tune, and deploy models really concern this, especially with this distributor training infrastructure, and those are some of the reasons why some of the leading generative AI startups are actually leveraging it, because they do not want a giant infrastructure team, which is constantly tuning and fine tuning, and keeping these clusters alive. >> It sounds like a lot like what startups are doing with the cloud early days, no data center, you move to the cloud. So, this is the trend we're seeing, right? You guys are making it easier for developers with Hugging Face, I get that. I love that GitHub for machine learning, large language models are complex and expensive to build, but not anymore, you got Trainium and Inferentia, developers can get faster time to value, but then you got the transformers data sets, token libraries, all that optimized for generator. This is a perfect storm for startups. Jon Turow, a former AWS person, who used to work, I think for you, is now a VC at Madrona Venture, he and I were talking about the generator AI landscape, it's exploding with startups. Every alpha entrepreneur out there is seeing this as the next frontier, that's the 20 mile stairs, next 10 years is going to be huge. What is the big thing that's happened? 'Cause some people were saying, the founder of Yquem said, "Oh, the start ups won't be real, because they don't all have AI experience." John Markoff, former New York Times writer told me that, AI, there's so much work done, this is going to explode, accelerate really fast, because it's almost like it's been waiting for this moment. What's your reaction? >> I actually think there is going to be an explosion of startups, not because they need to be AI startups, but now finally AI is really accessible or going to be accessible, so that they can create remarkable applications, either for enterprises or for disrupting actually how customer service is being done or how creative tools are being built. And I mean, this is going to change in many ways. When we think about generative AI, we always like to think of how it generates like school homework or arts or music or whatnot, but when you look at it on the practical side, generative AI is being actually used across various industries. I'll give an example of like Autodesk. Autodesk is a customer who runs an AWS and SageMaker. They already have an offering that enables generated design, where designers can generate many structural designs for products, whereby you give a specific set of constraints and they actually can generate a structure accordingly. And we see similar kind of trend across various industries, where it can be around creative media editing or various others. I have the strong sense that literally, in the next few years, just like now, conventional machine learning is embedded in every application, every mobile app that we see, it is pervasive, and we don't even think twice about it, same way, like almost all apps are built on cloud. Generative AI is going to be part of every startup, and they are going to create remarkable experiences without needing actually, these deep generative AI scientists. But you won't get that until you actually make these models accessible. And I also don't think one model is going to rule the world, then you want these developers to have access to broad range of models. Just like, go back to the early days of deep learning. Everybody thought it is going to be one framework that will rule the world, and it has been changing, from Caffe to TensorFlow to PyTorch to various other things. And I have a suspicion, we had to enable developers where they are, so. >> You know, Dave Vellante and I have been riffing on this concept called super cloud, and a lot of people have co-opted to be multicloud, but we really were getting at this whole next layer on top of say, AWS. You guys are the most comprehensive cloud, you guys are a super cloud, and even Adam and I are talking about ISVs evolving to ecosystem partners. I mean, your top customers have ecosystems building on top of it. This feels like a whole nother AWS. How are you guys leveraging the history of AWS, which by the way, had the same trajectory, startups came in, they didn't want to provision a data center, the heavy lifting, all the things that have made Amazon successful culturally. And day one thinking is, provide the heavy lifting, undifferentiated heavy lifting, and make it faster for developers to program code. AI's got the same thing. How are you guys taking this to the next level, because now, this is an opportunity for the competition to change the game and take it over? This is, I'm sure, a conversation, you guys have a lot of things going on in AWS that makes you unique. What's the internal and external positioning around how you take it to the next level? >> I mean, so I agree with you that generative AI has a very, very strong potential in terms of what it can enable in terms of next generation application. But this is where Amazon's experience and expertise in putting these foundation models to work internally really has helped us quite a bit. If you look at it, like amazon.com search is like a very, very important application in terms of what is the customer impact on number of customers who use that application openly, and the amount of dollar impact it does for an organization. And we have been doing it silently for a while now. And the same thing is true for like Alexa too, which actually not only uses it for natural language understanding other city, even national leverages is set for creating stories and various other examples. And now, our approach to it from AWS is we actually look at it as in terms of the same three tiers like we did in machine learning, because when you look at generative AI, we genuinely see three sets of customers. One is, like really deep technical expert practitioner startups. These are the startups that are creating the next generation models like the likes of stability AIs or Hugging Face with Bloom or AI21. And they generally want to build their own models, and they want the best price performance of their infrastructure for training and inference. That's where our investments in silicon and hardware and networking innovations, where Trainium and Inferentia really plays a big role. And we can nearly do that, and that is one. The second middle tier is where I do think developers don't want to spend time building their own models, let alone, they actually want the model to be useful to that data. They don't need their models to create like high school homeworks or various other things. What they generally want is, hey, I had this data from my enterprises that I want to fine tune and make it really work only for this, and make it work remarkable, can be for tech summarization, to generate a report, or it can be for better Q&A, and so forth. This is where we are. Our investments in the middle tier with SageMaker, and our partnership with Hugging Face and AI21 and co here are all going to very meaningful. And you'll see us investing, I mean, you already talked about CodeWhisperer, which is an open preview, but we are also partnering with a whole lot of top ISVs, and you'll see more on this front to enable the next wave of generated AI apps too, because this is an area where we do think lot of innovation is yet to be done. It's like day one for us in this space, and we want to enable that huge ecosystem to flourish. >> You know, one of the things Dave Vellante and I were talking about in our first podcast we just did on Friday, we're going to do weekly, is we highlighted the AI ChatGPT example as a horizontal use case, because everyone loves it, people are using it in all their different verticals, and horizontal scalable cloud plays perfectly into it. So I have to ask you, as you look at what AWS is going to bring to the table, a lot's changed over the past 13 years with AWS, a lot more services are available, how should someone rebuild or re-platform and refactor their application of business with AI, with AWS? What are some of the tools that you see and recommend? Is it Serverless, is it SageMaker, CodeWhisperer? What do you think's going to shine brightly within the AWS stack, if you will, or service list, that's going to be part of this? As you mentioned, CodeWhisperer and SageMaker, what else should people be looking at as they start tinkering and getting all these benefits, and scale up their ups? >> You know, if we were a startup, first, I would really work backwards from the customer problem I try to solve, and pick and choose, bar, I don't need to deal with the undifferentiated heavy lifting, so. And that's where the answer is going to change. If you look at it then, the answer is not going to be like a one size fits all, so you need a very strong, I mean, granted on the compute front, if you can actually completely accurate it, so unless, I will always recommend it, instead of running compute for running your ups, because it takes care of all the undifferentiated heavy lifting, but on the data, and that's where we provide a whole variety of databases, right from like relational data, or non-relational, or dynamo, and so forth. And of course, we also have a deep analytical stack, where data directly flows from our relational databases into data lakes and data virus. And you can get value along with partnership with various analytical providers. The area where I do think fundamentally things are changing on what people can do is like, with CodeWhisperer, I was literally trying to actually program a code on sending a message through Twilio, and I was going to pull up to read a documentation, and in my ID, I was actually saying like, let's try sending a message to Twilio, or let's actually update a Route 53 error code. All I had to do was type in just a comment, and it actually started generating the sub-routine. And it is going to be a huge time saver, if I were a developer. And the goal is for us not to actually do it just for AWS developers, and not to just generate the code, but make sure the code is actually highly secure and follows the best practices. So, it's not always about machine learning, it's augmenting with automated reasoning as well. And generative AI is going to be changing, and not just in how people write code, but also how it actually gets built and used as well. You'll see a lot more stuff coming on this front. >> Swami, thank you for your time. I know you're super busy. Thank you for sharing on the news and giving commentary. Again, I think this is a AWS moment and industry moment, heavy lifting, accelerated value, agility. AIOps is going to be probably redefined here. Thanks for sharing your commentary. And we'll see you next time, I'm looking forward to doing more follow up on this. It's going to be a big wave. Thanks. >> Okay. Thanks again, John, always a pleasure. >> Okay. This is SiliconANGLE's breaking news commentary. I'm John Furrier with SiliconANGLE News, as well as host of theCUBE. Swami, who's a leader in AWS, has been on theCUBE multiple times. We've been tracking the growth of how Amazon's journey has just been exploding past five years, in particular, past three. You heard the numbers, great performance, great reviews. This is a watershed moment, I think, for the industry, and it's going to be a lot of fun for the next 10 years. Thanks for watching. (bright music)

(gentle music) >> Hello, everyone. Welcome to this Cube Conversation. I'm John Furrier, host of theCUBE here, in our Palo Alto Studios. We're featuring OctoML. I'm with the CEO, Luis Ceze. Chief Executive Officer, Co-founder of OctoML. I'm John Furrier of theCUBE. Thanks for joining us today. Luis, great to see you. Last time we spoke was at "re:MARS" Amazon's event. Kind of a joint event between (indistinct) and Amazon, kind of put a lot together. Great to see you. >> Great to see you again, John. I really have good memories of that interview. You know, that was definitely a great time. Great to chat with you again. >> The world of ML and AI, machine learning and AI is really hot. Everyone's talking about it. It's really great to see that advance. So I'm looking forward to this conversation but before we get started, introduce who you are in OctoML. >> Sure. I'm Luis Ceze, Co-founder and CEO at OctoML. I'm also professor of Computer Science at University of Washington. You know, OctoML grew out of our efforts on the Apache CVM project, which is a compiler in runtime system that enables folks to run machine learning models in a broad set of harder in the Edge and in the Cloud very efficiently. You know, we grew that project and grew that community, definitely saw there was something to pain point there. And then we built OctoML, OctoML is about three and a half years old now. And the mission, the company is to enable customers to deploy models very efficiently in the Cloud. And make them, you know, run. Do it quickly, run fast, and run at a low cost, which is something that's especially timely right now. >> I like to point out also for the folks 'casue they should know that you're also a professor in the Computer Science department at University of Washington. A great program there. This is a really an inflection point with AI machine learning. The computer science industry has been waiting for decades to advance AI with all this new cloud computing, all the hardware and silicon advancements, GPUs. This is the perfect storm. And you know, this the computer science now we we're seeing an acceleration. Can you share your view, and you're obviously a professor in that department but also, an entrepreneur. This is a great time for computer science. Explain why. >> Absolutely, yeah, no. Just like the confluence of you know, advances in what, you know, computers can do as devices to computer information. Plus, you know, advances in AI that enable applications that you know, we thought it was highly futuristic and now it's just right there today. You know, AI that can generate photo realistic images from descriptions, you know, can write text that's pretty good. Can help augment, you know, human creativity in a really meaningful way. So you see the confluence of capabilities and the creativity of humankind into new applications is just extremely exciting, both from a researcher point of view as well as an entrepreneur point of view, right. >> What should people know about these large language models we're seeing with ChatGPT and how Google has got a lot of work going on that air. There's been a lot of work recently. What's different now about these models, and why are they so popular and effective now? What's the difference between now, and say five years ago, that makes it more- >> Oh, yeah. It's a huge inflection on their capabilities, I always say like emergent behavior, right? So as these models got more complex and our ability to train and deploy them, you know, got to this point... You know, they really crossed a threshold into doing things that are truly surprising, right? In terms of generating, you know, exhalation for things generating tax, summarizing tax, expending tax. And you know, exhibiting what to some may look like reasoning. They're not quite reasoning fundamentally. They're generating tax that looks like they're reasoning, but they do it so well, that it feels like was done by a human, right. So I would say that the biggest changes that, you know, now, they can actually do things that are extremely useful for business in people's lives today. And that wasn't the case five years ago. So that's in the model capabilities and that is being paired with huge advances in computing that enabled this to be... Enables this to be, you know, actually see line of sites to be deployed at scale, right. And that's where we come in, by the way, but yeah. >> Yeah, I want to get into that. And also, you know, the fusion of data integrating data sets at scales. Another one we're seeing a lot of happening now. It's not just some, you know, siloed, pre-built data modeling. It's a lot of agility and a lot of new integration capabilities of data. How is that impacting the dynamics? >> Yeah, absolutely. So I'll say that the ability to either take the data that has that exists in training a model to do something useful with it, and more interestingly I would say, using baseline foundational models and with a little bit of data, turn them into something that can do a specialized task really, really well. Created this really fast proliferation of really impactful applications, right? >> If every company now is looking at this trend and I'm seeing a lot... And I think every company will rebuild their business with machine learning. If they're not already doing it. And the folks that aren't will probably be dinosaurs will be out of business. This is a real business transformation moment where machine learning and AI, as it goes mainstream. I think it's just the beginning. This is where you guys come in, and you guys are poised for handling this frenzy to change business with machine learning models. How do you guys help customers as they look at this, you know, transition to get, you know, concept to production with machine learning? >> Great. Great questions, yeah, so I would say that it's fair to say there's a bunch of models out there that can do useful things right off the box, right? So and also, the ability to create models improved quite a bit. So the challenge now shifted to customers, you know. Everyone is looking to incorporating AI into their applications. So what we do for them is to, first of all, how do you do that quickly, without needing highly specialized, difficult to find engineering? And very importantly, how do you do that at cost that's accessible, right? So all of these fantastic models that we just talked about, they use an amount of computing that's just astronomical compared to anything else we've done in the past. It means the costs that come with it, are also very, very high. So it's important to enable customers to, you know, incorporate AI into their applications, to their use cases in a way that they can do, with the people that they have, and the costs that they can afford, such that they can have, you know, the maximum impacting possibly have. And finally, you know, helping them deal with hardware availability, as you know, even though we made a lot of progress in making computing cheaper and cheaper. Even to this day, you know, you can never get enough. And getting an allocation, getting the right hardware to run these incredibly hungry models is hard. And we help customers deal with, you know, harder availability as well. >> Yeah, for the folks watching as a... If you search YouTube, there's an interview we did last year at "re:MARS," I mentioned that earlier, just a great interview. You talked about this hardware independence, this traction. I want to get into that, because if you look at all the foundation models that are out there right now, that are getting traction, you're seeing two trends. You're seeing proprietary and open source. And obviously, open source always wins in my opinion, but, you know, there's this iPhone moment and android moment that one of your investors John Torrey from Madrona, talked about was is iPhone versus Android moment, you know, one's proprietary hardware and they're very specialized high performance and then open source. This is an important distinction and you guys are hardware independent. What's the... Explain what all this means. >> Yeah. Great set of questions. First of all, yeah. So, you know, OpenAI, and of course, they create ChatGPT and they offer an API to run these models that does amazing things. But customers have to be able to go and send their data over to OpenAI, right? So, and run the model there and get the outputs. Now, there's open source models that can do amazing things as well, right? So they typically open source models, so they don't lag behind, you know, these proprietary closed models by more than say, you know, six months or so, let's say. And it means that enabling customers to take the models that they want and deploy under their control is something that's very valuable, because one, you don't have to expose your data to externally. Two, you can customize the model even more to the things that you wanted to do. And then three, you can run on an infrastructure that can be much more cost effective than having to, you know, pay somebody else's, you know, cost and markup, right? So, and where we help them is essentially help customers, enable customers to take machine learning models, say an open source model, and automate the process of putting them into production, optimize them to run with the right performance, and more importantly, give them the independence to run where they need to run, where they can run best, right? >> Yeah, and also, you know, I point out all the time that, you know, there's never any stopping the innovation of hardware silicon. You're seeing cloud computing more coming in there. So, you know, being hardware independent has some advantages. And if you look at OpenAI, for instance, you mentioned ChatGPT, I think this is interesting because I think everyone is scratching their head, going, "Okay, I need to move to this new generation." What's your pro tip and advice for folks who want to move to, or businesses that want to say move to machine learning? How do they get started? What are some of the considerations they need to think about to deploy these models into production? >> Yeah, great though. Great set of questions. First of all, I mean, I'm sure they're very aware of the kind of things that you want to do with AI, right? So you could be interacting with customers, you know, automating, interacting with customers. It could be, you know, finding issues in production lines. It could be, you know... Generating, you know, making it easier to produce content and so on. Like, you know, customers, users would have an idea what they want to do. You know, from that it can actually determine, what kind of machine learning models would solve the problem that would, you know, fits that use case. But then, that's when the hard thing begins, right? So when you find a model, identify the model that can do the thing that you wanted to do, you need to turn that into a thing that you can deploy. So how do you go from machine learning model that does a thing that you need to do, to a container with the right executor, the artifact they can actually go and deploy, right? So we've seen customers doing that on their own, right? So, and it's got a bit of work, and that's why we are excited about the automation that we can offer and then turn that into a turnkey problem, right? So a turnkey process. >> Luis, talk about the use cases. If I don't mind going and double down on the previous answer. You got existing services, and then there's new AI applications, AI for applications. What are the use cases with existing stuff, and the new applications that are being built? >> Yeah, I mean, existing itself is, for example, how do you do very smart search and auto completion, you know, when you are editing documents, for example. Very, very smart search of documents, summarization of tax, expanding bullets into pros in a way that, you know, don't have to spend as much human time. Just some of the existing applications, right? So some of the new ones are like truly AI native ways of producing content. Like there's a company that, you know, we share investors and love what they're doing called runwayyML, for example. It's sort of like an AI first way of editing and creating visual content, right? So you could say you have a video, you could say make this video look like, it's night as opposed to dark, or remove that dog in the corner. You can do that in a way that you couldn't do otherwise. So there's like definitely AI native use cases. And yet not only in life sciences, you know, there's quite a bit of advances on AI-based, you know, therapies and diagnostics processes that are designed using automated processes. And this is something that I feel like, we were just scratching the surface there. There's huge opportunities there, right? >> Talk about the inference and AI and production kind of angle here, because cost is a huge concern when you look at... And there's a hardware and that flexibility there. So I can see how that could help, but is there a cost freight train that can get out of control here if you don't deploy properly? Talk about the scale problem around cost in AI. >> Yeah, absolutely. So, you know, very quickly. One thing that people tend to think about is the cost is. You know, training has really high dollar amounts it tends over index on that. But what you have to think about is that for every model that's actually useful, you're going to train it once, and then run it a large number of times in inference. That means that over the lifetime of a model, the vast majority of the compute cycles and the cost are going to go to inference. And that's what we address, right? So, and to give you some idea, if you're talking about using large language model today, you know, you can say it's going to cost a couple of cents per, you know, 2,000 words output. If you have a million users active, you know, a day, you know, if you're lucky and you have that, you can, this cost can actually balloon very quickly to millions of dollars a month, just in inferencing costs. You know, assuming you know, that you actually have access to the infrastructure to run it, right? So means that if you don't pay attention to these inference costs and that's definitely going to be a surprise. And affects the economics of the product where this is embedded in, right? So this is something that, you know, if there's quite a bit of attention being put on right now on how do you do search with large language models and you don't pay attention to the economics, you know, you can have a surprise. You have to change the business model there. >> Yeah. I think that's important to call out, because you don't want it to be a runaway cost structure where you architected it wrong and then next thing you know, you got to unwind that. I mean, it's more than technical debt, it's actually real debt, it's real money. So, talk about some of the dynamics with the customers. How are they architecting this? How do they get ahead of that problem? What do you guys do specifically to solve that? >> Yeah, I mean, well, we help customers. So, it's first of all, be hyper aware, you know, understanding what's going to be the cost for them deploying the models into production and showing them the possibilities of how you can deploy the model with different cost structure, right? So that's where, you know, the ability to have hardware independence is so important because once you have hardware independence, after you optimize models, obviously, you have a new, you know, dimension of freedom to choose, you know, what is the right throughput per dollar for you. And then where, and what are the options? And once you make that decision, you want to automate the process of putting into production. So the way we help customers is showing very clearly in their use case, you know, how they can deploy their models in a much more cost-effective way. You know, when the cases... There's a case study that we put out recently, showing a 4x reduction in deployment costs, right? So this is by doing a mix optimization and choosing the right hardware. >> How do you address the concern that someone might say, Luis said, "Hey, you know, I don't want to degrade performance and latency, and I don't want the user experience to suffer." What's the answer there? >> Two things. So first of all, all of the manipulations that we do in the model is to turn the model to efficient code without changing the behavior of the models. We wouldn't degrade the experience of the user by having the model be wrong more often. And we don't change that at all. The model behaves the way it was validated for. And then the second thing is, you know, user experience with respect to latency, it's all about a maximum... Like, you could say, I want a model to run at 50 milliseconds or less. If it's much faster than 15 seconds, you're not going to notice the difference. But if it's lower, you're going to notice a difference. So the key here is that, how do you find a set of options to deploy, that you are not overshooting performance in a way that's going to lead to costs that has no additional benefits. And this provides a huge, a very significant margin of choices, set of choices that you can optimize for cost without degrading customer experience, right. End user experience. >> Yeah, and I also point out the large language models like the ChatGPTs of the world, they're coming out with Dave Moth and I were talking on this breaking analysis around, this being like, over 10X more computational intensive on capabilities. So this hardware independence is a huge thing. So, and also supply chain, some people can't get servers by the way, so, or hardware these days. >> Or even more interestingly, right? So they do not grow in trees, John. Like GPUs is not kind of stuff that you plant an orchard until you have a bunch and then you can increase it, but no, these things, you know, take a while. So, and you can't increase it overnight. So being able to live with those cycles that are available to you is not just important for all for cost, but also important for people to scale and serve more users at, you know, at whatever pace that they come, right? >> You know, it's really great to talk to you, and congratulations on OctaML. Looking forward to the startup showcase, we'll be featuring you guys there. But I want to get your personal opinion as someone in the industry and also, someone who's been in the computer science area for your career. You know, computer science has always been great, and there's more people enrolling in computer science, more diversity than ever before, but there's also more computer science related fields. How is this opening up computer science and where's AI going with the computers, with the science? Can you share your vision on, you know, the aperture, or the landscape of CompSci, or CS students, and opportunities. >> Yeah, no, absolutely. I think it's fair to say that computer has been embedded in pretty much every aspect of human life these days. Human life these days, right? So for everything. And AI has been a counterpart, it been an integral component of computer science for a while. And this medicines that happened in the last 10, 15 years in AI has shown, you know, new application has I think re-energized how people see what computers can do. And you, you know, there is this picture in our department that shows computer science at the center called the flower picture, and then all the different paddles like life sciences, social sciences, and then, you know, mechanical engineering, all these other things that, and I feel like it can replace that center with computer science. I put AI there as well, you see AI, you know touching all these applications. AI in healthcare, diagnostics. AI in discovery in the sciences, right? So, but then also AI doing things that, you know, the humans wouldn't have to do anymore. They can do better things with their brains, right? So it's permitting every single aspect of human life from intellectual endeavor to day-to-day work, right? >> Yeah. And I think the ChatGPT and OpenAI has really kind of created a mainstream view that everyone sees value in it. Like you could be in the data center, you could be in bio, you could be in healthcare. I mean, every industry sees value. So this brings up what I can call the horizontally scalable use constance. And so this opens up the conversation, what's going to change from this? Because if you go horizontally scalable, which is a cloud concept as you know, that's going to create a lot of opportunities and some shifting of how you think about architecture around data, for instance. What's your opinion on what this will do to change the inflection of the role of architecting platforms and the role of data specifically? >> Yeah, so good question. There is a lot in there, by the way, I should have added the previous question, that you can use AI to do better AI as well, which is what we do, and other folks are doing as well. And so the point I wanted to make here is that it's pretty clear that you have a cloud focus component with a nudge focused counterparts. Like you have AI models, but both in the Cloud and in the Edge, right? So the ability of being able to run your AI model where it runs best also has a data advantage to it from say, from a privacy point of view. That's inherently could say, "Hey, I want to run something, you know, locally, strictly locally, such that I don't expose the data to an infrastructure." And you know that the data never leaves you, right? Never leaves the device. Now you can imagine things that's already starting to happen, like you do some forms of training and model customization in the model architecture itself and the system architecture, such that you do this as close to the user as possible. And there's something called federated learning that has been around for some time now that's finally happening is, how do you get a data from butcher places, you do, you know, some common learning and then you send a model to the Edges, and they get refined for the final use in a way that you get the advantage of aggregating data but you don't get the disadvantage of privacy issues and so on. >> It's super exciting. >> And some of the considerations, yeah. >> It's super exciting area around data infrastructure, data science, computer science. Luis, congratulations on your success at OctaML. You're in the middle of it. And the best thing about its businesses are looking at this and really reinventing themselves and if a business isn't thinking about restructuring their business around AI, they're probably will be out of business. So this is a great time to be in the field. So thank you for sharing your insights here in theCUBE. >> Great. Thank you very much, John. Always a pleasure talking to you. Always have a lot of fun. And we both speak really fast, I can tell, you know, so. (both laughing) >> I know. We'll not the transcript available, we'll integrate it into our CubeGPT model that we have Luis. >> That's right. >> Great. >> Great. >> Great to talk to you, thank you, John. Thanks, man, bye. >> Hey, this is theCUBE. I'm John Furrier, here in Palo Alto, Cube Conversation. Thanks for watching. (gentle music)

(upbeat music) >> Okay, welcome back everyone at Supercloud 2 event, live here in Palo Alto, theCUBE Studios live stage performance, virtually syndicating it all over the world. I'm John Furrier with Dave Vellante here as Cube alumni, and special influencer guest, Howie Xu, VP of Machine Learning and Zscaler, also part-time as a CUBE analyst 'cause he is that good. Comes on all the time. You're basically a CUBE analyst as well. Thanks for coming on. >> Thanks for inviting me. >> John: Technically, you're not really a CUBE analyst, but you're kind of like a CUBE analyst. >> Happy New Year to everyone. >> Dave: Great to see you. >> Great to see you, Dave and John. >> John: We've been talking about ChatGPT online. You wrote a great post about it being more like Amazon, not like Google. >> Howie: More than just Google Search. >> More than Google Search. Oh, it's going to compete with Google Search, which it kind of does a little bit, but more its infrastructure. So a clever point, good segue into this conversation, because this is kind of the beginning of these kinds of next gen things we're going to see. Things where it's like an obvious next gen, it's getting real. Kind of like seeing the browser for the first time, Mosaic browser. Whoa, this internet thing's real. I think this is that moment and Supercloud like enablement is coming. So this has been a big part of the Supercloud kind of theme. >> Yeah, you talk about Supercloud, you talk about, you know, AI, ChatGPT. I really think the ChatGPT is really another Netscape moment, the browser moment. Because if you think about internet technology, right? It was brewing for 20 years before early 90s. Not until you had a, you know, browser, people realize, "Wow, this is how wonderful this technology could do." Right? You know, all the wonderful things. Then you have Yahoo and Amazon. I think we have brewing, you know, the AI technology for, you know, quite some time. Even then, you know, neural networks, deep learning. But not until ChatGPT came along, people realize, "Wow, you know, the user interface, user experience could be that great," right? So I really think, you know, if you look at the last 30 years, there is a browser moment, there is iPhone moment. I think ChatGPT moment is as big as those. >> Dave: What do you see as the intersection of things like ChatGPT and the Supercloud? Of course, the media's going to focus, journalists are going to focus on all the negatives and the privacy. Okay. You know we're going to get by that, right? Always do. Where do you see the Supercloud and sort of the distributed data fitting in with ChatGPT? Does it use that as a data source? What's the link? >> Howie: I think there are number of use cases. One of the use cases, we talked about why we even have Supercloud because of the complexity, because of the, you know, heterogeneous nature of different clouds. In order for me as a developer, in order for me to create applications, I have so many things to worry about, right? It's a complexity. But with ChatGPT, with the AI, I don't have to worry about it, right? Those kind of details will be taken care of by, you know, the underlying layer. So we have been talking about on this show, you know, over the last, what, year or so about the Supercloud, hey, defining that, you know, API layer spanning across, you know, multiple clouds. I think that will be happening. However, for a lot of the things, that will be more hidden, right? A lot of that will be automated by the bots. You know, we were just talking about it right before the show. One of the profound statement I heard from Adrian Cockcroft about 10 years ago was, "Hey Howie, you know, at Netflix, right? You know, IT is just one API call away." That's a profound statement I heard about a decade ago. I think next decade, right? You know, the IT is just one English language away, right? So when it's one English language away, it's no longer as important, API this, API that. You still need API just like hardware, right? You still need all of those things. That's going to be more hidden. The high level thing will be more, you know, English language or the language, right? Any language for that matter. >> Dave: And so through language, you'll tap services that live across the Supercloud, is what you're saying? >> Howie: You just tell what you want, what you desire, right? You know, the bots will help you to figure out where the complexity is, right? You know, like you said, a lot of criticism about, "Hey, ChatGPT doesn't do this, doesn't do that." But if you think about how to break things down, right? For instance, right, you know, ChatGPT doesn't have Microsoft stock price today, obviously, right? However, you can ask ChatGPT to write a program for you, retrieve the Microsoft stock price, (laughs) and then just run it, right? >> Dave: Yeah. >> So the thing to think about- >> John: It's only going to get better. It's only going to get better. >> The thing people kind of unfairly criticize ChatGPT is it doesn't do this. But can you not break down humans' task into smaller things and get complex things to be done by the ChatGPT? I think we are there already, you know- >> John: That to me is the real game changer. That's the assembly of atomic elements at the top of the stack, whether the interface is voice or some programmatic gesture based thing, you know, wave your hand or- >> Howie: One of the analogy I used in my blog was, you know, each person, each professional now is a quarterback. And we suddenly have, you know, a lot more linebacks or you know, any backs to work for you, right? For free even, right? You know, and then that's sort of, you should think about it. You are the quarterback of your day-to-day job, right? Your job is not to do everything manually yourself. >> Dave: You call the play- >> Yes. >> Dave: And they execute. Do your job. >> Yes, exactly. >> Yeah, all the players are there. All the elves are in the North Pole making the toys, Dave, as we say. But this is the thing, I want to get your point. This change is going to require a new kind of infrastructure software relationship, a new kind of operating runtime, a new kind of assembler, a new kind of loader link things. This very operating systems kind of concepts. >> Data intensive, right? How to process the data, how to, you know, process so gigantic data in parallel, right? That's actually a tough job, right? So if you think about ChatGPT, why OpenAI is ahead of the game, right? You know, Google may not want to acknowledge it, right? It's not necessarily they do, you know, not have enough data scientist, but the software engineering pieces, you know, behind it, right? To train the model, to actually do all those things in parallel, to do all those things in a cost effective way. So I think, you know, a lot of those still- >> Let me ask you a question. Let me ask you a question because we've had this conversation privately, but I want to do it while we're on stage here. Where are all the alpha geeks and developers and creators and entrepreneurs going to gravitate to? You know, in every wave, you see it in crypto, all the alphas went into crypto. Now I think with ChatGPT, you're going to start to see, like, "Wow, it's that moment." A lot of people are going to, you know, scrum and do startups. CTOs will invent stuff. There's a lot of invention, a lot of computer science and customer requirements to figure out. That's new. Where are the alpha entrepreneurs going to go to? What do you think they're going to gravitate to? If you could point to the next layer to enable this super environment, super app environment, Supercloud. 'Cause there's a lot to do to enable what you just said. >> Howie: Right. You know, if you think about using internet as the analogy, right? You know, in the early 90s, internet came along, browser came along. You had two kind of companies, right? One is Amazon, the other one is walmart.com. And then there were company, like maybe GE or whatnot, right? Really didn't take advantage of internet that much. I think, you know, for entrepreneurs, suddenly created the Yahoo, Amazon of the ChatGPT native era. That's what we should be all excited about. But for most of the Fortune 500 companies, your job is to surviving sort of the big revolution. So you at least need to do your walmart.com sooner than later, right? (laughs) So not be like GE, right? You know, hand waving, hey, I do a lot of the internet, but you know, when you look back last 20, 30 years, what did they do much with leveraging the- >> So you think they're going to jump in, they're going to build service companies or SaaS tech companies or Supercloud companies? >> Howie: Okay, so there are two type of opportunities from that perspective. One is, you know, the OpenAI ish kind of the companies, I think the OpenAI, the game is still open, right? You know, it's really Close AI today. (laughs) >> John: There's room for competition, you mean? >> There's room for competition, right. You know, you can still spend you know, 50, $100 million to build something interesting. You know, there are company like Cohere and so on and so on. There are a bunch of companies, I think there is that. And then there are companies who's going to leverage those sort of the new AI primitives. I think, you know, we have been talking about AI forever, but finally, finally, it's no longer just good, but also super useful. I think, you know, the time is now. >> John: And if you have the cloud behind you, what do you make the Amazon do differently? 'Cause Amazon Web Services is only going to grow with this. It's not going to get smaller. There's more horsepower to handle, there's more needs. >> Howie: Well, Microsoft already showed what's the future, right? You know, you know, yes, there is a kind of the container, you know, the serverless that will continue to grow. But the future is really not about- >> John: Microsoft's shown the future? >> Well, showing that, you know, working with OpenAI, right? >> Oh okay. >> They already said that, you know, we are going to have ChatGPT service. >> $10 billion, I think they're putting it. >> $10 billion putting, and also open up the Open API services, right? You know, I actually made a prediction that Microsoft future hinges on OpenAI. I think, you know- >> John: They believe that $10 billion bet. >> Dave: Yeah. $10 billion bet. So I want to ask you a question. It's somewhat academic, but it's relevant. For a number of years, it looked like having first mover advantage wasn't an advantage. PCs, spreadsheets, the browser, right? Social media, Friendster, right? Mobile. Apple wasn't first to mobile. But that's somewhat changed. The cloud, AWS was first. You could debate whether or not, but AWS okay, they have first mover advantage. Crypto, Bitcoin, first mover advantage. Do you think OpenAI will have first mover advantage? >> It certainly has its advantage today. I think it's year two. I mean, I think the game is still out there, right? You know, we're still in the first inning, early inning of the game. So I don't think that the game is over for the rest of the players, whether the big players or the OpenAI kind of the, sort of competitors. So one of the VCs actually asked me the other day, right? "Hey, how much money do I need to spend, invest, to get, you know, another shot to the OpenAI sort of the level?" You know, I did a- (laughs) >> Line up. >> That's classic VC. "How much does it cost me to replicate?" >> I'm pretty sure he asked the question to a bunch of guys, right? >> Good luck with that. (laughs) >> So we kind of did some napkin- >> What'd you come up with? (laughs) >> $100 million is the order of magnitude that I came up with, right? You know, not a billion, not 10 million, right? So 100 million. >> John: Hundreds of millions. >> Yeah, yeah, yeah. 100 million order of magnitude is what I came up with. You know, we can get into details, you know, in other sort of the time, but- >> Dave: That's actually not that much if you think about it. >> Howie: Exactly. So when he heard me articulating why is that, you know, he's thinking, right? You know, he actually, you know, asked me, "Hey, you know, there's this company. Do you happen to know this company? Can I reach out?" You know, those things. So I truly believe it's not a billion or 10 billion issue, it's more like 100. >> John: And also, your other point about referencing the internet revolution as a good comparable. The other thing there is online user population was a big driver of the growth of that. So what's the equivalent here for online user population for AI? Is it more apps, more users? I mean, we're still early on, it's first inning. >> Yeah. We're kind of the, you know- >> What's the key metric for success of this sector? Do you have a read on that? >> I think the, you know, the number of users is a good metrics, but I think it's going to be a lot of people are going to use AI services without even knowing they're using it, right? You know, I think a lot of the applications are being already built on top of OpenAI, and then they are kind of, you know, help people to do marketing, legal documents, you know, so they're already inherently OpenAI kind of the users already. So I think yeah. >> Well, Howie, we've got to wrap, but I really appreciate you coming on. I want to give you a last minute to wrap up here. In your experience, and you've seen many waves of innovation. You've even had your hands in a lot of the big waves past three inflection points. And obviously, machine learning you're doing now, you're deep end. Why is this Supercloud movement, this wave of Supercloud and the discussion of this next inflection point, why is it so important? For the folks watching, why should they be paying attention to this particular moment in time? Could you share your super clip on Supercloud? >> Howie: Right. So this is simple from my point of view. So why do you even have cloud to begin with, right? IT is too complex, too complex to operate or too expensive. So there's a newer model. There is a better model, right? Let someone else operate it, there is elasticity out of it, right? That's great. Until you have multiple vendors, right? Many vendors even, you know, we're talking about kind of how to make multiple vendors look like the same, but frankly speaking, even one vendor has, you know, thousand services. Now it's kind of getting, what Kid was talking about what, cloud chaos, right? It's the evolution. You know, the history repeats itself, right? You know, you have, you know, next great things and then too many great things, and then people need to sort of abstract this out. So it's almost that you must do this. But I think how to abstract this out is something that at this time, AI is going to help a lot, right? You know, like I mentioned, right? A lot of the abstraction, you don't have to think about API anymore. I bet 10 years from now, you know, IT is one language away, not API away. So think about that world, right? So Supercloud in, in my opinion, sure, you kind of abstract things out. You have, you know, consistent layers. But who's going to do that? Is that like we all agreed upon the model, agreed upon those APIs? Not necessary. There are certain, you know, truth in that, but there are other truths, let bots take care of, right? Whether you know, I want some X happens, whether it's going to be done by Azure, by AWS, by GCP, bots will figure out at a given time with certain contacts with your security requirement, posture requirement. I'll think that out. >> John: That's awesome. And you know, Dave, you and I have been talking about this. We think scale is the new ratification. If you have first mover advantage, I'll see the benefit, but scale is a huge thing. OpenAI, AWS. >> Howie: Yeah. Every day, we are using OpenAI. Today, we are labeling data for them. So you know, that's a little bit of the- (laughs) >> John: Yeah. >> First mover advantage that other people don't have, right? So it's kind of scary. So I'm very sure that Google is a little bit- (laughs) >> When we do our super AI event, you're definitely going to be keynoting. (laughs) >> Howie: I think, you know, we're talking about Supercloud, you know, before long, we are going to talk about super intelligent cloud. (laughs) >> I'm super excited, Howie, about this. Thanks for coming on. Great to see you, Howie Xu. Always a great analyst for us contributing to the community. VP of Machine Learning and Zscaler, industry legend and friend of theCUBE. Thanks for coming on and sharing really, really great advice and insight into what this next wave means. This Supercloud is the next wave. "If you're not on it, you're driftwood," says Pat Gelsinger. So you're going to see a lot more discussion. We'll be back more here live in Palo Alto after this short break. >> Thank you. (upbeat music)