>> Announcer: Live from Las Vegas, it's theCUBE! Covering AWS re:Invent 2019. Brought to you by Amazon Web Services and Intel, along with its ecosystem partners. >> Hey welcome back, everyone, it's theCUBE's coverage here live in Las Vegas for re:Invent 2019, this is theCUBE's seventh year covering re:Invent, the event's only been going for eight years, it feels like a decade, so much growth, so much action, I'm John Furrier with my co-host Dave Vellante, here extracting the signal from the noise in the Intel AWS studio of theCUBE, thank you for that sponsorship. Mike Miller is our next guest, he's director of AI devices at AWS, super excited for this segment, because DeepRacer's here, and we got some music, AI is the front and center, great to see you again, thanks for coming on. >> Absolutely, thank you for having me on again, I appreciate it. >> All right, let's just jump right in, the toys. Developers are geeking out over DeepRacer and the toys you guys are putting out there as a fun way to play and learn. >> Absolutely, getting hands-on with these new broadly applicable machine learning technologies. >> Let's jump into DeepRacer, so first of all, give us a quick update on what's happened between last year and this year in the DeepRacer community, there's been a lot of froth, competitiveness, street battles, and then we'll get an update, give us a quick update on the community. >> So we launched DeepRacer last year as a 1/18 scale race car designed to teach reinforcement learning, so this thing drives by itself around the tracks. We've got an online experience where customers can train models, so we launched a DeepRacer league where we plan to visit 22 sites around the world at AWS summits, where developers can come visit us and race a car physically around a track, and we had online contests, so every month we had a new track for developers to be challenged by and race their cars around the track. We've seen tremendous engagement and excitement, a little bit of competition really gets developers' juices going. >> It's been a lot of fun, congratulations, by the way. >> Absolutely, thank you. >> All right, let's get into the new toy, so DeepRacer 2.0, whatever you're calling it, just DeepRacer-- >> DeepRacer Evo. >> Evo, okay. >> New generation, so we've basically provided more opportunities to race for developers, more challenges for them to learn, and more ways for them to win. So we integrated some new sensors on this car, so on top there's a LIDAR, which is a laser range finding device that can detect other cars or obstacles in the rear of the car and to the sides, and in the front of the car we have stereo cameras that we added so that the car can sense depth in front of it, so with those new sensors, developers can now be challenged by integrating depth sensing and object avoidance and head to head racing into their machine learning models. >> So currently it's not an obstacle course, correct, it's a race track, right? >> So we call it a time trial, so it's a single car on the track at a time, how fast can you make a lap, our world record actually is 7.44 seconds, set by a young lady from Tokyo this past year, really exciting. >> And she was holding up the trophy and said this is basically a dream come true. And so, what are they trying to optimize, is it just the speed at the turn, what are they sort of focused on? >> Yeah, it's a little bit of art and a little bit of science, so there's the reinforcement learning model that learns through what's called a reward function, so you give the car rewards for achieving specific objectives, or certain behaviors, and so it's really up to the developer to decide what kind of behaviors do they want to reward the car with, whether it's stay close to the center line, reduce the amount of turns, they can also determine its position on the track and so they can reward it for cutting corners close, speeding up or slowing down, so it's really a little bit of art and science through some experimentation and deciding. >> So we had Intel on yesterday, talking about some of their AI, Naveen Rao, great guy, but they were introducing this concept called GANs, Generative Adversarial Networks, which is kind of like neural network technology, lot of computer science in some of the tech here, this is not kiddie scripting kind of thing, this is like real deal. >> Yeah, so GANs actually formed the basis of the product that we just announced this year called DeepComposer, so DeepComposer is a keyboard and a cloud service designed to work together to teach developers about generative AI, and GANs are the technique that we teach developers. So what's interesting about generative AI is that machine learning moves from a predictions-based technology to something that can actually create new content, so create new music, new stories, new art, but also companies are using generative AI to do more practical things like take a sketch and turn it into a 3D model, or autocorrect colorize black and white photos, Autodesk even has a generative design product, where you can give, an industrial designer can give a product some constraints and it'll generate hundreds of ideas for the design. >> Now this is interesting to me, because I think this takes it to, I call basic machine learning, to really some more advanced practical examples, which is super exciting for people learning AI and machine learning. Can you talk about the composer and how it works, because pretend I'm just a musician, I'm 16 years old, I'm composing music, I got a keyboard, how can I get involved, what would be a path, do I buy a composer device, do I link it to Ableton Live, and these tools that are out there, there's a variety of different techniques, can you take us through the use case? >> Yeah, so really our target customer for this is an aspiring machine learning developer, maybe not necessarily a musician. So any developer, whether they have musical experience or machine learning background, can use the DeepComposer system to learn about the generative AI techniques. So GANs are comprised of these two networks that have to be trained in coordination, and what we do with DeepComposer is we walk users through or walk developers through exactly how to set up that structure, how these two things train, and how is it different from traditional machine learning where you've got a large data set, and you're training a single model to make a prediction. How do these multiple networks actually work against each other, and how do you make sure that they're generating new content that's actually of the right type of quality that you want, and so that's really the essence of the Generative Adversarial Networks and these two networks that work against each other. >> So a young musician who happens to like machine learning. >> So if I give this to my kid, he'll get hooked on machine learning? That's good for the college apps. >> Plug in his Looper and set two systems working together or against each other. >> When we start getting to visualization, that's going to be very interesting when you start getting the data at the fundamental level, now this is early days. Some would say day zero, because this is really early. How do you explain that to developers, and people you're trying to get attention to, because this is certainly exciting stuff, it's fun, playful, but it's got some nerd action in it, it's got some tech, what are some of the conversations you're having with folks when they say "Hey, how do I get involved, why should I get involved," and what's really going to be the impact, what's the result of all this? >> Yeah, well it's fascinating because through Amazon's 20 years of artificial intelligence investments, we've learned a lot, and we've got thousands of engineers working on artificial intelligence and machine learning, and what we want to do is try to take a lot of that knowledge and the experiences that those folks have learned through these years, and figure out how we can bring them to developers of all skill levels, so developers who don't know machine learning, through developers who might be data scientists and have some experience, we want to build tools that are engaging and tactile and actually tangible for them to learn and see the results of what machine learning can do, so in the DeepComposer case it's how do these generative networks actually create net new content, in this case music. For DeepRacer, how does reinforcement learning actually translate from a simulated environment to the real world, and how might that be applicable for, let's say, robotics applications? So it's really about reducing the learning curve and making it easy for developers to get started. >> But there is a bridge to real world applications in all this, it's a machine learning linchpin. >> Absolutely, and you can just look at all of the innovations that are being done from Amazon and from our customers, whether they're based on improving product recommendations, forecasting, streamlining supply chains, generating training data, all of these things are really practical applications. >> So what's happening at the device, and what's happening in the cloud, can you help us understand that? >> Sure, so in DeepComposer, the device is really just a way to input a signal, and in this case it's a MIDI signal, so MIDI is a digital audio format that allows machines to kind of understand music. So the keyboard allows you to input MIDI into the generative network, and then in the cloud, we've got the generative network takes that input, processes it, and then generates four-part accompaniments for the input that you provide, so say you play a little melody on the keyboard, we're going to generate a drum track, a guitar track, a keyboard track, maybe a synthesizer track, and let you play those back to hear how your input inspired the generation of this music. >> So GANs is a big deal with this. >> Absolutely, it forms the basis of the first technique that we're teaching using DeepComposer. >> All right, so I got to ask you the question that's on everyone's mind, including mine, what are some of the wackiest and/or coolest things you've seen this year with DeepComposer and DeepRacer because I can imagine developers' creativity straying off the reservation a little bit, any cool and wacky things you've seen? >> Well we've got some great stories of competitors in the DeepRacer league, so we've got father-son teams that come in and race at the New York summit, a 10 year old learning how to code with his dad. We had one competitor in the US was at our Santa Clara summit, tried again at our Atlanta summit, and then at the Chicago summit finally won a position to come back to re:Invent and race. Last year, we did the race here at re:Invent, and the winning time, the lap time, a single lap was 51 seconds, the current world record is 7.44 seconds and it's been just insane how these developers have been able to really optimize and generate models that drive this thing at incredible speeds around the track. >> I'm sure you've seen the movie Ford v Ferrari yet. You got to see that movie, because this DeepRacer, you're going to have to need a stadium soon, with eSports booming, this has got its own legs for its own business. >> Well we've got six tracks set up down at the MGM Grand Arena, so we've already got the arena set up, and that's where we're doing all the knock-out rounds and competitors. >> And you mentioned father-son, you remember when we were kids, Cub Scouts, I think it was, or Boy Scouts, whatever it was, you had the pinewood derby, right, you'd make a car and file down the nails that you use for the axles and, taking it to a whole new level here. >> It's a modern-day version. >> All right, Mike, thanks for coming on, appreciate it, let's keep in touch. If you can get us some of that B-roll for any video, I'd love to get some B-roll of some DeepRacer photos, send 'em our way, super excited, love what you're doing, I think this is a great way to make it fun, instructive, and certainly very relevant. >> Absolutely, that's what we're after. Thank you for having me. >> All right, theCUBE's coverage here, here in Las Vegas for our seventh, Amazon's eighth re:Invent, we're documenting history as the ecosystem evolves, as the industry wave is coming, IoT edge, lot of cool things happening, we're bringing it to you, we're back with more coverage after this short break. (techno music)

>> Announcer: Live from Las Vegas, it's theCUBE! Covering AWS re:Invent 2019. Brought to you by Amazon Web Services and Intel, along with its ecosystem partners. >> Welcome back to the Sands Convention Center in Las Vegas everybody, you're watching theCUBE, the leader in live tech coverage. My name is Dave Vellante, I'm here with my cohost Justin Warren, this is day one of our coverage of AWS re:Invent 2019, Naveen Rao here, he's the corporate vice president and general manager of artificial intelligence, AI products group at Intel, good to see you again, thanks for coming to theCUBE. >> Thanks for having me. >> Dave: You're very welcome, so what's going on with Intel and AI, give us the big picture. >> Yeah, I mean actually the very big picture is I think the world of computing is really shifting. The purpose of what a computer is made for is actually shifting, and I think from its very conception, from Alan Turing, the machine was really meant to be something that recapitulated intelligence, and we took sort of a divergent path where we built applications for productivity, but now we're actually coming back to that original intent, and I think that hits everything that Intel does, because we're a computing company, we supply computing to the world, so everything we do is actually impacted by AI, and will be in service of building better AI platforms, for intelligence at the edge, intelligence in the cloud, and everything in between. >> It's really come full circle, I mean, when I first started this industry, AI was the big hot topic, and really, Intel's ascendancy was around personal productivity, but now we're seeing machines replacing cognitive functions for humans, that has implications for society. But there's a whole new set of workloads that are emerging, and that's driving, presumably, different requirements, so what do you see as the sort of infrastructure requirements for those new workloads, what's Intel's point of view on that? >> Well, so maybe let's focus that on the cloud first. Any kind of machine learning algorithm typically has two phases to it, one is called training or learning, where we're really iterating over large data sets to fit model parameters. And once that's been done to a satisfaction of whatever performance metrics that are relevant to your application, it's rolled out and deployed, that phase is called inference. So these two are actually quite different in their requirements in that inference is all about the best performance per watt, how much processing can I shove into a particular time and power budget? On the training side, it's much more about what kind of flexibility do I have for exploring different types of models, and training them very very fast, because when this field kind of started taking off in 2014, 2013, typically training a model back then would take a month or so, those models now take minutes to train, and the models have grown substantially in size, so we've still kind of gone back to a couple of weeks of training time, so anything we can do to reduce that is very important. >> And why the compression, is that because of just so much data? >> It's data, the sheer amount of data, the complexity of data, and the complexity of the models. So, very broad or a rough categorization of the complexity can be the number of parameters in a model. So, back in 2013, there were, call it 10 million, 20 million parameters, which was very large for a machine learning model. Now they're in the billions, one or two billion is sort of the state of the art. To give you bearings on that, the human brain is about a three to 500 trillion model, so we're still pretty far away from that. So we got a long way to go. >> Yeah, so one of the things about these models is that once you've trained them, that then they do things, but understanding how they work, these are incredibly complex mathematical models, so are we at a point where we just don't understand how these machines actually work, or do we have a pretty good idea of, "No no no, when this model's trained to do this thing, "this is how it behaves"? >> Well, it really depends on what you mean by how much understanding we have, so I'll say at one extreme, we trust humans to do certain things, and we don't really understand what's happening in their brain. We trust that there's a process in place that has tested them enough. A neurosurgeon's cutting into your head, you say you know what, there's a system where that neurosurgeon probably had to go through a ton of training, be tested over and over again, and now we trust that he or she is doing the right thing. I think the same thing is happening in AI, some aspects we can bound and say, I have analytical methods on how I can measure performance. In other ways, other places, it's actually not so easy to measure the performance analytically, we have to actually do it empirically, which means we have data sets that we say, "Does it stand up to all the different tests?" One area we're seeing that in is autonomous driving. Autonomous driving, it's a bit of a black box, and the amount of situations one can incur on the road are almost limitless, so what we say is, for a 16 year old, we say "Go out and drive," and eventually you sort of learn it. Same thing is happening now for autonomous systems, we have these training data sets where we say, "Do you do the right thing in these scenarios?" And we say "Okay, we trust that you'll probably "do the right thing in the real world." >> But we know that Intel has partnered with AWS, I ran autonomous driving with their DeepRacer project, and I believe it's on Thursday is the grand final, it's been running for, I think it was announced on theCUBE last year, and there's been a whole bunch of competitions running all year, basically training models that run on this Intel chip inside a little model car that drives around a race track, so speaking of empirical testing of whether or not it works, lap times gives you a pretty good idea, so what have you learned from that experience, of having all of these people go out and learn how to use these ALM models on a real live race car and race around a track? >> I think there's several things, I mean one thing is, when you turn loose a number of developers on a competitive thing, you get really interesting results, where people find creative ways to use the tools to try to win, so I always love that process, I think competition is how you push technology forward. On the tool side, it's actually more interesting to me, is that we had to come up with something that was adequately simple, so that a large number of people could get going on it quickly. You can't have somebody who spends a year just getting the basic infrastructure to work, so we had to put that in place. And really, I think that's still an iterative process, we're still learning what we can expose as knobs, what kind of areas of innovation we allow the user to explore, and where we sort of walk it down to make it easy to use. So I think that's the biggest learning we get from this, is how I can deploy AI in the real world, and what's really needed from a tool chain standpoint. >> Can you talk more specifically about what you guys each bring to the table with your collaboration with AWS? >> Yeah, AWS has been a great partner. Obviously AWS has a huge ecosystem of developers, all kinds of different developers, I mean web developers are one sort of developer, database developers are another, AI developers are yet another, and we're kind of partnering together to empower that AI base. What we bring from a technological standpoint are of course the hardware, our CPUs, our AI ready now with a lot of software that we've been putting out in the open source. And then other tools like OpenVINO, which make it very easy to start using AI models on our hardware, and so we tie that in to the infrastructure that AWS is building for something like DeepRacer, and then help build a community around it, an ecosystem around it of developers. >> I want to go back to the point you were making about the black box, AI, people are concerned about that, they're concerned about explainability. Do you feel like that's a function of just the newness that we'll eventually get over, and I mean I can think of so many examples in my life where I can't really explain how I know something, but I know it, and I trust it. Do you feel like it's sort of a tempest in a teapot? >> Yeah, I think it depends on what you're talking about, if you're talking about the traceability of a financial transaction, we kind of need that maybe for legal reasons, so even for humans we do that. You got to write down everything you did, why did you do this, why'd you do that, so we actually want traceability for humans, even. In other places, I think it is really about the newness. Do I really trust this thing, I don't know what it's doing. Trust comes with use, after a while it becomes pretty straightforward, I mean I think that's probably true for a cell phone, I remember the first smartphones coming out in the early 2000s, I didn't trust how they worked, I would never do a credit card transaction on 'em, these kind of things, now it's taken for granted. I've done it a million times, and I never had any problems, right? >> It's the opposite in social media, most people. >> Maybe that's the opposite, let's not go down that path. >> I quite like Dr. Kate Darling's analogy from MIT lab, which is we already we have AI, and we're quite used to them, they're called dogs. We don't fully understand how a dog makes a decision, and yet we use 'em every day. In a collaboration with humans, so a dog, sort of replace a particular job, but then again they don't, I don't particularly want to go and sniff things all day long. So having AI systems that can actually replace some of those jobs, actually, that's kind of great. >> Exactly, and think about it like this, if we can build systems that are tireless, and we can basically give 'em more power and they keep going, that's a big win for us. And actually, the dog analogy is great, because I think, at least my eventual goal as an AI researcher is to make the interface for intelligent agents to be like a dog, to train it like a dog, reinforce it for the behaviors you want and keep pushing it in new directions that way, as opposed to having to write code that's kind of esoteric. >> Can you talk about GANs, what is GANs, what's it stand for, what does it mean? >> Generative Adversarial Networks. What this means is that, you can kind of think of it as, two competing sides of solving a problem. So if I'm trying to make a fake picture of you, that makes it look like you have no hair, like me, you can see a Photoshop job, and you can kind of tell, that's not so great. So, one side is trying to make the picture, and the other side is trying to guess whether it's fake or not. We have two neural networks that are kind of working against each other, one's generating stuff, and the other one's saying, is it fake or not, and then eventually you keep improving each other, this one tells that one "No, I can tell," this one goes and tries something else, this one says "No, I can still tell." The one that's trying with a discerning network, once it can't tell anymore, you've kind of built something that's really good, that's sort of the general principle here. So we basically have two things kind of fighting each other to get better and better at a particular task. >> Like deepfakes. >> I use that because it is relevant in this case, and that's kind of where it came from, is from GANs. >> All right, okay, and so wow, obviously relevant with 2020 coming up. I'm going to ask you, how far do you think we can take AI, two part question, how far can we take AI in the near to mid term, let's talk in our lifetimes, and how far should we take it? Maybe you can address some of those thoughts. >> So how far can we take it, well, I think we often have the sci-fi narrative out there of building killer machines and this and that, I don't know that that's actually going to happen anytime soon, for several reasons, one is, we build machines for a purpose, they don't come from an embattled evolutionary past like we do, so their motivations are a little bit different, say. So that's one piece, they're really purpose-driven. Also, building something that's as general as a human or a dog is very hard, and we're not anywhere close to that. When I talked about the trillions of parameters that a human brain has, we might be able to get close to that from a engineering standpoint, but we're not really close to making those trillions of parameters work together in such a coherent way that a human brain does, and efficient, human brain does that in 20 watts, to do it today would be multiple megawatts, so it's not really something that's easily found, just laying around. Now how far should we take it, I look at AI as a way to push humanity to the next level. Let me explain what that means a little bit. Simple equation I always sort of write down, is people are like "Radiologists aren't going to have a job." No no no, what it means is one radiologist plus AI equals 100 radiologists. I can take that person's capabilities and scale it almost freely to millions of other people. It basically increases the accessibility of expertise, we can scale expertise, that's a good thing. It makes, solves problems like we have in healthcare today. All right, that's where we should be going with this. >> Well a good example would be, when, and probably part of the answer's today, when will machines make better diagnoses than doctors? I mean in some cases it probably exists today, but not broadly, but that's a good example, right? >> It is, it's a tool, though, so I look at it as more, giving a human doctor more data to make a better decision on. So, what AI really does for us is it doesn't limit the amount of data on which we can make decisions, as a human, all I can do is read so much, or hear so much, or touch so much, that's my limit of input. If I have an AI system out there listening to billions of observations, and actually presenting data in a form that I can make better decisions on, that's a win. It allows us to actually move science forward, to move accessibility of technologies forward. >> So keeping the context of that timeframe I said, someday in our lifetimes, however you want to define that, when do you think that, or do you think that driving your own car will become obsolete? >> I don't know that it'll ever be obsolete, and I'm a little bit biased on this, so I actually race cars. >> Me too, and I drive a stick, so. >> I kind of race them semi-professionally, so I don't want that to go away, but it's the same thing, we don't need to ride horses anymore, but we still do for fun, so I don't think it'll completely go away. Now, what I think will happen is that commutes will be changed, we will now use autonomous systems for that, and I think five, seven years from now, we will be using autonomy much more on prescribed routes. It won't be that it completely replaces a human driver, even in that timeframe, because it's a very hard problem to solve, in a completely general sense. So, it's going to be a kind of gentle evolution over the next 20 to 30 years. >> Do you think that AI will change the manufacturing pendulum, and perhaps some of that would swing back to, in this country, anyway, on-shore manufacturing? >> Yeah, perhaps, I was in Taiwan a couple of months ago, and we're actually seeing that already, you're seeing things that maybe were much more labor-intensive before, because of economic constraints are becoming more mechanized using AI. AI as inspection, did this machine install this thing right, so you have an inspector tool and you have an AI machine building it, it's a little bit like a GAN, you can think of, right? So this is happening already, and I think that's one of the good parts of AI, is that it takes away those harsh conditions that humans had to be in before to build devices. >> Do you think AI will eventually make large retail stores go away? >> Well, I think as long as there are humans who want immediate satisfaction, I don't know that it'll completely go away. >> Some humans enjoy shopping. >> Naveen: Some people like browsing, yeah. >> Depends how fast you need to get it. And then, my last AI question, do you think banks, traditional banks will lose control of the payment systems as a result of things like machine intelligence? >> Yeah, I do think there are going to be some significant shifts there, we're already seeing many payment companies out there automate several aspects of this, and reducing the friction of moving money. Moving money between people, moving money between different types of assets, like stocks and Bitcoins and things like that, and I think AI, it's a critical component that people don't see, because it actually allows you to make sure that first you're doing a transaction that makes sense, when I move from this currency to that one, I have some sense of what's a real number. It's much harder to defraud, and that's a critical element to making these technologies work. So you need AI to actually make that happen. >> All right, we'll give you the last word, just maybe you want to talk a little bit about what we can expect, AI futures, or anything else you'd like to share. >> I think it's, we're at a really critical inflection point where we have something that works, basically, and we're going to scale it, scale it, scale it to bring on new capabilities. It's going to be really expensive for the next few years, but we're going to then throw more engineering at it and start bringing it down, so I start seeing this look a lot more like a brain, something where we can start having intelligence everywhere, at various levels, very low power, ubiquitous compute, and then very high power compute in the cloud, but bringing these intelligent capabilities everywhere. >> Naveen, great guest, thanks so much for coming on theCUBE. >> Thank you, thanks for having me. >> You're really welcome, all right, keep it right there everybody, we'll be back with our next guest, Dave Vellante for Justin Warren, you're watching theCUBE live from AWS re:Invent 2019. We'll be right back. (techno music)

(upbeat music) >> Hi, I'm Peter Burris, and welcome once again to another Cube Conversation from our studios here in beautiful Palo Alto, California. Today, we've got a really special guest. We're going to be talking about AI and some of the new technologies that are making that even more valuable to business. And we're speaking with Roy Kim, who's the lead for AI solutions at Pure Storage. Roy, welcome to theCUBE. >> Thank you for having me, very excited. >> Well, so let's start by just, how does one get to be a lead for AI solutions? Tell us a little bit about that. >> Well, first of all, there aren't that many AI anything in the world today. But I did spend eight years at Nvidia, helping build out their AI practice. I'm fairly new to Storage, I'm about 11 months into Pure Storage, so, that's how you get into it, you cut your teeth on real stuff, and start at Nvidia. >> Let's talk about some real stuff, I have a thesis, I (mumbles) it by you and see what you think about it. The thesis that I have: Wikibon has been at the vanguard of talking about the role that flash is going to play, flash memory, flash storage systems, are going to play in changes in the technology industry. We were one of the first to really talk about it. And well, we believe, I believe, very strongly that if you take a look at all the changes that are happening today with AI and the commercialization of AI and even big data and some other things that are happening, a lot of that can be traced back directly to the transition from memory, which had very very long lag times, millisecond speed lag times, to flash, which is microsecond speed. And, when you go to microsecond, you can just do so much more with data, and it just seems as though that transition from disk to flash has kind of catalyzed a lot of this change, would you agree with that? >> Yeah, that transition from disk to flash was the fundamental transition within the storage industry. So the fundamental thing is that data is now fueling this whole AI revolution, and I would argue that the big data revolution with Hadoop Spark and all that is really the essence underneath it is to use data get insight. And so, disks were really fundamentally designed to store data and not to deliver data. If you think about it, the way that it's designed, it's really just to store as much data as possible. Flash is the other way around, it's to deliver data as fast as possible. That transition is fundamentally the reason why this is happening today. >> Well, it's good to be right. (laughs) >> Yeah, you are definitely right. >> So, the second observation I would make is that we're seeing, and it makes perfect sense, a move to start, or trend to start, move more processing closer to the data, especially, as you said, on flash systems that are capable of delivering data so much faster. Is that also starting to happen, in you experience? >> That's right. So this idea that you take a lot of this data and move it to compute as fast as possible-- >> Peter: Or move the compute even closer to the data. >> And the reason for that, and AI really exposes that as much as possible because AI is this idea that you have these really powerful processors that need as much data as quickly as possible to turn that around into neural networks that give you insight. That actually leads to what I'll be talking about, but the thing that we built, this thing called AIRI, this idea that you pull compute, and storage, and networking all into this compact design so there is no bottleneck, that data lives close to compute, and delivers that fastest performance for your neural network training. >> Let's talk about that a little bit. If we combine your background at Nvidia, the fact that you're currently at Pure, the role that flash plays in delivering data faster, the need for that faster delivery in AI applications, and now the possibility of moving GPUs and related types of technology even closer to the data. You guys have created a partnership with Nvidia, what exactly, tell us a little bit more about AIRI. >> Right, so, this week we announced AIRI. AIRI is the industry's first AI complete platform for enterprises. >> Peter: AI Ready-- >> AI Ready Infrastructure for enterprises, that's where AIRI comes from. It really brought Nvidia and Pure together because we saw a lot of these trends within customers that are really cutting their teeth in building an infrastructure, and it was hard. There's a lot of intricate details that go into building AI infrastructure. And, we have lots of mutual customers at Nvidia, and we found is that there some best practices that we can pull into a single solution, whether it's hardware and software, so that the rest of the enterprises can just get up and running quickly. And that is represented in AIRI. >> We know it's hard because if it was easy it would've been done a long time ago. So tell us a little bit about, specifically about the types of technologies that are embedded within AIRI. How does it work? >> So, if you think about what's required to build deep learning and AI practice, you start from data scientists, and you go into frameworks like TensorFlow and PyTorch, you may have heard of them, then you go into the tools and then GPUs, InfiniBand typically is networking of choice, and then flash, right? >> So these are all the components, all these parts that you have access to. >> That's right, that's right. And so enterprises today, they have to build all of this together by hand to get their data centers ready for AI. What AIRI represents everything but data scientists, so start from the tools like TensorFlow all the way down to flash, all built and tuned into a single solution so that all, really, enterprises need to do is give it to a data scientist and to get up and running. >> So, we've done a fair amount of research on this at Wikibon, and we discovered that one of the reasons why big data and AI-related projects have not been as successful as they might have been, is precisely because so much time was spent trying to understand the underlying technologies in the infrastructure required to process it. And, even though it was often to procure this stuff, it took a long time to integrate, a long time to test, a long time to master before you could bring application orientations to bear on the problems. What you're saying is you're slicing all that off so that folks that are trying to do artificial intelligence related workloads can have a much better time-to-value. Have I got that right? >> That's right. So, think about, just within that stack, everything I just talked about InfiniBand. Enterprises are like, "What is InfiniBand?" GPU, a lot of people know what GPU is, but enterprises will say that they've never deployed GPUs. Think about TensorFlow or PyTorch, these are tools that are necessary to data scientists, but enterprises are like, "Oh, my goodness, what is that?" So, all of this is really foreign to enterprises, and they're spending months and months trying to figure out what it is, and how to deploy it, how to design it, and-- >> How to make it work together. >> How to make it work together. And so, what Nvidia and Pure decided to do is take all the learnings that we had from these pioneers, trailblazers within the enterprise industry, bring all those best practices into a single solution, so that enterprises don't have to worry about InfiniBand, or ethernet, or GPUs, or scale out flash, or TensorFlow. It just works. >> So, it sounds like it's a solution that's specifically designed and delivered to increase the productivity of data scientists as they try to do data science. So, tell us a little bit about some of those impacts. What kinds of early insights about more productivity with data science are you starting to see as a consequence of this approach. >> Yeah, you know, you'll be surprised that most data scientists doing AI today, when they kick off a job, it takes a month to finish. So think about that. When someone, I'm a data scientist, I come in on Monday, early February, I kick off a job, I go on vacation for four weeks, I come back and it's still running. >> What do you mean by "kicking off a job?" >> It means I start this workload that helps train neural nets, right? It requires GPUs to start computing, and the TensorFlow to work, and the data to get it consumed. >> You're talking about, it takes weeks to run a job that does relatively simple things in a data science sense, like train a model. >> Train a model, takes a month. And so, the scary thing about that is you really have 12 tries a year to get it right. Just imagine that. And that's not something that we want enterprises to suffer through. And so, what AIRI does, it cuts what used to take a month down to a week. Now, that's amazing, if you think about it. What used to, they only had 12 tries in a year, now they have 48 tries in a year. Transformative, right? The way that that worked is we, in AIRI, if you look at it there's actually four servers with FlashBlade. We figured out a way to have that job run across all four servers to give you 4X the throughput. Think that that's easy to do, but it actually is not. >> So you parallelized it. >> We parallelized it. >> And that is not necessarily easy to do. These are often not particularly simple jobs. >> But, that's why no one's doing it today. >> But, if you think about it, going back to your point, it's like the individual who takes performance-enhancement drugs so they can get one more workout than the competition and that lets them hit another 10, 15 home runs which leads to millions of extra dollars. You're kind of saying something similar. You used to be able to get only 12 workouts a year, now you can do 48 workouts, which business is going to be stronger and more successful as a result. >> That's a great analogy. Another way to look at it is, a typical data scientist probably makes about half a million dollars a year. What if you get 4X the productivity out of that person? So, you get the return of two million dollars in return, out of that $500,000 investment you make. That's another way of saying performance-enhancing drug for that data scientist. >> But I honestly think it's even more than that. Because, there's a lot of other support staff that are today, doing a lot of the data science grunt work, let's call it. Lining up the pipelines, building the, testing pipelines, making sure that they run, testing sources, testing sinks. And, this is reducing the need for infrastructure types of tasks. So, you're getting more productivity out of the data scientitists, but you're also getting more productivity out of all the people who heretofore were, you were spending on doing this type of stuff, when all they were doing was just taking care of the infrastructure. >> Yeah. >> Is that right? >> That's exactly right. We have a customer in the UK, one of the world's largest hedge fund companies that's publicly traded. And, what they told us is that, with FlashBlade, and not necessarily an AIRI customer at this time, but they're actually doing AI with FlashBlade today at Pure, from Pure. What they said is, with FlashBlade they actually got two engineers that were full time taking care of infrastructure, now they're doing data science. Right? To your point, that they don't have to worry about infrastructure anymore, because the simplicity of what we bring from Pure. And so now they're working on models to help them make more money. >> So the half a million dollars a year that you were spending on a data scientist and a couple of administrators, that you were getting two million dollars worth, that you're now getting two million dollars return, you can now take those administrators and have them start doing more data science, without necessarily paying them more. It's a little secret. But you're now getting four, five, six million dollars in return as a consequence of this system. >> That's right. >> As we think about where AIRI is now, and you think about where it's going to go, give us a sense of, kind of, how this presages new approaches to thinking about problem solving as it relates to AI and other types of things. >> One of the beauty about AI is that it's always evolving. What used to be what they call CNNs as the most popular model, now is GANs, which-- >> CNN stands for? >> Convolution Neural Nets. Typically used for image processing. Now, people are using things like Generative Adversarial Networks, which is putting two networks against each other to-- >> See which one works and is more productive. >> And so, that happened in a matter of a couple of years. AI's always changing, always evolving, always getting better and so it really gives us an opportunity to think about how does AIRI evolve to keep up and bring the best, state of the art technology to the data scientist. There's actually boundless opportunities to-- >> Well, even if you talk about GANs, or Generative Adversarial Networks, the basic algorithms have been in place for 15, 20, maybe even longer, 30 years. But, the technology wouldn't allow it to work. And so, really what we're talking about is a combination of deep understanding of how some of these algorithms work, that's been around for a long time, and the practical ability to get business value out of them. And that's kind of why this is such an exploding thing, because there's been so much knowledge about how this stuff, or what this stuff could do, that now we can actually apply it to some of these complex business problems. >> That's exactly right. I tell people that the promise of big data has been around for a long time. People have been talking about big data for 10, 20 years. AI is really the first killer application of big data. Hadoop's been around for a really long time, but we know that people have struggled with Hadoop. Spark has been great but what AI does is it really taps into the big data platform and translates that into insight. And whatever the data is. Video, text, all kinds of data can, you can use AI on. That really is the reason why there's a lot of excitement around AI. It really is the first killer application for big data. >> I would say it's even more than that. It's an application, but it's also, we think there's a bifurcation, we think that we're seeing an increased convergence inside the infrastructure, which is offering up greater specialization in AI. So, AI as an application, but it also will be the combination of tooling, especially for data scientists, will be the new platform by which you build these new classes of applications. You won't even know you're using AI, you'll just build an application that has those capabilities, right? >> Right, that's right, I mean I think it's as technical as that or as simple as when you use your iPhone and you're talking to Siri, you don't know that you're talking to AI, it's just part of your daily life. >> Or, looking at having it recognize your face. I mean, that is processing, the algorithms have been in place for a long time, but it was only recently that we had the hardware that was capable of doing it. And Pure Storage is now bringing a lot of that to the enterprise through this relationship with Nvidia. >> That's right, so AIRI does represent all the best of AI infrastructure from all our customers, we pulled it into what AIRI is, and we're both really excited to give it to all our customers. >> So, I guess it's a good time to be the lead for AI solutions at Pure Storage, huh? >> (laughs) That's right. There's a ton of work, but a lot of excitement. You know, this is really the first time a storage company was spotlighted and became, and went on the grand stage of AI. There's always been Nvidia, there's always been Google, Facebook, and Hyperscalers, but when was the last time a storage company was highlighted on the grand stage of AI? >> Don't think it will be the last time, though. >> You know, it's to your point that this transition from disk to flash is that big transition in industry. And fate has it that Pure Storage has the best flash-based solution for deep learning. >> So, I got one more question for you. So, we've got a number of people that are watching the video, watching us talk, a lot of them very interested in AI, trying to do AI, you've got a fair amount of experience. What are the most interesting problems that you think we should be focusing on with AI? >> Wow, that's a good one. Well, there's so many-- >> Other than using storage better. >> (laughs) Yeah, I think there's so many applications just think about customer experience, just one of the most frustrating things for a lot of people is when they dial in and they have to go through five different prompts to get to the right person. That area alone could use a lot of intelligence in the system. I think, by the time they actually speak to a real live person, they're just frustrated and the customer experience is poor. So, that's one area I know that there's a lot of research in how does AI enhance that experience. In fact, one of our customers is Global Response, and they are a call center services company as well as an off-shoring company, and they're doing exactly that. They're using AI to understand the sentiment of the caller, and give a better experience. >> All that's predicated on the ability to do the delivery. So, I'd like to see AI be used to sell AI. (Roy laughs) Alright, so Roy Kim, who's the lead of AI solutions at Pure Storage. Roy, thank you very much for being on theCUBE and talking with us about AIRI and the evolving relationship between hardware, specifically storage, and new classes of business solutions powered by AI. >> Thank you for inviting me. >> And again, I'm Peter Burris, and once again, you've been watching theCUBE, talk to you soon. (upbeat music)