(electronic music) >> Live from Orlando, Florida, it's theCUBE! Covering Microsoft Ignite. Brought to you by Cohesity, and theCUBE's ecosystem partners. >> Welcome back everyone to theCUBE's live coverage of Microsoft Ignite here in Orlando, Florida. I'm your host, Rebecca Knight, along with my cohost, Stu Miniman. We're joined by Jeffrey Snover. He is the technical fellow and chief architect for Azure Storage and Cloud Edge at Microsoft. Thanks so much for coming, for returning to theCUBE, I should say, Jeffrey, you're a CUBE alum. >> Yes, I enjoyed the last time. So can't wait to do it again this time. >> Well we're excited to have you. So before the camera's were rolling, we were talking about PowerShell. You invented PowerShell. >> Yeah, I did. >> It was invented in the early 2000's, it took a few years to ship, as you said. But can you give our viewers an update of where we are? >> Yeah, you know, it's 2018, and it's never been a better time for PowerShell. You know, basically the initial mission is sort of complete. And the mission was provide sort of general purpose scripting for Windows. But now we have a new mission. And that new mission is to manage anything, anywhere. So we've taken PowerShell, we've open sourced it. It's now running, we've ported it to macOS and Linux. There's a very large list of Linux distributions that we support it on, and it runs everywhere. And so, now, you can manage from anywhere. Your Windows box, your Linux box, your Mac box, even in the browser, you can manage, and then anything. You can manage Windows, you can manage Linux, you can manage macOS. So manage anything, anywhere. Any cloud, Azure, or AWS, or Google. Any hypervisor, Hyper-V or VMware, or any physical server. It's amazing. In fact, our launch partners, when we launched this, our launch partners, VMware, Google, AWS. Not Microsoft's traditional partners. >> That's great to hear. It was actually, one of the critiques we had, at the key note this morning, was partnerships are critically important. But felt that Satya gave a little bit of a jab towards, the kind of, the Amazon's out there. When we talk to customers, we know it's a heterogeneous, multi-cloud world. You know, you work all over the place, with your solutions that you had. There's not, like, Azure, Azure Stack, out to The Edge. The Edge, it is early, it's going to be very heterogeneous. So connect the dots for us a little. You know, we love having the technical fellows on, as to, you go from PowerShell, to now this diverse set of solutions that you work on today. >> Yeah, exactly. So basically, from PowerShell, they asked me to be the chief architect for Windows Server. Right, because if you think about it, an operating system is largely management, right? And, so, that's what I did, resource management. And, so, I was the chief architect for that, for many years, and we decided that, as part of that, we were developing cloud-inspired infrastructure. So, basically, you know, Windows Server had grown up. You know, sort of focused in on a machine. Azure had gone and needed to build a new set of infrastructure for the cloud. And we looked at what they were doing. And they say, hey, that's some great ideas. Let's take the ideas there, and put them into the general purpose operating system. And that's what we call our software-defined data center. And the reason why we couldn't use Azure's directly is, Azure's, really, design center is very, very, very large systems. So, for instance, the storage stamp, that starts at about 10 racks. No customer wants to start with 10 racks. So we took the inspiration from them and re-implemented it. And now our systems can start with two servers. Our Azure Stack systems, well, so, then, what we decided was, hey, this is great technology. Let's take the great cloud-inspired infrastructure of Windows Server, and match it with the Azure services themselves. So we take Azure, put it on top of Windows Server, package it as an appliance experience, and we call that Azure Stack. And that's where I have been mostly focused for the last couple of years. >> Right, can you help us unpack a little bit. There's a lot of news today. >> Yes. >> You know, Windows 2019 was announced. I was real interested in the Data Box Edge solution, which I'm sure. >> Isn't that crazy? >> Yeah, really interesting. You're like, let's do some AI applications out at the Edge, and with the same kind of box that we can transport data. Because, I always say, you got to follow customers applications and data, and it's tough to move these things. You know, we've got physics that we still have to, you know, work on until some of these smart guys figure out how to break that. But, yeah, maybe give us a little context, as to news of the show, things your teams have been working on. >> Yeah, so the Data Box Edge, big, exciting stuff. Now, there's a couple scenarios for Data Box Edge. First is, first it's all kind of largely centered on storage and the Edge. So Storage, you've got a bunch of data in your enterprise, and you'd like it to be in Azure. One flavor of Data Box Edge is a disk. You call us up, we send you a disk, you fill up that disk, you send it back to us, it shows up in Azure. Next. >> A pretty big disk, though? >> Well, it can be a small disk. >> Oh, okay. >> Yeah, no, it can be a single SSD, okay. But then you can say, well, no, I need a bunch more. And so we send you a box, the box is over there. It's like 47 pounds, we send you this thing, it's about 100 terabytes of data. You fill that thing up, send it to us, and we upload it. Or a Data Box Heavy. Now this thing has a handle and wheels. I mean, literally, wheels, it's specially designed so that a forklift can pick this thing up, right? It's like, I don't know, like 400 pounds, it's crazy. And that's got about a petabyte worth of storage. Again, we ship it to you, you fill it up, ship it back to us. So that's one flavor, Data Box transport. Then there's Data Box Edge. Data Box Edge, you go to the website, say, I'd like a Data Box Edge, we send you a 1u server. You plug that in, you keep it plugged in, then you use it. How do you use it? You connect it to your Azure storage, and then all your Azure storage is available through here. And it's exposed through SMB. Later, we'll expose it through NFS and a Blob API. But, then, anything you write here is available immediately, it gets back to Azure, and, effectively, it looks like near-infinite storage. Just use it and it gets backed up, so it's amazing. Now, on that box, we're also adding the ability to say, hey, we got a bunch of compute there. You can run IoT Edge platforms. So you run the IoT Edge platform, you can run gateways, you can run Kubernetes clusters on this thing, you can run all sorts of IoT software. Including, we're integrating in brainwave technology. So, brainwave technology is, and, by the way, we'll want to talk about this a little bit, in a second. It is evidence of the largest transformation we'll see in our industry. And that is the re-integration of the industry. So, basically, what does that mean? In the past, the industry used to be, back when the big key players were digital. Remember digital, from DEC? We're all Massachusetts people. (Rebecca laughs) So, DEC was the number one employer in Massachusetts, gone. IBM dominant, much diminished, a whole bunch of people. They were dominant when the industry was vertically integrated. Vertically integrated meant all those companies designed their own silicone, they built their own boards, they built their own systems, they built their OS, they built the applications, the serviced them. Then there was the disintegration of the computer industry. Where, basically, we went vertically integrated. You got your chips from Intel or Motorola. The operating system, you got from Sun or Microsoft. The applications you got from a number of different vendors. Okay, so we got vertically integrated. What you're seeing, and what's so exciting, is a shift back to vertical integration. So Microsoft is designing its own hardware, right? We're designing our own chips. So we've designed a chip specially for AI, we call it a brainwave chip, and that's available in the Data Box Edge. So, now, when you do this AI stuff, guess what? The processing is very different. And it can be very, very fast. So that's just one example of Microsoft's innovation in hardware. >> Wow, so, I mean. >> What do you do with that? >> One of the things that we keep hearing so much, at this conference, is that Microsoft products and services are helping individual employees tap into their own creativity, their ingenuity, and then, also, collaborate with colleagues. I'm curious about where you get your ideas, and how you actually put that into practice, as a technical fellow. >> Yeah. >> How do you think about the future, and envision these next generation technologies? >> Yeah, well, you know, it's one of those things, honestly, where your strength is your weakness, your weakness is your strength. So my weakness is, I can't deal with complexity, right. And, so, what I'm always doing is I'm taking a look at a very complex situation, and I'm saying, what's the heart of it, like, give me the heart of it. So my background's physics, right? And so, in physics, you're not doing, you're looking for the F equals M A. And if you have that, when you find that, then you can apply it over, and over, and over again. So I'm always looking at what are the essential things here. And so that's this, well, you see a whole bunch of confusing things, like, what's up with this? What's with this? That idea of there is this narrative about the reintegration of the computer industry. How very large vendors, be it Microsoft, or AWS, are, because we operate at such large scales, we are going to be vertically integrated. We're developing our own hardware, we do our own systems, et cetera. So, I'm always looking for the simple story, and then applying it. And, it turns out, I do it pretty accurately. And it turns out, it's pretty valuable. >> Alright, so that's a good set up to talk about Azure Stacks. So, the value proposition we heard, of course, is, you know, start everything in the cloud first, you know, Microsoft does Azure, and then lets, you know, have some of those services in the same operating model in your data center, or in your hosting service provider environment. So, first of all, did I get that right? And, you know, give us the update on Azure Stack. I've been trying to talk to customers that are using it, talking to your partners. There is a lot of excitement around it. But, you know, proof points, early use cases, you know, where is this going to be pointing towards, where the future of the data center is? >> So, it's a great example. So what I figured out, when I thought about this, and kind of drilled in, like what's really, what really matters here? What I realized was that what the gestalt of Azure Stack is different than everything we've done in the past. And it really is an appliance, okay? So, in the past, I just had a session the other day, and people were asking, well, when are you going to, when is Azure Stack going to have the latest version of the operating system? I said, no, no, no, no, no. Internals are internal, it's an appliance. Azure Stack is for people who want to use a cloud, not for people who want to build it. So you shouldn't be concerned about all the internals. You just plug it in, fill out some forms, and then you use it, just start using it. You don't care about the details of how it's all configured, you don't do the provisioning, we do all that for you. And so that's what we've done. And it turns out that that message resonates really well. Because, as you probably know, most private clouds fail. Most private clouds fail miserably. Why? And there's really two reasons. There's two flavors of failure. But one is they just never work. Now that's because, guess what, it's incredibly hard. There are so many moving pieces and, guess what, we learned that ourselves. The numbers of times we stepped on the rakes, and, like, how do you make all this work? There's a gazillion moving parts. So if any of your, you have a team, that's failed at private cloud, they're not idiots. It's super, super, super hard. So that's one level of failure. But even those teams that got it working, they ultimately failed, as well, because of lack of usage. And the reason for that is, having done all that, they then built a snowflake cloud. And then when someone said, well, how do I use this? How do I add another NIC to a VM? The team that put it together were the only ones that could answer that. Nope, there was no ecosystem around it. So, with Azure Stack, the gestalt is, like, this is for people who want to use it, not for people who want to build it. So you just plug it in, you pick a vendor, and you pick a capacity. This vendor, four notes, this vendor 12 or 16 notes. And that's it. You come in, we ask you what IP range is, how do I integrate with your identity? Within a day, it's up and running, and your users are using it, really using it. Like, that's craziness. And then, well what does it mean to use it? Like, oh, hey, how do I ad a NIC to a VM? It's Azure, so how does Azure do it? I have an entire Azure ecosystem. There's documentation, there's training, there's videos, there's conferences. You can go and put on a resume, I'd like to hire someone with Azure skills, and get someone, and then they're productive that day. Or, and here's the best part, you can put on your resume, I have Azure skills, and you knock on 10 doors, and nine of them are going to say, come talk to me. So, that was the heart of it. And, again, it goes back to your question of, like, the value, or what does a technical fellow do. It's to figure out what really matters. And then say, we're all in on that. There was a lot of skepticism, a lot of customers like, I must have my security agent on there. It's like, well, no, then you're not a good candidate. What do you mean? I say, well, look, we're not going to do this. And they say, well you'll never be able to sell to anyone in my industry. I said, no, you're wrong. They say, what do you mean, I'm wrong? I say, well, let me prove it to ya, do you own a SAN? They say, well, of course we own a SAN. I said, I know you own a SAN. Let me ask you this, a SAN is a general purpose server with a general purpose operating system. So do you put your security and managing agents on there? And they said, no, we're not allowed to. I said, right, and that's the way Azure Stack is. It's a sealed appliance. We take care of that responsibility for you. And it's worked out very, very well. >> Alright, you got me thinking. One of the things we want to do is, we want to simplify the environment. That's been the problem we've had in IT, for a long time, is it's this heterogeneous mess. Every group did their own thing. I worry a multi-cloud world has gotten us into more silos. Because, I've got lots of SAS providers, I've got multiple cloud providers, and, boy, maybe when I get to the Edge, every customer is going to have multiple Edge applications, and they're going to be different, so, you know. How do you simplify this, over time, for customers? Or do we? >> Here's the hard story, back to getting at the heart of it. Look, one of the benefits of having done this a while, is I've stepped on a lot of these rakes. You're looking at one of the biggest, earliest adopters of the Boolean cross-platform, Gooey Framework. And, every time, there is this, oh, there's multiple platforms? People say, oh, that's a problem, I want a technology that allows me to bridge all of those things. And it sound so attractive, and generates a lot of early things, and then it turned out, I was rocking with this Boolean cross-breed platform. I wrote it, and it worked on Mac's and Windows. Except, I couldn't cut and paste. I couldn't print, I couldn't do anything. And so what happens is it's so attractive, blah, blah, blah. And then you find out, and when the platforms aren't very sophisticated, the gap between what these cross-platform things do, and the platform is not so much, so it's like, eh, it's better to do this. But, over time, the platform just grows and grows and grows. So the hard message is, people should pick. People should pick. Now, one of the benefits of Azure, as a great choice, is that, with the other guys, you are locked to vendor. Right, there is exactly one provider of those API's. With Azure, you can get an implementation of Azure from Microsoft, the Azure Public Cloud. Or you can get an implementation from one of our hardware vendors, running Azure Stack. They provide that to you. Or you can get it from a service provider. So, you don't have to get, you buy into these API's. You optimize around that, but then you can still use vendor. You know, hey, what's your price for this? What's your price for that, what can you give me? With the other guys, they're going to give you whatcha give ya, and that's your deal. (Rebecca laughs) >> That's a good note to end on. Thank you so much, Jeffrey, for coming on theCUBE again. It was great talking to you. >> Oh, that was fast. (Rebecca laughs) Enjoyed it, this was great. >> Great. I'm Rebecca Knight, for Stu Miniman, stay tuned to theCUBE. We will have more from Microsoft Ignite in just a little bit. (electronic music)

>> Hi, this is George Gilbert. We have an extra-special guest today on our CUBEcast, Aman Naimat, Senior Vice President and CTO of Demandbase started with a five-person startup, Spiderbook. Almost like a reverse IPO, Demandbase bought Spiderbook, but it sounds like Spiderbook took over Demandbase. So Aman, welcome. >> Thank you, excited to be here. Always good to see you. >> So, um, Demandbase is a Next Gen CRM program. Let's talk about, just to set some context. >> Yes. >> For those who aren't intimately familiar with traditional CRM, what problems do they solve? And how did they start, and how did they evolve? >> Right, that's a really good question. So, for the audience, CRM really started as a contact manager, right? And it was replicating what a salesperson did in their own private notebook, writing contact phone numbers in an electronic version of it, right? So you had products that were really built for salespeople on an individual basis. But it slowly evolved, particularly with Siebel, into more of a different twist. It evolved into more of a management tool or reporting tool because Tom Siebel was himself a sales manager, ran a sales team at Oracle. And so, it actually turned from an individual-focused product to an organization management reporting product. And I've been building this stuff since I was 19. And so, it's interesting that, you know, the products today, we're going, actually pivoting back into products that help salespeople or help individual marketers and add value and not just focus on management reporting. >> That's an interesting perspective. So it's more now empowering as opposed to, sort of, reporting. >> Right, and I think some of it is cultural influence. You know, over the last decade, we have seen consumer apps actually take a much more, sort of predominant position rather than in the traditional, earlier in the 80s and 90s, the advanced applications were corporate applications, your large computers and companies. But over the last year, as consumer technology has taken off, and actually, I would argue has advanced more than even enterprise technology, so in essence, that's influencing the business. >> So, even ERP was a system of record, which is the state of the enterprise. And this is much more an organizational productivity tool. >> Right. >> So, tell us now, the mental leap, the conceptual leap that Demandbase made in terms of trying to solve a different problem. >> Right, so, you know, Demandbase started on the premise or around marketing automation and marketing application which was around identifying who you are. As we move towards more digital transaction and Web was becoming the predominant way of doing business, as people say that's 70 to 80 percent of all businesses start using online digital research, there was no way to know it, right? The majority of the Internet is this dark, unknown place. You don't know who's on your website, right? >> You're referring to the anonymity. >> Exactly. >> And not knowing who is interacting with you until very late. >> Exactly, and you can't do anything intelligent if you don't know somebody, right? So if you didn't know me, you couldn't really ask. What will you do? You'll ask me stupid questions around the weather. And really, as humans, I can only communicate if you know somebody. So the sort of innovation behind Demandbase was, and it still continues to be to actually bring around and identify who you're talking to, be it online on your website and now even off your website. And that allows you to have a much more sort of personalized conversation. Because ultimately in marketing and perhaps even in sales, it comes down to having a personal conversation. So that's really what, which if you could have a billion people who could talk to every person coming to your website in a personalized manner, that would be fantastic. But that's just not possible. >> So, how do you identify a person before they even get to a vendor's website so that you can start on a personalized level? >> Right, so Demandbase has been building this for a long time, but really, it's a hard problem. And it's harder now than ever before because of security and privacy, lots of hackers out there. People are actually trying to hide, or at least prevent this from leaking out. So, eight, nine years ago, we could buy registries or reverse DNS. But now with ISBs, and we are behind probably Comcast or Level 3. So how do you even know who this IP address is even registered to? So about eight years ago, we started mapping IP addresses, 'cause that's how you browse the Internet, to companies that they work at, right? But it turned out that was no longer effective. So we have built over the last eight years proprietary methods that know how companies relate to the IP addresses that they have. But we have gone to doing partnerships. So when you log into certain websites, we partner with them to identify you if you self-identify at Forbes.com, for example. So when you log in, we do a deal. And we have hundreds of partners and data providers. But now, the state of the art where we are is we are now looking at behavioral signals to identify who you are. >> In other words, not just touch points with partners where they collect an identity. >> Right. >> You have a signature of behavior. >> That's right. >> It's really interesting that humans are very unique. And based on what they're reading online and what they're reading about, you can actually identify a person and certainly identify enough things about them to know that this is an executive at Tesla who's interested in IOT manufacturing. >> Ah, so you don't need to resolve down to the name level. >> No. >> You need to know sort of the profile. >> Persona, exactly. >> The persona. >> The persona, and that's enough for marketing. So if I knew that this is a C-level supply chain executive from Tesla who lives in Palo Alto and has interests in these areas or problems, that's enough for Siemens to then have an intelligent conversation to this person, even if they're anonymous on their website or if they call on the phone or anything else. >> So, okay, tell us the next step. Once you have a persona, is it Demandbase that helps them put together a personalized? >> Profile. >> Profile, and lead it through the conversation? >> Yeah, so earlier, well, not earlier, but very recently, rebuilding this technology was just a very hard problem. To identify now hundreds of millions of people, I think around 700 are businesspeople globally which is majority of the business world. But we realize that in AI, making recommendations or giving you data in advanced analytics is just not good enough because you need a way to actually take action and have a personalized conversation because there are 100 thousand people on your website. Making recommendations, it's just overwhelming for humans to get that much data. So the better sort of idea now that we're working on is just take the action. So if somebody from Tesla visits your website, and they are an executive who will buy your product, take them to the right application. If they go back and leave your website, then display them the right message in a personalized ad. So it's all about taking actions. And then obviously, whenever possible, guiding humans towards a personalized conversation that will maximize your relationship. >> So, it sounds like sometimes it's anticipating and recommending a next best action. >> Yeah. >> And sometimes, it's your program taking the next best action. >> That's right, because it's just not possible to scale people to take actions. I mean, we have 30, 40 sales reps in Demandbase. We can't handle the volume. And it's difficult to create that personalized letter, right? So we make recommendations, but we've found that it's just too overwhelming. >> Ah, so in other words, when you're talking about recommendations, you're talking about recommendations for Demandbase for? >> Or our clients, employees, or salespeople, right? >> Okay. >> But whenever possible, we are looking to now build systems that in essence are in autopilot mode, and they take the action. They drive themselves. >> Give us some examples of the actions. >> That's right, so some actions could be if you know that a qualified person came to your website, notify the salesperson and open a chat window saying, "This is an executive. "This is similar to a person who will buy "a product from you. "They're looking for this thing. "Do you want to connect with a salesperson?" And obviously, only the people that will buy from you. Or, the action could be, send them an email automatically based on something they will be interested in, and in essence, have a conversation. Right? So it's all about conversation. An ad or an email or a person are just ways of having a conversation, different channels. >> So, it sounds like there was an intermediate marketing automation generation. >> Right. >> After traditional CRM which was reporting. >> Right, that's true. >> Where it was basically, it didn't work until you registered on the website. >> That's right. >> And then, they could email you. They could call you. The inside sales reps. >> That's right. >> You know, if you took a demo, >> That's right. >> you had to put an idea in there. >> And that's still, you know, so when Demandbase came around, that was the predominant between the CRM we were talking about. >> George: Right. >> There was a gap. There was a generation which started to be marketing. It was all about form fills. >> George: Yeah. >> And it was all about nurturing, but I think that's just spam. And today, their effectiveness is close to nothing. >> Because it's basically email or outbound calls. >> Yeah, it's email spam. Do you know we all have email boxes filled with this stuff? And why doesn't it work? Because, not only because it's becoming ineffective and that's one reason. Because they don't know me, right? And it boils down to if the email was really good and it related to what you're looking for or who you are, then it will be effective. But spam, or generic email is just not effective. So it's to some extent, we lost the intimacy. And with the new generation of what we call account-based marketing, we are trying to build intimacy at scale. >> Okay, so tell us more. Tell us first the philosophy behind account-based marketing and then the mechanics of how you do it. >> Sure, really, account-based marketing is nothing new. So if you walk into a corporation, they have these really sophisticated salespeople who understand their clients, and they focus on one-on-one, and it's very effective. So if you had Google as a client or Tesla as a client, and you are Siemens, you have two people working and keeping that relationship working 'cause you make millions of dollars. But that's not a scalable model. It's certainly not scalable for startups here to work with or to scale your organization, be more effective. So really, the idea behind account-based marketing is to scale that same efficacy, that same personalized conversation but at higher volume, right? And maximize, and the only way to really do that is using artificial intelligence. Because in essence, we are trying to replicate human behavior, human knowledge at scale. Right? And to be able to harvest and know what somebody who knows about pharma would know. >> So give me an example of, let's stay in pharma for a sec. >> Sure. >> And what are the decision points where based on what a customer does or responds to, you determine the next step or Demandbase determines what next step to take? >> Right. >> What are some of those options? Like a decision tree maybe? >> You can think of it, it's quite faddish in our industry now. It's reinforcement learning which is what Google used in the Go system. >> George: Yeah, AlphaGo. >> AlphaGo, right, and we were inspired by that. And in essence, what we are trying to do is predict not only what will keep you going but where you will win. So we give rewards at each point. And the ultimate goal is to convert you to a customer. So it looks at all your possible futures, and then it figures out in what possible futures you will be a customer. And then it works backwards to figure out where it should take you next. >> Wow, okay, so this is very different from >> They play six months ahead. So it's a planning system. >> Okay. >> Cause your sales cycles are six months ahead. >> So help us understand the difference between the traditional statistical machine learning that is a little more mainstream now. >> Sure. >> Then the deep learning, the neural nets, and then reinforcement learning. >> Right. >> Where are the sweet spots? What are the sweet spots for the problems they solve? >> Yeah, I mean, you know, there's a lot of fad and things out there. In my opinion, you can achieve a lot and solve real-world problems with simpler machine learning algorithms. In fact, for the data science team that I run, I always say, "Start with like the most simplest algorithm." Because if the data is there and you have the intuition, you can get to a 60% F-score or quality with the most naive implementation. >> George: 60% meaning? >> Like accuracy of the model. >> Confidence. >> Confidence. Sure, how good the model is, how precise it is. >> Okay. >> And sure, then you can make it better by using more advanced algorithms. The reinforcement learning, the interesting thing is that its ability to plan ahead. Most machine learning can only make a decision. They are classifiers of sorts, right? They say, is this good or bad? Or, is this blue? Or, is this a cat or not? They're mostly Boolean in nature or you can simulate that in multi-class classifiers. But reinforcement learning allows you to sort of plan ahead. And in CRM or as humans, we're always planning ahead. You know, a really good salesperson knows that for this stage opportunity or this person in pharma, I need to invite them to the dinner 'cause their friends are coming and they know that last year when they did that, then in the future, that person converted. Right, if they go to the next stage and they, so it plans ahead the possible futures and figures out what to do next. >> So, for those who are familiar with the term AB testing. >> Sure. >> And who are familiar with the notion that most machine learning models have to be trained on data where the answer exists, and they test it out, train it on one set of data >> Sure. >> Where they know the answers, then they hold some back and test it and see if it works. So, how does reinforcement learning change that? >> I mean, it's still testing on supervised models to know. It can be used to derive. You still need data to understand what the reward function would be. Right? And you still need to have historical data to understand what you should give it. And sure, have humans influence it as well, right? At some point, we always need data. Right? If you don't have the data, you're nowhere. And if you don't have, but it also turns out that most of the times, there is a way to either derive the data from some unsupervised method or have a proxy for the data that you really need. >> So pick a key feature in Demandbase and then where you can derive the data you need to make a decision, just as an example. >> Yeah, that's a really good question. We derive datas all the time, right? So, let me use something quite, quite interesting that I wish more companies and people used is the Internet data, right? The Internet today is the largest source of human knowledge, and it actually know more than you could imagine. And even simple queries, so we use the Bing API a lot. And to know, so one of the simple problems we ran into many years ago, and that's when we realized how we should be using Internet data which in academia has been used but not as used as it should be. So you know, you can buy APIs from Bing. And I wish Google would give their API, but they don't. So, that's our next best choice. We wanted to understand who people are. So there's their common names, right? So, George Gilbert is a common name or Alan Fletcher who's my co-founder. And, you know, is that a common name? And if you search that, just that name, you get that name in various contexts. Or co-occurring with other words, you can see that there are many Alan Fletchers, right? Or if you get, versus if you type in my name, Aman Naimat, you will always find the same kind of context. So you will know it's one person or it's a unique name. >> So, it sounds to me that reinforcement learning is online learning where you're using context. It's not perfectly labeled data. >> Right. I think there is no perfectly labeled data. So there's a misunderstanding of data scientists coming out of perfectly labeled data courses from Stanford, or whatever machine learning program. And we realized very quickly that the world doesn't have any perfect labeled data. We think we are going to crowdsource that data. And it turns out, we've tried it multiple times, and after a year, we realized that it's just a waste of time. You can't get, you know, 20 cents or 25 cents per item worker somewhere in wherever to hat and label data of any quality to you. So, it's much more effective to, and we were a startup, so we didn't have money like Google to pay. And even if you had the money, it generally never works out. We find it more effective to bootstrap or reuse unsupervised models to actually create data. >> Help us. Elaborate on that, the unsupervised and the bootstrapping where maybe it's sort of like a lawnmower where you give it that first. >> That's right. >> You know, tug. >> I mean, we've used it extensively. So let me give you an example. Let's say you wanted to create a list of cities, right? Or a list of the classic example actually was a paper written by Sergey Brin. I think he was trying to figure out the names of all authors in the world, and this is 1988. And basically if you search on Google, the term "has written the book," just the term "has written the book," these are called patterns, or hearse patterns, I think. Then you can imagine that it's also always preceded by a name of a person who's an author. So, "George Gilbert has written the book," and then the name of the book, right? Or "William Shakespeare has written the book X." And you seed it with William Shakespeare, and you get some books. Or you put Shakespeare and you get some authors, right? And then, you use it to learn other patterns that also co-occurred between William Shakespeare and the book. >> George: Ah. >> And then you learn more patterns and you use it to extract more authors. >> And in the case of Demandbase, that's how you go from learning, starting bootstrapping within, say, pharma terminology. >> Yes. >> And learning the rest of pharma terminology. >> And then, using generic terminology to enter an industry, and then learning terminology that we ourselves don't understand yet it means. For example, I always used this example where if we read a sentence like "Takeda has in-licensed "a molecule from Roche," it may mean nothing to us, but it means that they're partnered and bought a product, in pharma lingo. So we use it to learn new language. And it's a common technique. We use it extensively, both. So it goes down to, while we do use highly sophisticated algorithms for some problems, I think most problems can be solved with simple models and thinking through how to apply domain expertise and data intuition and having the data to do it. >> Okay, let's pause on that point and come back to it. >> Sure. >> Because that sounds like a rich vein to explore. So this is George Gilbert on the ground at Demandbase. We'll be right back in a few minutes.