 Live from New York, it's theCUBE. Covering Big Data New York City 2016. Brought to you by headline sponsors, Cisco, IBM, NVIDIA, and our ecosystem sponsors. Now, here are your hosts, Dave Vellante and Jeff Frick. We're back at NYC. Jim McHugh was here as the vice president and general manager at NVIDIA. Jim, good to see you again. Good to see you guys. It was good. Monday night, we had an awesome event with you guys. We were thrilled to partner up with you. You must be happy. It was really successful. Yeah, a lot of great customers came out of that. A lot of conversations going on. That was a great event. So you gave a keynote on Tuesday. You gave another talk in like an hour. What are you even talking about? I mean, it's the hot topic here, is AI and deep learning. Yeah, so yeah, that's exactly my talk. Was really pretty fantastic on the other day. If you do say so yourself. You know, I was in the company of geniuses, really. So when you follow Yann Lacoon, who is like kind of the godfather of machine learning, AI, and he was on stage and you get to go up and follow him, it's pretty exciting. He was one of the first customers of my DJX One. So he was the first customer on the East Coast that we delivered to a couple of weeks ago. They've already using it. But the talk there was about what's deep learning doing? How are we using it? And how is everybody taking advantage of it? And focused a lot on our activities and autonomous driving, because there's been a lot of talk about self-driving cars in the market. And it's really interesting how people don't even really understand what it's about. Like they think smart cameras, radar, and there's been some debates and the press about that. But it's much more than that. Because those are just detectors. And really is, you got to detect, but then you actually have to understand. You have to be able to see, hear, understand what's going on, make intuitive decisions. And then there's an action plan from there. And so that's what we're working on, is that whole component. In fact, we have a car that's built. It's called BB-8. And BB-8 didn't just, wasn't programmed. BB-8 is learning. So what we did, we put eight cameras around the car and just drove BB-8. And so when you drove it for the first little bit and you go about in a hundred hours, yeah, it took over the cones. You got to test it in the parking lot. After 3,000 hours, it's driving on the streets of Jersey. And it's going down country roads. Most people think in self-driving cars, it knows the lanes, right? And knows how to stop. There's no lanes on the country roads in Jersey, right? So he's just driving around in the rain, doesn't care, right? Because it's super human. AI cars, they never stop thinking, they never stop working, and they never get tired either. So it's pretty cool. They can see both ways at the same time. We've got the Google cars come down our street and make a very challenging left hand turn. They come in batches, they'll come 10 in 15 minutes. And I'm sure they're just teaching them because there's all kind of trees in this, that. But they can literally look both ways at the same time. Oh, they're looking all around it. They know exactly where it's going on in their component. Well, I remember several years ago when I was out visiting, I saw my first Google car, you know, a while ago, right? And now every time I go, you see multiple. But it's interesting to me how that whole business has exploded. I mean, so what happened is just that people saw the possibilities, or a bunch of Google engineers left, and it was just sort of open source. And the guys at Google are smart, but they're not the only ones leading, right? It's honestly, it's deep learning. It's the ability to teach the car, right? And the car is learning intuitively, because you're just throwing all this data at it. It's getting all this camera components going from there. But it keeps, it's even actually more interesting than that, right? Nvidia, right? Where the graphics guys, you know, I like to say we became known for handling the most demanding users in the industry, gamers, right? Yeah, right. Well, guess what games are? You're simulating, you're creating a reality. We can now do augmented reality or using that to train cars. Because you can't find every possible situation on the streets of New York or on the streets of Palo Alto or wherever. You actually have to create it, right? You know these things are going to come up, and you don't have to put the car through that to teach it. So we're using augmented reality to do a lot of this as well. And I think so many people don't understand too kind of the second order impact. Like you said, while the car is driving around, whether it generally doesn't crash, but it's picking up all types of data on the way that roads actually work, the way bicyclists actually behave, the way pedestrians actually behave. So every run, it's really getting so much data in the way the real world of the streets work that most people don't really get that. Yeah, and what's crazy though, the first AI robot is a car. But AI robots are coming like crazy. Berkeley has one that's called Brett. It's one of my favorites because it's the Berkeley robot for the elimination of tedious tasks. And that's how it came up with the name. So it's a great name. But Honda has Isimo. And really what these things are doing is they're learning. Like Brett was taught to stand. And the goal was get your head as high as possible. No programming, no other than that. And it just kept trying to get his head higher and higher. And then it stood up and it learned how to stand up. Then learned to screw a cap on. Isimo's kicking soccer balls, going all this stuff. But what they're going to eventually be is be able to learn the task that we don't want to do, the tasks that are too tedious for us to do, but also tasks that are harmful for us to do. And in Japan it's really big because of the aging population. They need caregivers, right? They need people to take care of people. People that can know when pills need to be taken, know how to do all that kind of stuff. And it's really astounding how fast these things are coming. So I love that because that's kind of incremental. But it's also, people are concerned, right? They say, well, machines are going to replace humans. But machines have always replaced humans. But now they're starting to do so in cognitive tasks. But the list of things that machines can do, or if humans can do that machines can't, is dwindling, right? Every year you look at it and say, okay, because robots used to not be able to climb stairs and other climbing stairs. Climbing stairs and just climbing stairs. And there's so much research going on. And it's funny because people say, well, yeah, it's in research. No, but the way this space is working, AI and deep learning is happening so fast, it starts in research. It's the Yanlacuns who work at Facebook as well, right? You know, you guys met Claudio the other night for Claudio Silva who's at NYU, but he's the guy behind baseball doing a lot of the smart stuff they're doing there. So it's really, really cool stuff going on. I saw some tweet about, will robots take the place of lawyers? And I'm like, well, who can go through how much case law faster? I mean, are you kidding? Absolutely, yeah, they can learn, they can just study. That's the thing about this, they don't forget. They just continually learn. And then that knowledge that they learn, it stays with them. Somebody said to me, well, machines won't be able to negotiate. I'm like, why not? Yeah, how do you program it? I mean, you know, between marketplaces and just learning how to go back and forth. Well, the big component really why this became popular, and it's kind of funny, but it's only been since last December, the game ago lost to the champion lost to a computer. And there was a move, I think it was like move 37, the computer chose the center square. And the commentators, if you were watching said, oh yeah, it's a glitch. No one would take that move or whatever. It shut down the game really quick. It won. It was just like the move. The judo move, let's go. That no one would have thought. And it was the turning point in the game. And the computer just won, because no human would have been able to see all those patterns and going from that. To get there. And to get there. The other fun thing that you mentioned on Monday night, and I learned long ago in describing services, the key is not to have a better numerator with the variables to change the variable. And you asked a bunch of the panelists really to compare before and after using the GPU technology. And it's not, it's fewer days. It's from days to hours to minutes. You know, you're changing the variable by which you measure what you can accomplish. Pretty crazy numbers that came out of that. Well, that's what was kind of crazy the transition I made this week where I'm standing with the geniuses of AI and deep learning. And then I go to the big data and analytic conference, you know, just across the hall. And now we're talking about acceleration of database, acceleration of analytics and going from there. And, you know, I'm kind of telling people our goal is so you don't have to sit through another 4V presentation, right? If we have to talk about volume, velocity, variety, I think that the first time has been mentioned in three days. Let's just put it in there. And because if you have to wait for a query for 10 seconds, but on your average dashboard or as you're looking at the data, there's 10 to 15 queries. 10 seconds times each of those just adds up to too much time. You're not exploring the data. You're not going through it. So when a GPU solution like Kinetico or MapD or Scream that was here the other night, you're already on to your third or fourth exploration of the data while the first guy is still waiting for it to return. Yeah, well. Just because milliseconds versus over a minute is... Especially when you're buying it to big, big, big data sets, it really adds up. Well, the funny thing is a lot of people say, well, how many use cases are there really for that? And there's a lot more than people realize because it's a game changer. So it opens up whole new sets of applications. Well, isn't it funny that one of the most expensive, one of the most treasured people who work for you, a data scientist, we frustrate them. I mean, where's the goal in that? You know, you're trying to keep the guy there, give him the tools that allow him to explore. Data scientists are really creative people. They like to explore data. They like to go through the data. But if you make them wait, they're going to get frustrated. They're going to go, okay, how do I get to the answer? And they're going to get the answer and they're going to go from there. Because if you're not going to give them the tools, they can't get to, they can't discover those new things. The outliers, the long tails, they're going to be frustrated and stuck. Yeah, there's artists, they're artists in a way. You mentioned a couple of your partners, Connecticut, MAPTI, others. The ecosystem is really interesting around NVIDIA. Talk about that a little bit. So it's come quite fast actually, like everything that's going on. You have people start seeing what you can do. And you know, I'd say fast, but these are guys that have been working on their product for a couple of years, right? So they've been at it and doing it. But people are starting to see what accelerated databases can do. And you're not learning a new skill set, right? SQL and going from there. And then you're seeing what you can do with accelerated interaction with the data, accelerated visualization, right? So the dashboards that we're all looking at right now are nothing compared to what people are going to be using a year or two from now, where you're actually going to be intuitively going through the data. And then it keeps banding out, right? So you saw the other night, how many accelerated database companies there are now, right? So, and different options coming. Accelerated visualizations can be one of the coolest things. You know, there's a company called Graphistry. Graphistry will be a part of my presentation for about an hour. But Leo's here as well, the CEO, and he's going to be presenting. And they are able to point Graphistry at a Splunk log. And if you look through a Splunk log, it's not the most exciting thing that you're going to be looking through. But if you can actually visualize it. So imagine it's like the top security things that happen that day. And I can visualize these IP addresses or where the big clusters have been. And I can see the entry points in the exit and how everything comes together. And I'm looking at hundreds of millions, right? Going across the networking activity. I can then narrow in further and get down this, you know, two million. I can get down to 100,000. I can keep exploring the data to where I want. And it's doing all the correlations that I'm seeing, which has been quite impressive. Well, what people need to understand, if you try doing this with a conventional visualization tool, like a tableau or a click, it would just bring it to its knees. It would just stop. It would choke. Yeah, you can't scale out enough. There's not enough real estate to keep adding, you know, more power, more power, more power. So that's why, I mean, my DGX, it's actually quite popular. It's kind of nice to see people come over and petting a server, right? It's like, it's a super company, right? It has eight GPUs that are on an NVLink board. There's no bottlenecks. They all communicate with each other. So it's the equivalent of 250 X86 servers. And people are like, well, how much is that? I'm like, well, if you want to say 10, a rack, that's 25. That's basically a row in a lot of data centers. We got a lot of questions after Monday night in terms of the market for GPU versus CPU. What's it going to look like in five years? You must get that question a lot. Well, I tell people, it's a new computing model. It's really a change in the world where you're allowing billions of software neurons to talk to trillions of connections and they're all being trained and learned at the same time. It's parallel processing. Everything we do is parallel these days. So you'll need CPUs. There's certain things that are serial, but the market's changing. It is the new platform and a new way of doing software. All right, good, we got to go. Jim, thanks very much for coming on theCUBE. Good luck on the talk. All right, thanks so much. It was fantastic partnering with you guys on Monday night. Really pleasure. Okay. All right, keep right there, everybody. We'll be back with our next guest of theCUBE we're live from Big Data NYC. Right back.