 Hi everybody, we're back. This is Dave Vellante of Wikibon.org. I'm here with Jeff Kelly. We're here live at the Vertica user event. We're here in Boston, Massachusetts, the Western Hotel and the Waterfront. It's been two days of talking to practitioners and partners and HP executives. Brian Weiss is here. He's the Vice President of Marketing of Autonomy. He's a CUBE alum. Brian, welcome back. Good to see you again. Thanks a lot, Dave. It's nice to be here. Yeah, so we're seeing, last time we talked, you know, we talked about the deliberate attempts to integrate autonomy into many, many parts of the HP portfolio. We're now seeing the results of that. We heard today in the keynotes, Robert Young-Johns, George Kadifa, talking a lot about Haven and giving some demos. So we're really starting to see this strategy come together. And you know, I said to my colleague, John Furrier, when you're on the outside looking in, you know, and there's all this noise going on in the marketplace. They say, what the heck's going on inside there? But inside, you guys were doing some really deliberate planning, doing road maps. I mean, it's very clear now that we're seeing the results of that planning. So talk about where we're at and how do you feel? Yeah, you know, this is really exciting for us. You're right, we've been through that sort of year of deliberate strategy and planning about how we're going to take these very, very unique technology assets that we've got within Vertica for managing just excruciating amounts of extreme data, all of machine data. And then you've got the autonomy portfolio, which is fundamentally something very different. It understands the nuance and the nearness and the sort of things that human brains do that maybe databases don't. And so it's always been incredibly apparent to us that there's an unbelievable amount of synergy there, the drive of value that probably most business applications can't conceive of. And certainly in the face of the types of data we're trying to get in the value again. So, you know, with Haven as a platform strategy, really it represents our commitment to get that unique value both integrated and out to the market, whether it's as a development platform or underpinning applications that'll tell you not only what happened and how fast it happened and, you know, but as well as what it means. So those two things together. How many widgets, et cetera. And, you know, it was interesting to see George Kedifus slide where he gave a little history lesson on sort of the transaction piece of the business. And now, you know, there's a 10X growth there. And then all of a sudden he's got this 100X vision. And the thing about that, I wonder if you could comment, is a lot of that value, he talked about monetization, a lot of that value creation is, it's kind of fuzzy, you're drawing inferences, it doesn't have to be like perfectly precise, you know, a false positive, it doesn't kill you, you know, it's not like a zero in a bank account. And that's where the values come up, and that's where autonomy shines. Yeah, and you're right. So, and I think, you know, for those of you who didn't see the slide that you were talking about, it's really the movement of being able to provide insight and analysis on data that fits in a Rona column. And then the next sort of wave of that is that your computer can actually do things that are very similar to what you and I do with our brains, which is, yeah, it's not the same, but it's close, right? So I want to know all of the, you know, I want to not only have phone calls that came in between 10 and 11, if there's 20 million of them, but I want to know what they were about and more of the people mad. And if they were, I want to know the stuff they were mad about, but, and show me things, don't just give me the answer that seems right, show me the next most likely answer. So it's a question of the, you know, the data talking to you in a way and putting those two things together means that compute systems will do something very different. Very, very different. Well, you know, you're talking about sort of more like the human brain, but of course, you can't take the humans out of the equation. Now, what you've done is you're compressing the time it takes to actually get to, you're actually making it feasible because it used to take, you know, days, sometimes months, and at that point, you're just like, the market's changed, you know, it's too hard, IT's saying no. So how is that affecting the way in which customers are changing their processes specifically to take advantage of that? I think that's really interesting. It's part of, I think it's part of why everybody's so hyped up about big data in general. So when we used to deal with the volumes of the data, it was all about, well, it costs a lot to store it and it's a pain and I don't know what's in it. And now what customers are saying is, I've got all this data, how do I get the, I think there's a lot of value in this information. If it only could figure out a way to get it out. And one way to get it out is to hire room pulls of people and have them read the documents and tell you what they're about but you can't do that. And so the realization is you can't do that human task on the amount of data we're generating or the types we're generating. And if you could, you'd be in good shape. Like a lot of great value you can find out. People are beginning to realize that it's not as possible but it's happening. And so we see that all over the place. People come to us with a big data problem and they say, I want to know how my customers feel about X, Y and Z and I'm going to get that data from YouTube, from Twitter, from social media, from surveys. And I'm basically going to couple that with information that I'm getting off of telemetry coming from their cars. I was going to say, Jeff's going to get into use cases with you but we saw that at Discover in June in Las Vegas with NASCAR. Yes. And I had some great extensive conversations with those guys in terms of how they were making decisions and virtually in close to real time as possible as to what videos to keep up there. They were taking down videos and getting instant feedback. No, get that video back. Exactly. And they're almost like a command center. And they could see the results to the bottom line. Right, right. And it goes well beyond just saying, I want to know how many times people are saying NASCAR on Twitter. It's actually what are they saying where they care about and what are the main ideas and how they matter. I mean, there's a whole nuance to it that is very hard to get to a traditional computer. You're right. It's like, hey, people are talking about us. Golf claps. What do you do with that? Right. You know, you got a lot of mentions. Okay, good. But what did they say? Like tell me if I've got a problem. I mean, a great example I'll give you of, you know, I did a presentation today on how we're being used in some movie studios to analyze information out there in the world before a movie launch. So what are people saying? How are they saying it? Main ideas? How are they responding to our actors? How are they responding to our ads? Right, are we doing the right thing or not? Right, that kind of thing. And you can imagine the sources that you- Is this going to flop or is it going to hit a home run? Yeah, or, and by the way, I'm leading up to my release so I still can make changes. Yeah, right. Like I can still adjust, you know, which actor I send to what location depending on how they're trending and how do you get that data, right? So there was this case- You can move the needle with that data. You can absolutely move the needle on it. But there was a good use case where we were doing that and it just happened to be after a film had released and we started seeing these Twitter patterns of, it was a kid's movie and the kids were like, okay, and then we saw things like crying, children screaming from the theater, running and it was geo-targeted and we knew where it was coming from and somebody had accidentally loaded up the Ghost Rider trailer for this kid's movie. And so you go in to see a kid's movie and there's melting faces and skulls but in a particular, some maul and so that was, of course, they were able to see that and spot it and get ahead of it because along with what they were tracking, they saw these other patterns coming out of it. So Brian, I'm running to a panel with my colleagues Vinny, Mersendani and Kurt Monash. We're going to go talk about big data. So Jeff's going to take it from here, talk about use cases and other examples. Really great to see you again, always a pleasure. So yeah, so before we move into some of those use cases, I wanted to follow up on your answer to the last question about, so you've got these insights, right? The movie studio example, I think is a good one. But we often hear about, and this relates to the, you can't take the human out of the process, how do you help customers not just understand the data and the insights, but then actually take the next best action to improve the situation if it's, in the case of movie studio example. Yeah, there's two parts of not taking the human out. The machine's never going to make the decision for you. What it's going to do is surface, it's going to be machine assisted, if you will. It's going to surface intelligence that you might not have been able to get from data that which is too big for you to read by yourself. But you're exactly right. And one of the things that we're focused on with Vertica and with the Haven platform is not just, let me show you a picture of what's happening, but let me give you the tools to take action on it. And I think that's one of the places that, people get excited about big data and it's, but they're going to be told something and then what? Now what? Where's my ROI? Right, super. So if I, for a lot of the things that we're focused on doing is saying, let's take this insight, let's take the information and let me take the people associated with it, whether it's the influencers who are tweeting the most about it or whether it's the customers who feel a certain way about a topic or a group of folks in a certain geography and then I want to take that, cluster those individuals, put them into a workflow and start a campaign around them. Say for example, when I'm marketing, right? I can take directs action on that, target a group of people who happen to be talking about this subject, but I had no idea it would be hitting my screen when I woke up this morning and it's about my movie, right? Or I can market this particular way to this. So we're looking, it's important for us to see and drive workflow out of it that delivers results. So not just dashboards. Right, so it's a lot more than just saying, here's some sentiment people like this or don't like this, it's directing next best actions to actually. Well, so let's take it a step further, either, right? Another core part of, and you heard about it a little bit this morning in the security part of the presentation, that we're doing is, we're taking the ability of something like Artsite Logger, right? And experimenting with the fact that we can now take the events that have been logged, which tell us what happened and when and some information about it. And we can also then look at the information itself. So maybe I'm interested in people who send data outside of my environment within 30 days of leaving the company. Maybe it's something I'm very interested in if I'm a pharmaceutical company or anything else. And so, well, while traditional systems can tell you that's happening, they maybe tell you what's happening a lot and they can't tell you what the document's about. So maybe you send an intellectual property out, maybe you did it on whether you were mad or maybe you just sent your soccer schedule. So here's another point where Idle can come in or Autonomy can come in and go, actually, no, we'll have the computer read the document and compare it to things that we know to be very likely about intellectual property that maybe you've given us. And we can come back and say, okay, red alert, documents are leaving, that's fine, people send them out all the time, but this particular set of documents is actually very risky because we understand what that is and there's no way you could do that without a person reading the doc. So that's kind of the type of place where or we can monitor the social media sites where people who might be talking about your company and perhaps intending or plotting something. Before a document goes outside the firewall, we can monitor that, not only the activity, but also what's being said and what it might mean for you. So it's a great example of where we're taking this security part of Haven, right? And looking at ways to create that synergy that's not possible with traditional tech. So as a marketer, what are some of the challenges that you have? Because we're talking about actually using computers to kind of mimic some of the things that human brain can do. And I'm sure there's some skepticism out there. I'm not too far with that, right? I don't want to, let's be very clear here. We're not talking artificial intelligence, but it will do things that are similar to the inferences that you and I would do. So not necessarily AI, but it's still pretty sophisticated stuff. And I imagine you come up against some skepticism when you go into potential clients. So as a marketer, how do you tell that story? How do you get that across to a skeptical end user who wants to believe what you're saying, but is saying, is this science fiction kind of stuff? Yeah, so the best way to do that is to talk about the way customers use it. And we can give you clear examples. And we have lots of demonstrations that will show you in real time what this like news. Pick out all the key ideas and news. Actually, we released a website that's completely public. Anybody on the planet can get to it. It's called news slash spectrum.com. And it uses autonomy to segregate the news automatically. It reads all the stories and decides whether about sports or politics, entertainment, and it drives the threads and the ideas within them in the same way that it would if I'd hire a roomful of people to read all the news and categorize it for me. So we do things like that, but really what's very interesting is to take the use cases where customers have taken what is a very expensive human problem like. You're a law firm and you need to read 10 million documents about a case on behalf of your client, which you're getting paid $350 an hour for, and find the relevant documents. Now what's relevancy? The judge does not say I want you to find the documents with the word bridge in them. The judge says, find me the relevant documents. Here's the case. So you have to do this exercise. And so it's a great example of where what we can do is say, all right, well, first of all, have the computer organize all the data. Put it into clusters. If this thing's about bridge, but it's bridges that go over water and not bridges as dental work and bridges a card game or bridges a musical form, there's all different contexts of that. And so we'll divide it up and say, no, it means different things. And if I then feed the machine an example of five things that are relevant, by the way, I don't have to tell it why. I don't have to say it's relevant because it's got this word and this word and this word. Nope, things like this. This counts. And it can do a nuanced analysis of the patterns in that and say, okay, great, here's all the ones. I'm going to train it again. You classify four more. So I can turn a review cycle. You know, for an attorney who's very expensive and I can use the machine to effectively do the lawyer's work. Of course, you're never going to say, okay, just let the machine run and submit. What you do is they have this iterative quality check. Is that relevant? Yes, no, okay, computer, that one doesn't count. This one does. Go again, but you can increase it by 200 fold. The speed of review and the quality of by 200 fold. So that's a significant improvement in the productivity of essentially that example of a team of lawyers going through these documents. Well, now they don't have to spend as much time doing that. They can spend times and more on higher value activities. Higher value activities, that's right. And it's going to save you money because as you said, the lawyers are very expensive. And you know, people are skeptical about this. In that business, in that industry, there's a lot of like, is the computer going to get it right and how do I trust it and sort of these things have gone all the way up into the courts who have evaluated this type of machine-assisted. They've actually come down and said, no, this is much better than people. And lots of, like, you get to be timid with people, you're going to have a margin of error. And you're going to, you know, but actually the computer does a better job in many cases of finding what you need, at least getting that first cut of relevance. Well, that's a key kind of line across when the computer is better than the person. That is a lot easier to sell that. I think a lot of what, in the answer to a question like, how do I know it's going to be, look, you can't, it's never going to be perfect, but neither are people. And so the probability modeling that we're doing to come up with this is probably what you're looking for because we think it's related to things that are similar to things that you've told us, or it's all, we've got to be good enough. We've got to end it well. Well, that's the thing, people aren't perfect either. And that's some people, I think when people are evaluating this kind of technology, they hear about it, they take, they have that assumption, well people always get it right. And that's literally not the case. So an improvement over what a human can do is where you kind of cross that barrier where this is really valuable. But plus, even if you could get it right, like you look at automatic document categorization, right? I mean, what it would take you to do that would be, we're going to read every document. That's crazy, you can't do that. I mean, that's what we do keyword search for. Okay, I can't read all of these. Find me the one with the word bridge in it and give them all of those. Okay, there's a bunch in there about card games. That's no good. But you see that perfect being the enemy of the good a little bit in here and realize that we're just like, you don't have any hope otherwise. Because you're not going to get through these 15 million documents unless you spend $20 million. And that's what the judge says. No, the computer's a lot better. Right. So we're here, there's a lot of Vertica customers here. Talk a little bit about how you're kind of talking to these customers about some of the things they're doing. Some, I imagine are autonomy customers as well, but not all. How do you kind of communicate that, the value proposition to some of these customers here that are using Vertica? You know, I think what's, people often ask me, okay, what's the difference between idle and Vertica and how do we work together and what do we do? Look, Vertica is an extraordinary engine for blazing fast query on excruciating amount of data. To be able to load and query that amount of structured information and the amount of time it can is, and that's not what autonomy does. You could do that, but you wouldn't be even anywhere near. It's not what it's built to do. And what autonomy is built to do is give you a different answer. So Vertica's going to give you the answer based on the structure of the data and the query. Autonomy's going to give you that answer and it's going to say, and by the way, here are the next things that are most likely probably related to what you've asked for based on the patterns that we've inferred out of the information. And so they kind of go hand in glove. Like I can take a use case problem. I mean, I gave you the example earlier of someone, I've got 40,000 phone calls that come in between 10 and 11 and they're this long in between this period and they come from this GPS location and they're this customer and that's all metadata and I've got those, but how many of them were upset customers? In my call center, right? And by the way, what did they talk about? So I can analyze that data and say, okay, great, now I've got 40,000, now I've got another 50 or 60 more data points that I've used autonomy to tell me what's happening and I can now, now I've got a much richer view of what's happening. That's an example of combining a structured analytic query with sort of a deep conceptual one. And it works great. I mean, it's great for things like audio and video and really hard sources to get. Well, that's, yeah, when you start talking about audio and video, now, you know, that's really, the amount of data being created in that genre, it's just outstanding. It's unbelievable, the amount of data points inside of video, so we do this deep video indexing. Interestingly enough, when you look at video, I can do a search on YouTube or something. You're searching for the label. You realize that you're not actually searching the video content. If I do a video about rabbits entitled dragons and you search for rabbits, you're not going to find my video, right? So what we do is actually watch the video and we look for the words on the screen, the words that are being said, the stuff at the bottom, the shape that's in it, and if it changes and, you know, rabbit-shaped things, it'll find rabbit-shaped things. So I think there's a lot of richness in that because we have a history in video surveillance and facial recognition and those sort of technologies, which are the same thing, it's pattern. Does it match or does it not match or is it close enough? Because it's never going to match. If it's never going to match, what's close enough? Right? It's close enough and then it gets the job done and that's where the value comes in. Exactly, it's the signal from noise and that. So in your answer to the question, working with Vertica is very exciting because we generate a lot of data. We generate, and so we can give that unique answer, but we can also then pair it with, you know, all that extreme structured information which gives us a really, I mean, it gives a holistic view of the big data problem. Fantastic. Well, Brian, we are out of time. Brian Weiss. My pleasure. Vice President of Marketing and Autonomy, thanks so much for coming on. Thank you, everybody. Nice to be here. We'll love to have you on again in future events, looking forward to it. Everybody, stick around. We have more coverage today. We're going until mid-afternoon today, covering all the action here at the HP Vertica Big Data Conference in Boston and we'll be right back. Thanks.