 Live from Houston, Texas. Extracting the signal from the noise. It's theCUBE. Covering Grace Hopper Celebration of Women in Computing. Now your host, Jeff Brick. Hey, welcome back everybody. Jeff Brick here with theCUBE. We are live in Houston, Texas at the Grace Hopper Celebration of Women in Computing 2015. It was 8,000 women last year, 12,000 people this year, probably 12,000 women, 1,000 men. The conference is growing like crazy. It's a, celebration is the right word. It's like no other conference that we cover. And so we're really excited to get the keynote. Hillary Mason, Cube alumni. We haven't seen Hillary since Big Data, New York City, 2013 I think was the last time we had you. It's a welcome back. Yeah, that's right. It's nice to see you again. Absolutely. So when we last talked to you, you were doing the VC thing. You were helping out at Excel. Now you've founded another company, Fast Forward Labs. So congratulations. Thank you. So give us an update. What's Fast Forward Labs all about? So what we do at Fast Forward Labs is look for technologies around data and machine learning that are emerging, that might open up possibilities for new products or new features or new business opportunities, but that are not really well understood or commoditized. And we pack them up in a form where our clients can understand what it is, how it works, who's doing it, what's in the market, and really figure out how to plan their product and business roadmaps around these machine learning technologies. And just to give you an example, we just finished a project on deep learning for image analysis, which is having algorithms look at images and tell you, oh, there's a dog and there's a can of soda in this picture. And that's something that's new. It's a breakthrough over the last couple of years. And now people are starting to really build it into their product roadmaps. It's a really different process because it used to be, if you built a technology and a product without an application, oftentimes you just burned up a bunch of VC money and didn't go anywhere, right? The much better path was to have the problem first and then build the technology to go attack it. But you're kind of flipping it the other way because these algorithms and stuff are moving so fast and the technology's moving so fast that the business people and the people thinking up the applications can't hardly keep up. So it's a pretty, it's an interesting kind of twist on it. Exactly. So our goal is to make our clients as smart as we are about what's possible with their data and machine learning and data science capabilities. So one of the things that's come up around machine learning and kind of algorithmic suggestions, right, is if it's done right, it's magic. If it's done poorly, it's creepy, right? So not right actually. I take issue with the phrase of that. Because machine learning is not magic. It is fairly simple mathematics deployed on data at scale. And we act like it's something that's inscrutable, that it's a black box no one can understand. But we exist to say that's not true. You can understand it. And yes, we do see these bizarre cultural consequences of machine learning applied sometimes in products where people are not aware of what could happen in edge cases where you might classify a photo, say of a concentration camp as a jungle gym, which was a recent issue that came up with a photo product from a major tech company. And that's not algorithmically wrong, but obviously very upsetting in the product. So part of what we do is actually explain how these algorithms work, how the math works, how the code works, how the data works, and then help people think through the potential cultural side effects of deploying one of these things in a product. And it is not magic. It's, but it feels like, but it should feel like magic, from my, I mean, just, it's crazy. Yesterday, what did I search like Southwest Airlines or something, or no, I searched, I think Grace Hopper, I was trying to get to the agenda. And it came up with my registration number. And my whole thing's like, wow. And you know, and of course the classic case is the target example, not to beat a dead horse, but that's the one that everybody knows about. So even if it's not magic in the machine that delivers the result, there still now has to be this kind of next factor, this, you know, is it appropriate, is it ready? Do we have to retune the algorithm? Do we have to run it through a filter? How do those things kind of get addressed? So the best data products are the ones that you can use without any awareness of the technology behind them. And in my keynote, I talked a bit about Google Maps as the best example, in my opinion, of a product that integrates a huge amount of complex technology. Like if you bring up the Google Maps traffic view, you look at it, you make your decision about where to drive and you turn it off. You can use it without any idea about the prediction algorithms going on, the data they're collecting, how they're displaying it to you. It's a great example of a product that is really simple to use, but you don't have to understand that. From the perspective of the product designer, I think we have to grow our engineering and product design processes to take this sort of evaluation into account where it's not just, is the product usable, but is the product providing value to the person using it in a way that is not going to upset them or make them creepy, of course? Also, we tend to ask things like, what are the ethical considerations of building a product around this technology and how could it possibly go wrong? Because you need to think about that stuff in your product development process. Which brings up another question, a great topic. We're at another show and they were talking about the moral issues around writing the algorithm. What are you creating this algorithm for? What is your expected outcome? But it seems like kind of a fallacy to even attempt that beyond kind of the first order, because now its algorithms are built on algorithms are built on algorithms, right? And the data keeps increasing and changing. So even if I have all the best intentions today, and we bring you guys in and you make a great little product for us, who knows what either the data or the input or the social mores are going to be two years, three years, four years. I don't even want to think beyond that. So how do you help customers kind of think those issues through or do, is it more let's do it, see what comes back and then we can start to make some of those judgment calls. There's an element of both. So you really have to be thoughtful about this from the beginning of the product design process to think about, if I build this algorithm, what are the potential edge cases and how might this go in a way that I don't want it to go in my product? But then of course there are things you can't predict. And in the keynote I shared this example of Sony created a product called the Ibo, which was this robot dog, it's really cute, right? And they've debuted it in 1999 and they finally ended technical support for it and repair service in 2013, which means that now these dogs are dying and people are holding funerals for them. They've become incredibly emotionally attached to their little robot dogs, which I can understand, dogs are great. And you don't have to scoop the poop on the Ibo too, I don't think maybe it does a little plastic one. So this is the kind of cultural consequence you could never have predicted when you're an engineer building this little robot dog, and yet it had some real emotional impact for some people who really love this device. And so you can never predict everything, but if you don't even think about it, you're not trying to be thoughtful about it in your product design process, you're never going to do a good job to begin with. Right, so you are such a pioneer in this space and obviously back at Bit.ly, where we all just thought we were using URL shorteners, we never really thought through the fact that this is providing all kinds of good information about who's sharing what. The world has really changed since then. People are realizing there's so many data sources, your own data, other data, outside data, public data, the speed of the data creation, now the speed of the tools, Hadoop, and Spark, share with us your perspective on how those types of technologies are really advancing, you know, kind of the machine algorithm world, and we're just going to go next. That's a great question, and it's something that I'm pretty enthusiastic and optimistic about because if you think back, even just five or six years ago, getting Hadoop to work on a large data set in a reliable way with a good response time was itself a technical challenge, and now I can spin up an elastic MapReduce cluster on Amazon Web Services with one line of code, and it works most of the time, and so the reduction in cost, in time, and in friction that we've seen over the last five years has been remarkable, and then when you add in things like Spark and other in-memory computation technologies, we get to a point where we can now do online machine learning and online computation in a really cheap way, from an infrastructure perspective, so what we've seen is that people have been able to now play with the technology to start to build it into their products, to see a lot of success with that, and when you look at what's going next, you can sort of think about it coming up the stack, so first we had to get the data in one place, then we had to build a way to analyze it, but we didn't really care how long it took, then we cared how long it took, and now it's real time and in memory, and the next thing, you know, I have a lot of theories on this, but I think we'll start to see some algorithms commoditized in that process, so you'll no longer have to write your own code for a recommendation algorithm, or for, you know, sort of an online averaging or variance model, and a lot of that is happening in the open source community. We're starting to see the same thing around analyzing data that was previously too difficult to analyze, either because it was too dirty, so we have products emerging to clean data, or because it was too complex, as in the case of image data, and I'll stop there, because I could talk about this whole thing. No, I love it, I'm just, as you saw, we have to have like Hillary Mason Day sometime, we'll come out and do like eight segments with Hillary Mason. No, I mean, just concepts are just coming in my head as you're talking, the whole concept of sampling where before everything had to be sampled, right, you didn't have the opportunity to take the whole data set, that's no longer the case anymore. And not in the technical age that we are in right now. And the other thing is just the opening of innovation, especially here with, you bring in some young engineers, they don't have to set up the cluster, they can actually start to execute their theories, their ideas and kind of get out of setting up the infrastructure, which I think is the big power of AWS and how it is really fundamentally changing industry. Absolutely, and I first realized this back when I joined Bitly in 2009, which is a social media analytics company gathering a massive amount of data for the size of the team and for what the product was, which is a, it was a URL shortener, right, short links on the social web. And I realized that this tiny little company with relatively modest funding was able to do analytics at a scale that previously would have been so expensive, it was just out of reach. And that that data was such a rich, interesting resource. Right, and it's often spoke about, we're certainly in a little bit of a bubble right now, compared to the old bubble, but the old bubble you had to take all that VC funding and buy sun boxes and build out data centers and Oracle databases and all this other stuff where now, like you say, you can spin up an instance on AWS, get started, get your app out there with a much less significant investment upfront. Absolutely, and we're seeing that for not just building a web product or a mobile product, but for building a data product, which is pretty exciting. So let's talk about one of the other things going on here to need a board is the variety of companies that are here, right? Googles here, Yahoo's here, LinkedIn, Uber. Those two aren't that different. Right, the usual suspects, right? But then you look around and you see nationwide insurance, we just had Bank of America on, Target is here, Best Buy, Johnson & Johnson. So clearly, and we talk about it all the time in theCUBE, every company's a software company now. It's just what they happen to wrap their software around. So it really opens up opportunities from a computer science point of view to go into lots of different places, not just kind of the traditional things where you might suspect. No, that's absolutely true. There are finance companies here, insurance companies here, retailers. I think any company large enough to get a team here might be here. And it's great to see that, and it's great to see the variety of career opportunities that are on offer here for all of the young engineers in the room. And I've talked to a bunch of them, and they are all so excited. That's great. So what are you working on next? When we talk to you next year, when we have Hillary Mason Day, back at fast forward later this year, what are you working on next? Building the team? Any kind of big things you're working on? Yeah, so we work in generally a quarterly cycle where we pick one idea to look at every quarter. And currently we're looking at language summarization. So this idea that you might have an article and you want to pull out the most interesting sentences, or you might have 10,000 articles, say on the Democratic debate, and you might want to understand what's the distribution of different viewpoints in all of the coverage, and what does that look like? Or you might be, say, an insurance company who has 10,000 customer profiles and want to understand what kinds of customers do we have? And when a customer appears, what products are best for them, how do I really think about that? So we're looking at neural models for language analysis and summarization right now, but we have a long list of ideas. So next year it'll be something else entirely. You must love all the horsepower that you could bring to bear on these problems like you did in that. I love what we're doing. It's getting cheaper and cheaper and cheaper. And we monitor that pretty closely. I bet. The economic constraints on some of these techniques, once it gets below a certain point, it becomes something you can play with, and then you start to see this plowering of creative products around the different techniques. It's amazing. That's all. We had Jason Stowell on from Cycle Computing, and they basically get excess capacity from Amazon that people have purchased but aren't using and spin up massive capacity for heavy, heavy jobs that are single-run, life-testing, et cetera. So it's a really crazy time to be able to have that at your disposal. Yes, and it just gets more interesting every year. So last question. What's your favorite hamburger? Cheeseburger. Can you say it? I'm Hillary Mason, I love cheeseburgers. What do you recommend? Well, so your access to cheeseburgers is somewhat geographically constrained, but if you happen to be in New York City, I recommend a little pub called Corner Bistro in the West Village. The menu's on the wall, and they serve you everything on plastic plates, and it's amazing. Corner of what and what? Corner of what and what? Oh, it's on Greenwich Ave and Jane Street, I think. All right, and what about not in New York City when you travel? So I am always up for trying burgers. I haven't found one in Houston yet, but if I'm on the West Coast, I always swing by In-N-Out, which is a, you know, it's in and out. We know In-N-Out, for sure. Get the animal, protein style, all kind of crazy stuff off the menu. Hillary, thanks for stopping by. Thank you so much. Best of luck to you at Fast Forward Labs, and we really enjoy getting you on theCUBE. Likewise. Absolutely, so I'm Jeff Frick. Next week, or no, two weeks from now, we're going to be at Oracle Open World, one of the biggest tech conferences there are. We are really excited. We'll be out in the middle of Howard Street. So hopefully you can tune in or stop by if you're in San Francisco. I'm Jeff Frick. We're in Houston at the Grace Hopper Celebration of Luminant Computing. We'll be right back with our next guest after this short break.