 Hey, welcome back everybody. Jeff Frick here with theCUBE. We're on the ground at the Cosmo Pollutant Hotel and Resort in lovely Las Vegas. It's going to be, I think, 120 degrees today. Very appropriate for talking about IoT and Edge machines because it's not a nice cool data center like we see when you go to a regular data center. So we're really excited for this next segment to be joined by Jeff, or excuse me, Greg Petroff. You are the Chief Experience Officer and Chief Evangelist of Predix. Welcome. Thank you. So we're talking a little bit offline. You were employee number four and kind of this Predix journey up there with Bill who's been on theCUBE a number of times. Wow, how far have you come? We were there at the 2013 launch event at the Jewish History Museum in downtown San Francisco. Wow, you've come a long way. Yeah, I know, it's kind of crazy. I mean, my first day at GE Digital, I walked in and the only person there was the security guard. So we're 1,800 people in San Ramon now and 28,000 people across GE. So it has been a remarkable rocket ship to be on. It's amazing to me every time we go into the San Ramon office, how much it grows and grows and grows because it's not easy to hire good talent in the Bay Area. It's a super competitive marketplace and the fact that Bill and the team have been able to attract that much talent that fast. I think it really speaks to kind of the opportunity and also, you know, we always joke because of the cute GE commercial that's on TV that a lot of people see with the software developer kid who's not working on locomotive, she's working on software. So you've really been able to build a significant presence in culture right in the heart of Silicon Valley. Yeah, I mean, I think a big part of our appeal is that people are working on things that really matter. If you can make an airplane use its fuel more efficiently, that's a huge outcome for economies and business but also it takes carbon out of the atmosphere. These are all things that people actually feel good about when they go home from work and say, what did you do today? Well, I actually changed something that's going to change the world. So that's been a big part of our appeal to developers is to say, look, you know, you can go work in a lot of different places in the valley but this place, this place is really purpose driven towards infrastructure and making it work well. And then the problem space is just really fascinating. I mean, it's totally different than enterprise software. There's a lot of different problems. It really benefits from all of the new cloud tech, the tech stack in ways that are really unimaginable before. So it's a lot of fun. But you were in enterprise software before and then looking at your background on LinkedIn, you're at the NASDAQ for a while. But that's another mission critical system that can't go down. It's not like Pokemon Go suddenly if your ball's not working, right? So these are really important systems that got to stay functional. So then the developer world, how do you compete? I mean, you need your own developers. You need those 1800 people in San Ramon but to really make this thing go, you got to get outside developers. And clearly, to a 1700 here on your first ever developer show, that's a pretty good statement. Yeah, yeah, that's right. I mean, we're excited to have the community here and we're trying to build a tribe that really is interested in industrial solutions and industrial outcomes. It's a challenge. I mean, building a new community is always difficult but I think one of the key decisions that GE made early was to make predicts an open platform. So it's not just GE that's building capability in the platform. We've got partners like Technohendra and emphasis building capability. We have customers who are buying and making solutions but also adding capability like Pitney Bowes. We're going to be out later today. They're going to talk about a location data service that they've created and added to the platform. So it's really exciting. We have all these people making stuff and adding capability. And our belief is that, you know, no one company can solve the problem. I mean, GE is a big company. We've got a lot of expertise in operational technology. We get the industrial world but we can't do it alone. We need an active community of people who are on the journey with us and we want to incent them. We want to make them successful. And if they're successful, we will be too. Right. And how does that conversation go in the early days? Because clearly, not only are you not the only company that can execute it, but there's a lot of installed gear out there of which yours is probably part of that operational chain but there's other stuff as well. You mentioned you were on the IoT consortium. Explain a little bit how that plays because we don't see as many consortiums as back in the day kind of with the rise of open source. That seems to be more the mechanism for setting standards, if you will, versus classic consortium. But there is an IoT consortium. Yeah, I mean, we're early, right? So there's kind of these things, groups emerging. So the industrial internet consortium, which GE is one of the founding members of, is not really a standards body. It's really a group that's trying to facilitate a conversation with technology companies about this marketplace. Get people literate about the opportunities, help people understand through test beds that the technology is mature and capable of solving real problems. There are a lot of companies in that. I think GE's perspective is really focused on interoperability. I think it benefits everybody. Our customers are asking for single pane of glass to view their environments. And we may have what we think is the most important real estate, the gas turbine or the jet engine or the locomotive. But there's a whole bunch of other pieces of equipment from other vendors at our customer sites. And they're frustrated because they can't see across those systems. So we know we have to play well with others. We have to have people who are building complementary technologies have a big tent. And it's really important for us to be in these organizations to express our point of view and to get others to come join. IIC is pretty interesting. It's grown. There's 220 members now around the world. It's very active in China and Europe and in the US. So that's been rewarding to see. And then we're seeing other groups like the Open Connectivity Foundation, which is looking at machine to machine protocols. That's a standards body. That's got strong membership from Samsung and Cisco and Intel and Microsoft. And that group is really looking at how devices can talk to each other. And then there's a new group that we've just joined, the OpenCloud OpenFog Consortium, which is looking at the edge. We call it the edge. They call it the fog. And this is really about compute at the edge. And we're really interested in that because if you look at the industrial world, we don't always have connectivity. The devices are in very remote locations. They really need to be able to have some autonomous degree of computation and experience. Can we pull the cloud and move it all the way to the edge? Instead of having an on-prem site solution, it's really cloud thinking close to the device. And so it's really important for us to participate in these groups. So let's unpack that a little bit because I think that's interesting. The conversation around Predix is all about the cloud, cloud, cloud benefits of cloud, elasticity, huge impact on big data. But there's this whole edge component which isn't necessarily at the top level of the conversation because oftentimes, as you just mentioned, you don't have time for latency. You don't have time for connectivity. You don't have time for all kinds of reasons. So now in this edge piece extension of the Predix, if you will, now you're bringing compute, store directly to however close it needs to be, device controller, you know, all these different pieces of the product. Yeah, I mean, we're at an interesting moment. I mean, if you look at the cost of computing, right? I mean, Edison chips what? This size, size of an SD card. You know, we can put 100 cores next to a machine and it doesn't really change the cost of goods of that machine. I mean, especially for high value assets. So why not have a whole bunch of really awesome analytics running right next to the machine and then that information gets pushed up to the cloud and then we can use that with a cohort of machines that are similar, drive that to get meaning and understanding, bring that information and package it back down. We call it digital twins. This idea of a model of that machine running in simulation mode all the time. So we know the operational behavior and we know it's expected behavior at the machine. So that's a big part of our strategy and if you look at Predix, I mean, we sometimes call it an operating system because it's cloud all the way to the edge and it has to exist in different layers. Not only at the edge device, but in the communication infrastructure that connects the edge to the cloud. There are variants of Predix technology along the way to add, you know, and you can imagine why that's important. Industrial space, security is sort of paramount, right? So you really have to choke old moments along the way. Very, very carefully. And that's a big part of what Predix is trying to do. And then there's the appropriate place to do the appropriate information processing and that's an important part. And then our customers also have needs. They may say, look, I don't want certain kinds of data leaving my industrial site and we're okay with that. So we have to kind of have a system that allows us to do that. Yeah, or they may want to optimize for different things. I think a lot of times when people are kind of nascent to the IT discussion, they talk about sensors. It's all about sensors. Sensors are just the itty-bitty little piece at the very, very tip of the sphere. There's all kinds of controllers and as you said, kind of macro levels, different levels of continuity in which you want to manage things, not down to the individual sensor. So there's a lot of complexity opportunity to tweak and change things. Yeah, and just on the sensor side, I mean, you saw today we showed our Predix kit. It's a box developer can grab, it's got an Intel Edison chip on a board, bunch of sensors, Predix machine, which is the software instantiation that sits on the edge built in, developer like 10 minutes can get up and running and connected to the Predix cloud. Those are kinds of things that we're trying to develop and that when we have industrial kits, they'll mean it allows to go out to existing infrastructure and basically connect them very, very quickly. And if you think about the industrial world, I mean, there are systems that are out there still being productive that were built 40 years ago and don't have any sensors on. So if we can retrofit those environments and make it like the cable guy, they basically going out, installing a box, putting a bunch of sensors, turning it on and the system recognizes that system right away. And then we start listening to that data, we can help our customers get a lot more out of their existing assets, even if they're older assets. So is that the on-ramp path? I mean, I'm just curious, if you've got a big contiguous system, it's a field of pump and oil or wind turbines or jet engine, this is not something you can have a stop start, introduce some new technology and reboot the thing up, you've got a big system. So how do people kind of get started? How do they kind of start their journey, take advantage of predicts? Or even in the one you said, a old factory where it's working. I mean, I think the first thing is getting access to your data, getting connected, right? And there are a lot of different ways you can do that. There are, predicts has a bunch of APIs that allows you to ingest data from different control systems. We have members of our ecosystem are building those capabilities for areas that we don't have domain expertise on. You can do the low tech version, which is you just install sensors on the equipment, have predicts start reading it. And at first, it won't mean very much, but if you get a month, two months, three months of real time data, you'll start to see signal to noise in that data and get information and allow you to monitor it more effectively. I mean, if you're a developer, I mean, it's really about learning about time series information, like how do you take machines, which for the most part operate very well over long periods of time. But then when they do have problems, there's little signatures, and it's hard to figure out what those are, right? So that getting instrumented, learning how to play with it, using the tools, building your first app. These are all things that we wanna make really simple for people so that they get up to speed with the value. And then once you do that, you can start to make these really kind of hybrid mashup applications where you might have your ticketing system and all of the information coming off a set of machines and access to your field force, put those things together and really make applications that add productivity to teams and improve services. Okay, so you've been at it for a couple of three, four years now. What are you looking for next? What's kind of the next big hill to climb, next big challenge you're looking forward to taking on as this kind of evolution continues on IoT specifically, and Industrial Internet is a subset, and G predicts is a subset of that. Well, I think there's a couple of things. I mean, part of it is getting the critical mass and scale so customers feel confident that they can grow quickly. So that's part of an event like this where we're building a developer community inside. We've got 1700 people here today. We hope to have 20,000 developers by the end of this year. We're over 10,000 right now trained in predicts. So that's part of it. So I'd like to see us grow the community of people of makers who are making stuff. And then we had a hackathon yesterday and there were so many awesome ideas in that hackathon that industrial customers would eat up in a second. What one? One was great. So I'll give you, I can't remember the winner, but we had one that I really loved. They're a fun one that you have stuck out. So it was for autonomous vehicles or for self-driving cars, right? So you get the car, you turn it on for the first time, and they built a service that automatically recognizes that vehicle into the predicts asset model and starts tracking it immediately. No human intervention, no capabilities necessary, and the person who's managing that fleet all of a sudden sees a new vehicle show up. And it sounds really stupid and simple, but that just by itself is incredibly valuable because you don't need to have someone set it up. Automatically you have visibility into the performance of and the key statistics and health of that particular machine. And this group built this really nice service that showed that in a 24-hour period of time, right? So that was awesome. You asked earlier about some things I think we need to focus on. I think we're really doubling down on machine learning. We know that there are kind of three buckets to understanding the industrial world. There's the sort of typical data science things that you might get from the business intelligence world and we've got to be great at that. There's the sort of social knowledge of operational technology that you can embed in algorithms. It's like, you know, what is the guy who's been there for 30 years and he knows what the sound sounds like, me, right? Can you capture that information, that tacit knowledge and actually make that into algorithms? And then the last part, machines have physics models, right? We know the material science of the machines. Can we build machine learning algorithms that understand how machines operate over time so that if we know a particular operator runs their equipment hot or harsh or there's a lot of humidity in their environment, what's going to happen to the machine over time? Is it going to get rust? Do we need to change the oil more often? So we can actually get down to, if we get that right, we can maintain each machine individually. We don't have to do machine-based maintenance where it's like every 10,000 hours. It'll be based on the way you operate it in the construct of its location and its behavior. Here's what we predict is going to happen to it in the future based on how you're going to operate it. And then that makes a huge difference in terms of how you maintain it and it gives economic value to our customers because imagine arbitraging your assets, right? You have the ability to say, you know what, here's how I have behaved, but I have an economic opportunity. What if I run my systems really, really hard? I know it's not going to be good for them but if I know exactly how bad it will be, I can measure that against the economic opportunity that I have and I can make a business decision. You can't do that today. So if you can, that's really going to change we're going to get more resilient infrastructure. We're going to have people and businesses are going to be able to take advantage of economic opportunities and they're going to be able to be more capable of servicing their customers in ways that we don't even know. Right, I love that example. I love what you didn't say, with the digital twin you can actually test it too. That's right. You can run it out for a while. That's right. See what you think is going to happen. This is a cool thing about digital twins. Every asset has a past and a present. That's all we know right now. With the digital twin, we know all of its potential futures. Not just the future if you carry it forward, but we can simulate a whole multitude of different ways that you get scenarios. So that part of it I think is really cool. Yeah, it's amazing. I mean, I think just this whole, moving away from a world of sampling and moving away from a world of averages and moving away from a world of means to moving into the specifics of that individual unit, whether it's a person wearing their Fitbit or whether it's a locomotive or whether it's one individual piece within a big factory floor. I mean, we're super interested in this idea because what we're seeing is, if you look at the last 20 years, you had ERP systems, which were systems of record, and we had lots and lots of applications. You needed an application for your travel. You needed an application to hire someone. You needed an application to do a PO. We all suffer by having millions of different applications. They're really useful and they made enterprise really successful, but it's about that record, right? And then we had systems of engagement. Easy like Salesforce or Facebook that connected people like you and I together are human capital networks together. And now we're moving into this system of assets where in the moment, right? I know where you are. I know why you're there. I know what you're interacting with. Why don't I collect all the information from the system of record, from the system of engagement, from the systems of asset and put it all at your fingertips at that moment so you can be the most productive, smartest guy at any moment in time when you're interacting with the things that you need to interact with. And we see that happening. You're starting to see that in our personal lives, right? So you see that with applications like the New York Times where they kind of track what you like to read and they customize it for you automatically or your Facebook feed where they, the people you interact with the most show up in the top. We're building applications that recognize how you work with the industrial equipment that you work with and are going to start placing that information at your disposal based on what you're doing which is like mind blowing. Mind blowing. Well, exciting times, Greg. I think they got the right guy for the job. So congratulations to you and building the team and really for a great event. 1700 people for the first ever developer conference. Like I said, we go to a lot of shows. I don't know that we've ever been to one where the first had that many people. So congratulations. All right, thanks very much. Hey, thank you. Thanks for stopping by. I'm Jeff Frick, you're watching theCUBE. We're in Las Vegas at the Cosmo G predicts developer 2016. Be right back.