EP016: Cutting Edge Computing - An Interview With GE Digital's Mike Dolbec
|Dec 04, 2017|
According to the 2017 Gartner Hype Cycle, many emerging technologies linked to IoT like machine learning, cognitive computing and IoT platforms find themselves in the "peak of inflated expectations" phase. At such an inflexion point, it is vital to accurately evaluate the promise these technologies hold for us.
In the first part of the very first episode of this exciting new series, Mike Dolbec, Managing Director of venture investments and M&A at GE Digital has a candid discussion about edge computing, blockchain and other nascent technologies in the Industrial IoT space.
Erik: Welcome to the Industrial IoT Spotlight. Your number one spot for insight from industrial IoT thought leaders who are transforming businesses today with your host, Erik Walenza.
This is an episode of The Ventures and Industrial IoT series brought to you by GE Ventures. In the series, we explore success factors and challenges in industrial IoT markets with CEOs, investors and experts. Welcome back to the industrial IoT Spotlight. I'm joined today by Mike Dolbec. Mike is managing director of Venture Capital and M&A at GE Digital. He is also an investor with Alchemist Accelerator and with i3 Equity Partners as well as a board observer on FogHorn Systems, Data Torrent, Mocana, MAANA and Ayasdi.
We're very excited to have Mike here. Obviously, he has a depth of experience in the space. We're going to be talking about three topics today. First, what's hot versus what's hype in the industrial IOT space. I believe we have a lot of technologies and companies that could be classified as either or both of those, second challenges and opportunities, specifically from the adoption standpoint; and then finally, looking at partnering beyond funding, so how are companies like GE, partnering with startups, not just for venture funding but also for go to market for support, with strategic direction in R&D and in other aspects.
We're going to break this podcast up into three parts as normal. We'll be publishing the first on a Monday, the second on a Wednesday and the third on a Friday. So make sure you join us for all three.
Mike, to kick off, give me an overview of the technologies that you think in the edge computing space are most likely to radically change industries, potentially change the world in the coming 5-10 years?
Mike: So I think Edge Computing is much more of a trend now than it has ever been in the last 12 months. It's both hype and hot. But it's been hyped for a long time and now it's turning into reality and picking up steam. But the simplest one is computing that isn't happening centrally in the cloud and maybe is intelligently the work, is intelligently split up and distributed and happens for good reasons where it should happen best, usually, for economic and latency reasons.
So I think the thing that's rocking the world or is in the process of doing that right now is edge computing used to be a novelty, kind of, wouldn't it be great if that happened someday. And now that the least as far as we can see, our customers are extremely excited about the concept of not having to do everything in the cloud. And they surprised us with how interested they were in things happening in a distributed way at a variety of places, which collectively we call the Edge. Most people may think about it as on the device or in the factory, but there's really a variety of edges. I think it's surprised people who had strategies based on cloud computing exclusively or had just thought they convinced all their customers to trust cloud computing.
But it's really turned into a world where every company we meet and talk about potentially investing in or partnering, one of the very early questions is okay, that's great, that's a wonderful strategy. What's your edge strategy? What's your strategy to split the work as I described it so that efficiently parts of it happen in variety of places at the edge and then the rest of it, maybe the heavy lifting happens centrally? And I think that trend of entrepreneurs and large companies thinking through okay, what is our strategy and we no longer sell the one monolithic thing, we have to be smart about where the work happens and smart about thinking about how do we split it and break it up? That's a huge trend I think that's changing quite a few people's business plans.
Erik: And I think you're spot on that this caught a lot of companies off guard, the focus had been on cloud and certainly cloud is going to be a very important aspect of the industrial IoT. But a lot of companies now for latency purposes and security and other reasons are looking now towards the Edge. I think from most people's perspective, when we're looking at industrial IoT solutions, we're not looking at the infrastructure, we're looking at the SaaS aspect or the application that a company might actually be engaging with. But maybe the most important part is the piping out the actual infrastructure that's making this possible where it wasn't possible a few years ago. What are the actual technology advances that are making Edge now technically feasible cost-wise feasible where it just wasn't there a few years ago?
Mike: To me, it feels like the capabilities have been in place for some time. But I think the generic answer would be the typical computing power, embedded systems are now super capable and memory is almost free and bandwidth is a lot faster but still has a expensive price. The laws of physics are still in place. So it's still cost something to move from where you find it to where you wished it was.
And cost region, compute in cloud computing is very attractive and falling rapidly, but it's still what it is. It's almost as if the pieces had been in place for some time. It had been a vision that people knew was going to happen. And then a few early innovative people started building Edge, what we would think of as traditional Edge gateways and demonstrating what's possible confirming the vision that many people had, kind of saying, yeah, I told you so, I knew this was going to happen. I told you it was a good idea. And then it sort of snowballed after that.
But we started telling our customers about this particular the GE version of this vision four years ago. And like I said, before, we were overwhelmed by their response, even though our version of it wasn't demonstrable until like Witcher maybe three years ago, two years ago, two and a half years ago. But it just like snowball. Once it starts going and things started creating to it, it picks up speed and heads downhill pretty well. I think it's just everybody coming together just the right time and then swing set off on Earth.
Erik: One of the interesting aspects of the industrial IoT is certainly the ecosystem play, which is that it's not one company with one killer app or new technological breakthrough that makes it possible. It's 10 different things that are all maturing, and when they reach a point, then the system becomes viable. But let's look at a couple of specific technologies.
Mike: The industrial world had a bit of a head start because there's a long history of machine to machine networking. So it was kind of the Jurassic period of edge computing, unlike autonomous vehicles. In that case, there was no other way to get a vehicle to make its course correction in time before it hit something, so edge computing was critically important to them. The machine and machine phase of industrial computing has been going for some decades. In some cases, it was difficult for some people to unlearn some lessons they had learned before. But the transition to edge computing was pretty smooth relative to some other industries.
Erik: So looking at a couple of technologies and specific blockchain, what do you feel is the potential for blockchain on the edge? There's a company called IYOTO which is building and they call it the tangle not the blockchain, but the tangle that specifically designed for machine to machine, no transaction costs, scalable communication, other companies are also looking at applying blockchain in industrial settings. Do you see this as a significant opportunity? Do you think it's a concept that doesn't necessarily have proof of concept or proofs in operations yet and we'll have to just see if this actually pans out as a viable technology?
Mike: So I'd say we're very excited about the prospect of what might be achieved because of blockchain, extremely excited about the number of people, they're innovating. They're trying to come up with a number of commercially significant uses of blockchain in IoT, almost as many as there are in FinTech. The uses tend to be slightly different. They often revolve around identity, authentication, access control.
And yet, we're still in the very early innings of a long baseball game, because I think our criteria, it has to be more than novel. Just because you can do something in a novel way with blockchain doesn't mean that's a commercially attractive investment. So we placed a pretty high bar on these things, it has to be better, not just different. And so our investigations of the use of blockchain often happen in this weird world where if you can picture the IP network is extensive and imagine the edge of the IP network is all the end points that have IP addresses. Well, in industrial, there's stuff beyond that edge, the non-IP devices that are connected eventually to IP network.
And we can't use today's wealth of rich infrastructure to authenticate non-IP address things that least not easily. So that's one thing we've been engaging with a number of entrepreneurs on that use blockchain to extend the capabilities of the network. If you have a policy for identifying and make sure that the jet engine is still what it says it is, but something that says it's a physical device, it's even harder to verify but it's still what it says it is and it’s not spoofing you when it's not an IP address accessible device.
One of the other novel issues we face which I didn't realize until an architect pointed it out to me is most people in the modern infrastructure use some sort of PKI to authenticate IP things. You have credentials, and you say your computer and your company about this computer and it's okay for this computer to be on that network. Well, we make devices that are like computers but they last longer than anybody's Public Key Infrastructure is designed to last 20, 30, 35, 40 years.
And so we are almost guaranteed that whatever method we use today to authenticate that, that thing is okay and it's still the thing it's claiming to be. So we have to think about situations where the infrastructure we use to identify and authenticate things is wholesale obsoleted and somehow survives a very long lifetime of product. And we've seen some interesting uses of blockchain in that perspective as well, either lives a long time, or it's an easy way to come back to a thing and re-authenticate it. You have to go out in the field and re-identify it and make sure it's still good. But you have to hook it up to a new authentication infrastructure at some point because we think that the ones we use today they won't last the 10, 20, 30 year cycle that we have for some products.
Erik: I suppose with a high value asset like a jet engine, maybe the costs can be borne right of every 5 or 10 years re-authenticating. But for a lot of IoT devices, the cost of the asset is low enough that it's just not going to be feasible to send people out and manually do any work, it's going to have to be an automated process.
Mike: And you don't usually worry about that because consumer products like your phone generally has 9-18 month, maybe 24 month lifecycle before eventually that version of the operating system isn't supported anymore or a new version won't fit in your phone because there's no memory. So eventually, the end of life of those things is not 10 years, not 20 years. So these long life time, very expensive, exotic assets have unique requirements that nobody thought about when they were building today's public key infrastructure, at least I don't think they did. So many of the assets have the problem, but ours do and so we're thinking about…
Erik: Mike, what are a few companies that are the most exciting that you've had the chance to work with? I think FogHorn, Data Torrent, doing very interesting work in the edge, what are the companies that you most enjoy or that you're most excited to watch as they mature over the coming years in the space?
Mike: We have an investment in FogHorn and so I love to talk about them. They're a leader in analytics for edge computing, and now machine learning. But the challenge is if you had edge computing as a resource at your disposal, how do you make the effort to exploit it a low bar so that people in factories and people who develop software for trains and jet engines and so forth can exploit it without needing to become super experts in exotic real time programming languages?
I think that's the real breakthrough with FogHorn is there ultimately making it accessible for a relatively talented but not PhD enterprise employee to move things that were usually reserved for doing later by data scientists in the cloud, move them all the way out to the edge, maybe next to a device so that you arbitrage either the cost of moving and storing the bits and/or exploit the latency of being able to react, notice something important has happened, and then react to it very efficiently in a short amount of time. And those two things combined together most industrial situations like saving money and doing things fast and on time, so people tend to respond to it.
The other fun thing for me is FogHorn, we partner with them extensively. What's more important is us sharing our requirements for different customer use cases which has allowed them to kind of see into the future a little bit farther than other companies. But I think that they're now focusing on ease of use features. They actually understand the lifecycle of what it's like for somebody to engage with a customer, deliver a solution, configure it, set it up, maintain it over time, and they're starting to work on the whole lifecycle issue.
So it's probably the industrial equivalent of the out-of-box experience that consumer goods companies are really good at. So it's not just the phone and how you use the phone in my analogy. It's how easy is it to get the phone provision to the networks. So I think they're delighted that they're in danger of growth where they can now learn enough from customers to start hitting the secondary and tertiary requirements.
Erik: A lot of the heavy lifting in terms of the core technology development we seem to be there at least to provide value in a lot of spaces. But when we talk to end users, a lot of the barrier is around their ability to implement these solutions or even to evaluate how do I compare two different solutions, how do I anticipate changes in technology and so forth? I'm based here in China and it's kind of a running joke that a lot of state-owned enterprises, they might have $10 billion in revenue and 10,000 employees, but their websites look like they were from the late 1990s because they're very engineering heavy on mechanical engineering, heavy on electrical engineering. But their IT teams are relatively immature.
A lot of the IT talent is sucked up by the Tencent and Alibaba is and the IT team. So then you have all these traditional companies that are being faced with a lot of new technology that requires some degree of expertise on the IT side and they don't have the in-house teams. So we need to have solutions that are built for non-technical operators to make sense of.
Mike: Our aesthetic for what's our attention span and our aesthetic for what's acceptable user experience is quite high, thanks to excellent UXs like the iPhone. And it's spilled over into industrial. When you hire millennials into your companies to replace the old farts like me, their expectations for figuring out how to do something without a manual and repetitive tasks better be much easier, the second or third time because it's learned what it is you want. Just mentioning a number of different UX factors that navigation ought to resemble, it shouldn't be a whole lot worse than the amount of effort I'm willing to navigate through to get something done in the social network. And it's not something traditionally that industrial companies have invested in because they could force their employees to use whatever they developed.
But I think now, the rising tide of consumer expectations and experience at home and elsewhere in the enterprise is washed over. So yes, something that each of our companies eventually has their customer feedback tasked them with is to, thanks for making this possible, now make it way more easy to use and way more valuable in terms of my productivity, don't waste my time.
Erik: And this is a great point to wrap up the first part of a discussion, which is what's hot versus what's hype. We'll be discussing next challenges and opportunities from the adoption standpoint. This again, will be published on the Wednesday following the podcast that you're listening to today. So please look out for this coming podcast. Thanks again for joining.
Thanks for tuning in to another edition of the Industrial IoT Spotlight. Don't forget to follow us on Twitter at IoTONEHQ, and to check out our database of case studies on IoTone.com. If you have unique insight or a project deployment story to share, we'd love to feature you on a future edition. Write us at Erik.walenza@Iotone.com
Thank you for joining us for another episode of The Ventures in Industrial IoT series. You can learn how GE Ventures goes beyond funding to support their partners in technology development and commercialization at www.geventures.com