From left, Zachary Karabell, Kelly Liang, Prith Banerjee, Charles Meyers
Mark Bonchek of SHIFT Academy
McKinsey's Bertil Chappuis
Managing Director of Global Technology R&D, Accenture
Senior Vice President of Business Development, SmartThings
Chief Operating Officer, Equinix, Inc.
Head of Global Strategy, Envestnet
Karabell: So as you know from these sessions, these are meant to be discursive and interactive and chatty and not nearly as formal. So let’s have that spirit to it. I am—nor are we supposed to do elaborate introductions. That’s all on the app. So for Charles, Prith, and Kelly, you can find their voluminous and depressively impressive CVs on the participant app online. I’m definitely the odd person out in this just because I think I know what the Internet of Things is, but I’ve had absolutely no role whatsoever in constructing it, and won’t. But hopefully I can facilitate a discussion. I’ve written about these things, I’ve been in business, I know a tad about what these things are about, but I have a humanities background rather than the science background, which I clearly have never emotionally resolved.
But this is one of these patois of Internet of Things and connectivity that has become sort of, like many of the things that this kind of conference and in this area, something that everybody talks about without necessarily a great deal of clarity about what exactly does this mean, is it just another catchphrase that we’ve applied to some degree of communications integration between devices, people, things, all funneled through different systems. And each of you have a slightly different take and position on that from some more experience in the hardware and telecom side and Charles on the security side, and, Kelly, you as well. But I would like each of you, as we’ve discussed, to kind of describe what is actually going on. Is there a qualitative difference, not just a quantitative difference in this interactive network effects? Are we talking about something that is very different or just an extension of whatever’s been going on in the past ten years? And then we can—that will lead us pretty organically into a discussion and then all of you should feel free to jump in, but you do have to jump in with the mics because we are recording, so—not because we can’t hear you, but because of the recording factor. And I think we thought Prith, who’s recently, at Accenture, done a—well why don’t you describe what you’ve done.
Banerjee: Sure, thank you, Zach. So I’m Prith Banerjee, I’m managing director of global technology R&D at Accenture. Accenture, as many of you know, is a fairly big consulting company, technology, strategy, and outsourcing types of consulting. As part of our global R&D, I run our technology labs open innovation program, and so one of the very, very big bets that Accenture is making is in the area of Internet of Things, and I’ve kind of used some of my background. So last job was, I was CTO at ABB, a power and automation company, which was sort of working at the operational technology, how do you make sort of power plants, industries, etcetera, work. And two jobs ago I was at HP labs, working at the IT part. And so the whole Internet of Things is what people say is the IT-OT convergence. Through my IT background and the OT background enough, I feel I’m in a really good place.
So the opening statement I want to make is we just published a point of view on this industrial Internet of Things from our labs and it just got picked up by the Wall Street Journal and CIO Journal today. So if you go to the Wall Street Journal you can actually see it talked about. The key thing that we talk is, the Internet of Things, people are now talking about sort of the last ten years was about the consumer Internet, where we connected the five billion people on the planet. The next sort of ten years is about connecting 50 billion sensors to 100 billion sensors and people and things connected together through this thing called the Internet of Things. And if you think about it, the Internet of Things has been around for a long time, right? People have talked about connected medicine cabinets, connected cars, etcetera, etcetera. So what’s new? Why now? So there’s three things. One is the cost of sensors has actually gone down tremendously, right? So people used to talk about Internet of Things, sensors, etcetera, but the cost has really gone down in terms of the dollar cost and the energy cost. You could actually put a sensor in there and that thing could run on batteries for sort of six months to a year and you don’t have to do anything. So that cost of sensor coming down has enabled these 50 billion sensors to be possible.
The second thing is that the cost of storage has gone down tremendously. What used to cost hundreds, thousand dollars per gigabyte, hundred dollar per gig, now it’s like free, right? You can store whatever you want to do on Amazon or Azure. So with all these sensors, you collect a lot of data. I don’t know what to do with it and let’s store it and we’ll figure out what to do with it. So the cost of storage going down has had an impact.
On top of that, then you have the ability through communication, sort of wired and wireless communication. So ubiquitous communication, 3G, 4G, Bluetooth NRG, Zigbee, etcetera, etcetera, right? So you have these cheap sensors, lots of communication, and cheap storage that allows you to do a lot of analytics on it, which is now what has enabled this big push. And according to some latest studies, people have said, like GE and Cisco, they say some number between $14 trillion dollars to $19 trillion dollars of economic impact will be enabled by the Industrial Internet. So while everybody talks about Industrial Internet, well, it’s smart, I’ll put a sensor on a thermostat, sensor in a car or whatever, everybody talks about operation efficiency, how by putting sensors in your car, instead of running at 72 percent efficiency, it would run at 78 percent efficiency. So that’s kind of impacting the bottom line. What we say and our point of view is how it unleashes unconventional growth in terms of new services that will be enabled by the Internet of Things, because you’re now collecting data, you own the data, and now what kind of new services will be enabled by this?
So I’ll stop here and hopefully through the dialogue we can talk about all kinds of interesting services that are being enabled in different industries, in healthcare and farming, agriculture, and how product companies are transitioning to service companies. Thank you.
Karabell: Just a quick follow up on that—and for those of you who don’t know, and I hadn’t actually realized when I was looking this up prior to the panel, Accenture has 300,000 people, which is a lot of people. How does this play—I mean when you think about this, it’s one thing to talk about these as macro-trends that are going to be important. How do you actually think about that as a set of challenges or something that a company like Accenture, in terms of helping other companies do whatever they’re going to do, that it becomes part of an implementation of something, rather than just this is happening, isn’t it transformative, and it’s going to create some aggregate amount of future growth.
Banerjee: So the huge opportunity, if you look at what it takes to make an IoT real, there are the sensors, so this is actual hardware at the edge, but then there is some sort of embedded software that has to run on those sensors. But then those sensors will have to communicate that data up to the cloud, so there will be some software that will run on the cloud, software that will run on the sensors, and you have to coordinate the whole thing through some kind of a platform. Again, my colleagues, we talk about some of the platform work. So there’s tremendous work to be done in this area. Some of the challenges that we have identified are, when we have all kind of sensors, some of this sort of stuff is sort of legacy, right? How do you connect your newer systems to older systems? How do you have interoperability? When I was at ABB running these sort of—they have this big power transformer and switchgears and so on. They don’t want their switchgears to talk to Siemen’s, right? Because they want to own their customer. Yet, from an Accenture perspective you’d like to do large deployments where a customer like PG&E would use some Siemens’ transformers with ABB switchgear and so on, right? So there’s the interoperability part.
There’s tremendous opportunity for security, or lack thereof. I mean if you have a connected world with 50 billion sensors, every one of these things is now a point at which it becomes insecure. You can tap into it, you are open to cyberattacks. And then now there is a lot of discussion that’s going around with some data privacy, data locality, and so on, so forth. So imagine that you have a transformer in Germany and you are going to do a service of your transformer through a bunch of engineers in Bangalore. Can you actually transfer that data from Germany if say ABB equipment—it’s a Swiss company’s equipment, located in Germany, being serviced by Indian engineers, right? What’s local, right? So there’s so many challenges around interoperability, around legacy systems, around security, around this data privacy issue. There’s tremendous opportunities for a company like Accenture to help our clients navigate through this maze.
Karabell: And I think Charles probably has some to talk about this as well, but, Kelly, why don’t you talk a little bit about actually creating those connections between multiple and disparate objects and services, and also you talked a little bit about SmartThings and Samsung, which is obviously a lot of connectivity with a lot of disparate parts.
Liang: I will try to touch on all of that, but before I do that let me—if you guys will indulge me for a minute, I want to kind of take a step back and kind of set up the conversation and talk about a couple of little quick things, just really how the Internet more broadly has evolved over the past 15 to 20 years in terms of paradigm shifts. So early days of the web, companies like Google and Yahoo started indexing all the world’s information, bringing all that knowledge online, creating what we of course now call the Knowledge Graph. A few years later companies like Facebook and Foursquare started taking virtual representations of people and mapping those relationships, creating what we now know as the Social Graph. Now fast forward to today, everyday ordinary objects are now getting connected. So companies like SmartThings, where I work today, we’re taking virtual representations of these physical objects and making them programmable, and that is incredibly powerful. And we’re creating what we know the Physical Graph. So what is the Physical Graph? Let’s think about that. When you have in a home lights and switches, doors and locks that all the sudden can trigger events, can take action, or sensors that have on/off, that have capabilities to change states, all the sudden, things that are happening in the physical world can be tracked and monitored in the virtual world, and vice versa, in the virtual world we can make these changes and have it manifest itself in the physical world. That’s what the Physical Graph is.
Now, I actually want to introduce some data, and it’s some data that Prith just mentioned, which is, by 2020 you guys are estimating that 50 billion devices will be connected. Now, our friends at IDC project that by 2020, 212 billion devices will be connected. Now, that is about 30 devices per man, woman, child on the planet. That is a lot of devices. I’m going to say something a little controversial, which is, just because these devices are connected doesn’t mean that they are actually valuable, that they are delivering value. Again, going back to your house, you’ve got the smart light switches, doors. In and of itself, they deliver some value, but the real value, we believe, comes from installing smart apps directly into your home, just like you can install apps on your phone. The value comes from the interaction between these devices that these apps facilitate. And these interactions span the range of use cases from very mainstream ones like security, convenience, lights automatically shut off when you go to sleep, to those that are a little bit more niche, like for example, feed the pets, feed the dog when your kids forget to do so, or you know what, shut off the TV because your kids have consumed too much television and they’ve hit their daily TV allowance. Now, I’ve got two kids at home and for me those use cases are extremely valuable. They’re personalized. They’re relevant to me. So I think for the mass adoption of the IoT, of Internet of Things, to really take off not only do we have to solve for these mainstream use cases like security, peace of mind, convenience, we actually have to address the long tail, because those use cases are important. They have emotional value and real value to people.
Now, SmartThings, we are an open platform. We do quite a few things well, but I’d like to highlight three things that we do to solve this issue. One, we allow all devices that are connectable to hook into our platform. We’re not limited to just a small number of devices that are tied to clothes, to proprietary platforms. Two, we have a platform that’s open for developers and not just very technical firmware engineers, but, you know, software developers, those that can code, to be able to build interesting smart apps to solve these long-tail and mainstream use cases. And three, we’ve got the ability to allow users to install smart apps into their home to make their connected devices now useful.
Now, with the 212 billion devices that are coming online in the next 15 years or so, there’s going to be an inordinate amount of data that’s generated, and that is incredibly powerful. Now, then if we can take that intelligence of that data and create these following use cases, let me give you examples. One, what if there’s a water leak behind your washer and the house detects it, knows it, can actually shut off the water for you, notify you, call a plumber for you, and, oh, by the way, unlock the door because you can’t take the time off of work to be able to do so? What if your alarm wakes you up a little bit earlier today because it knows that there was a huge snowstorm last night and you need to get up a little bit earlier so you can shovel the snow and the sidewalks and scrape your car so you can get to work on time? What if your child inadvertently opens the cabinet in your home that contains all the household chemicals and you are immediately notified and you can go and stop that from happening, or you could stop one of 10,000 trips to the emergency room that happen each year because toddlers ingest harmful chemicals from the house? So the potential in the programmable world is absolutely enormous, and I would say there is no one company today that can solve all the connectivity issues, but it’s going to shape lots and lots different industries, be it healthcare, telecommunications, insurance, utilities. And it extends beyond home automation, extends beyond consumer, but into enterprise, as we were talking about. And what I get most excited about really is the fact that companies from different industries are going to have to come together and collaborate to solve real human problems in innovative and creative ways. And I think that’s why we are all here today to talk about this.
Karabell: I particularly like the turning off the TV when you’ve had the TV limit reached, and being able to do that offsite, because, you know, on the web no one can hear you scream, and to be able to just do that without any real life consequences would be great.
So Charles, maybe from the view from Equinix, and talk about what you wish, but also the issues this raises, not just in terms of the storage of this information and the integration of it, but also the security of it. You know, you don’t want to be able to turn off somebody else’s TV. I mean, you may want to be able to—it’d be great to be able to have this interactivity in a hotel room, you know, the guy’s TV is real loud in the room next to you, you just go online and turn it off.
Meyers: Well I’ll give a little perspective and try to pivot off some of the comments that both Kelly and Prith have made. And I think you started the question with—or the session here with the question of is this either qualitatively and/or quantitatively different in some way in terms of representing a shift in what we’re seeing. And I think the answer to that is both of those are true. And I think that Prith and Kelly’s comments indicate that. And I think both of them will have a much richer perspective on the use cases, some which Kelly just outlined here, that are made possible by this, but I think our perspective, Equinix—I think many of you may be familiar with this, but we would come at this problem and support it more from the infrastructure side of things. We’re very much a platform company. That’s probably an overused term, but given the $7 billion dollars of capital we’ve invested into the ground building data centers probably qualifies in that regard. We own and operate about 100 data centers around the world, comprising about six million plus square feet of data center space in 32 of the most highly-connected metros around the globe. And while we monetize our value in the form of what people would traditionally refer to as colocation services or interconnection, we really think of ourselves in many respects as sort of a curator of digital ecosystems, and this opportunity as it relates to sort of the Internet of Things is a pretty compelling one in terms of being able to sort of build and curate these ecosystems.
And in fact, as Kelly talked about, we’re preparing ourselves for what is inevitably going to be a tsunami of data. And the question which Kelly also posed was, can we effectively—you know, can our downstream digital infrastructure adapt properly to respond to that and make it so that we could somehow translate that tsunami of data into useful and actionable information that creates value and creates sort of high return on investment use cases for people? And so what we’re saying is that the infrastructure is going to have to respond in a way that it sort of parses the problem in some effective way, and a lot of the technologies that are happening today around—you know, the cost of putting a sensor in and having it collect data and store that, and even having the wireless capability to transmit that data and store it somewhere else is one thing, but then the ability to actually have the compute power to process that information, being able to put it from a storage perspective into somewhere that’s viable from a cost perspective over time and then be able to transform that into something useful and put it back out to a user that can gain value from it is a challenging problem and it is one that really needs to leverage the connectivity that is out there on a global basis, and also the technologies that are—really, I think the intersection of cloud computing, broadly stated, with this sort of endpoint technology that makes the sensors cost-effective and possible is really an exciting opportunity and one that we think holds a huge amount of potential.
Karabell: And what about the security parts of it? There was a—and this is a little more what you were talking about, but there was an episode of “Homeland” last year that kind of raised the whole what’s the future of Internet of Things and how that could impact personal stuff. Anyone remember this? They managed to get the code for a high politician’s pacemaker and the way in which they assassinated this politician was by remotely resetting the code of the pacemaker to have a different oscillation and then, you know, he has a heart attack. And that was sort of […]
Karabell: But is there something particular about data security that is in any way different from the data security we struggle with already?
Meyers: Well, I mean I think this just is a different level of the problem in terms of, you know, when you start to extend data collection and transmission out to a number of endpoints that is the kinds of numbers that Prith is talking about here, it magnifies the security for sure. But it’s going to have to be broken down and dealt with in a layered fashion as it has been for a very long time. And so I do think what is happening is that there is—and we see it every day, in fact in terms of people consuming—you know, even cloud computing, which is not all the way out at the sort of ridiculous number of distributed devices, but even with just mobile employees, which those represent a big enough number of endpoints on their own. That security challenge is becoming really difficult to manage for people, and increasingly they’re finding that both from a performance and a privacy and a security perspective, sort of over the top Internet is no longer meeting the needs of most companies relative to the security and privacy challenges that they face. They talked a little bit about Merkel wanting to sort of build a German Internet. While there’s a variety of people who are responding to sort of the challenges from a security perspective in similar ways in terms of saying some level of sort of private-public-hybrid connectivity is going to be required to eventually solve the business problems and the security problems that exist.
Karabell: So I want to ask Prith and Kelly ,and also we do have this mic, but as you want to ask questions just raise your hand and we’ll pass it around. Please feel free to start jumping in as you wish, when you wish. On this kind of qualitative/quantitative question, so a lot of what—separate from just the magnification of devices, but some of the things that you point to, Kelly, and even in the case of an ABB and an efficiency of how you’re managing turbines, some of that where there’s a clear business case for efficiencies—I mean you can understand why Google pays $3 billion dollars for Nest purely from a, if it’s going to be a data hub and an efficiency and all of that stuff that people are going to want and desire and it makes sense. I still want to push on the question of, is that so much different in the way in which our lives are actually being lived, or is it just a different tool than insulating your home or putting a hand lock on the kitchen cabinet that had the ammonia, or any number of things that people have always done, this just being the latest technical version of it? Do you actually see it changing things in other ways?
Banerjee: So let me address this issue, kind of what I highlighted in our sort of recent point of view. So a lot of the IoT play has been around operational efficiency, and so we have this joint venture between Accenture and GE called Taleris, where we monitor aircraft engines. There’s sort of all kinds of stuff on engines, and as that plane is flying from New York to San Francisco, you collect all kinds of data, analyze it, and you strive—So basically, an aircraft engine has to be sort of maintained every 30,000 hours, just like you rotate your car tire every 5,000 miles. So by monitoring the aircraft engines sort of performance you can say, hey, you could postpone that preventive maintenance, instead of 30,000 hours, to 40,000 hours, and thereby increase your efficiency of that plane, right? You don’t have to do this thing. Or you just found out this thing is broken, send this spare part to the gate with the repairperson, and you reduce the downtime of that aircraft and increase the efficiency. And those kinds of examples are—there are lots and lots of examples of operation efficiency. But the real play that we make is about these new services, as we call, unleashing unconventional growth. Let me give you an example. All of you know of John Deere as a farm equipment, tractor company, these green tractors that are driving around, right? So what they are putting—and again, we are working with John Deere on this kind of solutions—is to put sensors on all their farm equipment, on their tires, so as that tractor is going on the soil, that is sensing the soil, and if that particular region requires more pesticide, more carbon, less nitrogen, whatever, you actually know exactly where the tractor has gone with a GPS location, so you could do much better farming and thereby you drive what is called higher business outcomes. It’s the outcome economy that we are going after, right? So instead of John Deere being a tractor company, they are becoming your precision farming as a service company. The same thing happens in Michelin, sort of a tire company going to tires as service and so on.
So there is just so much stuff that is being unleashed in terms of the data that you collect. You can convert that data into a service in terms of higher business outcomes. So I could just literally go into lots and lots of examples in this area, and Kelly talked about healthcare examples and the Fitbits that you have, right? I mean, that’s just collecting your blood pressure information, but just imagine tying that with your health records and so on, and these are these whole new things that are being enabled by this technology. Zachary mentioned, so what’s the kind of stuff that you have to do in terms of the innovative software world that is needed. I talked about the software running on the edge devices, software in the cloud, but if you think about the new things that is needed, if these 50 billion to 200 billion devices all collect data, there’s just so much data you’d send, you’d crash entire systems because they would not be able to process that much data. So what you have to do is figure out some of the computations you would do at the sensor. This is called sensor-based computing. Cisco calls it fog computing, others call it different, but so there’s a lot of interesting work around doing stuff at the edge, edge analytics, sensor-based computing, fog computing, edge computing. Then you do sort of this analytics on the cloud. You do the sort of analytics on this thing, and then ultimately you have to build applications, as Kelly was talking about, these smart water, smart whatever, right? You have all these millions of apps on your Android or iPhone. There are apps that you could build, you should build for each of those equipment, your smart thermostats or whatever, and you have to do that in a secure manner. Tremendous opportunities, lots of interesting innovations, but driving much better business outcomes. It’s transforming companies from product companies to service companies, that’s the opportunity.
Liang: I was just going to add to what Prith was saying in terms of new services being created as a result, and I really see this as growing the pie versus it really being a zero sum game. So take for example your HVAC system in the home. Imagine if before it actually dies—in the middle of the summer were your air conditioning to die, that’s a horrible scenario. But before it happens, because we’re collecting all of this data, your service provider is actually notified and they are coming to you to fix a problem before it happens. Now, this is a service that doesn’t exist today. It can’t exist today because we haven’t made that data available. But now we’re creating new use cases, new services as a result of this fad, or trend, or whatever you’re calling it. I don’t think that this is just a trend. This is the world we’re going to be living in. This is intelligent living, no longer just connected living, but intelligent living, and the predictive models and behaviors that result because of all of this enormous amount of data will help us live more intelligently, and it creates a whole new era of how we interact with things and people around us. So just kind of adding on to what Prith was saying and giving some more examples of new services that are created.
Karabell: Yeah, I wasn’t going to […] I was just being the moderator and trying to get you to talk about […]
Meyers: I would expect that the—I think that there are probably use cases that we don’t yet even envision. It may represent meaningful shifts in quality of life or adjustments that we don’t even fully understand, but I think that in many respects what both Kelly and Prith here are talking about—you know, Geoffrey Moore talks about this concept of trapped value, and I think there are so many sources of trapped value in terms of feedback loops, that either today are very long or that don’t exist at all, that this will enable that I think will unlock massive trapped value out there in the broad enterprise that I think will then be able to fuel other sources of innovation, and I think quantitatively it will be very significant.
Karabell: Please just identify yourself.
Bonchek: Oh great, thank you. I’m Mark Bonchek with SHIFT Academy. I’d love to get the panel’s thoughts on an idea I’ve had that, I don’t know if it’s shaped too much by my bias or whether there may be something there. And it comes from looking at trying to understand what this is going to mean, and all I keep seeing is people talking about data, but if I look at the Internet evolution as a model for this Internet of Things evolution, I kind of take the 1.0 is when we got computers connected together, and then 2.0 is when we got people connected together, and the social revolution is when it really started to pick up. And so if I look at that here it feels like we’re talking about the Internet of Things in kind of a 1.0 way. It’s still about data and machines and the analytics and so on. And when I take the kind of stack that you laid out of data, computing, analytics, and apps, where are the relationships in that? I mean, if devices and things are going to start to become intelligent then they’re going to have relationships with each other. Take Nest, right? The thermostat and the smoke detector are talking to each other. How are they talking to each other? In what way? In what language? Is there trust? Is there reputation? These things that start to sound like kind of anthropomorphized views are in some ways not. What’s the anthropology of the Internet of Things? So I’m wondering, I look at things through a community lens, but is this a social network of things and how do we think about this social network of things?
Banerjee: I think what you said is absolutely right. In fact, people have done analyses of sort of the Industrial Revolution that happened sort of in the 1800s, water, electricity, manufacturing, etcetera. Then there’s sort of the IT revolution, the Internet revolution, and now in fact there’s a term called the Industrial Internet, which is sort of a combination of the Industrial Revolution and the Internet revolution. We have defined sort of a term called—again, we have leveraged some of the earlier work that has been done by GE, but we have defined that Industrial Internet as machines, processes, and humans all connected together through Internet. So the first generation Internet, 1.0, what you call it, right, was designed to connect sort of people and electronic commerce. What I said was the consumer Internet, that impacted only 20 percent of the economic value of the world. If we look at a GDP of the world, that consumer Internet only impacted 20 percent. The remaining 80 percent is, by connecting those machines, like Kelly was talking about, that is the 80 percent impact. Huge. The kinds of connectivity requirements would be quite different. You and I, when we talk, we send you an email, you understand that and so on. But how does a machine, an ABB machine talk to a Siemen’s machine using legacy protocols? Those things have not been figured out. How does a machine generating voltage and currents talk to a human? How does the human understand certain ways of—so this is a huge opportunity, so your 1.0 Internet will actually not solve it. This requires the 3.0.
Liang: So I want to jump in. I don’t have the answer, but I will pick up on your point on going back through anthropology and history and look at, in the early 1990s, when email started proliferating, and there were services like AOL and Prodigy and everybody saying, “Come sign up for my service because you can email the 50 million people in my network.” This was before the emergence of SMTP, a standard, an open standard that everybody could embrace, and what it did was level the playing field for all the players and let companies compete based on UX, user experience, functionality, features. And again, I don’t have the answer, but I know that that’s the world that we need to get to, where there are standards embraced by the industrial Internet, companies, people, machines, devices, sensors, etcetera. So just taking a step back and looking at history, how do we take what we know and has worked and how the Internet has evolved and think about it in today’s terms.
Meyers: Yeah, I think if you look at—and some of this has been discussed over the last couple of days—innovation and how it is evolving, the effect of open source, all these things, a lot of the dynamics of the web world and Internet X.Y, in terms of however many variations we’ve gone through, but I think that it is going to be—this notion of ecosystems and bringing together folks who can innovate in sort of, even if they’re not geographically in proximity to one another per se, that the digital infrastructure can be crafted in a way that allows them to innovate and collaborate effectively, I think that magnifies the potential of this opportunity—however it plays out, because I certainly don’t have any of those answers either. But I do think that this notion of sort of open, collaborative, innovative ecosystems is fundamental to it. But then they also have to be incredibly scalable, incredibly secure, completely global, because that’s the other thing, I mean it’s been pretty cool watching the panels and some of these things in terms of where innovation is coming from, and the notion that we had with the panel this morning of empowering people in their local communities rather than third parties coming in to try to save the day, so to speak, to really foster innovation in a local way, those are incredibly powerful things, but the digital infrastructure is going to have to keep up with that.
Anderson: Mark Anderson. I agree with everything you guys have said so far. I’d like to push it a little bit. Often in these IoT conversations what I find missing is the actual mechanism: here are the problems, here are the opportunities. So I want to give you a couple examples of what I’d like to hear you respond to. I designed a chip, which I think was the first ever done, called a PRP, a pattern recognition processor, a couple of years ago, and I did that because it seemed to me that the old Von Neumann chips and computing technologies that we run today are completely outdated. It can’t be linear anymore with all these things we’re talking about. It’s going to be all about billions of things and very large scale integration of pattern recognition on those things. That’s how we’ll get these solution sets. It won’t be with Von Neumann. And so we need to have a whole new way of running things, of having chips at the core of these operations which could do that, and which by their nature saw patterns and conveyed that information in various ways. And there has been a new chip just announced by IBM under Dharmendra Modha, through the SyNAPSE project at DARPA called TrueNorth, and if you haven’t looked at it I would recommend that you—yeah, you probably know Dharmendra, right? So there’s a colloquium in two days. If you guys are around I’m going to be there. But this is an amazing chip. It’s the largest chip ever made. Samsung manufactured it, and it combines PRPs, pattern recognition processors, with brain-inspired compute hardware at very low power, record breaking low power and record breaking number of transistors on a—all that stuff at once, but made to work both to see patterns and to figure out what to do with them. I think that’s the kind of technology we’re going to need to get beyond the “oh my gosh” conversation about how big it’s going to be and start having this conversation that you’re alluding to about how do we get solutions and relationships and everything else out of all of this.
Banerjee: So I am—because I run R&D, it’s my job to kind of follow all these things. I’m very familiar with the IBM work, but those kind of technologies have been around by many—other companies have been working on neural network kind of things as well. But that gets to the core of what I had said earlier. If you have 20 billion, 50 billion, 200 billion sensors collecting data and if you say I’m just going to collect it and shove it up to a grand computer that’s going to crank it all up, that computer will die, right? So you have to allow—and this is why this concept of edge computing, sensor-based computing, edge analytics have come in. Just look at a very simple example, when you send a spacecraft to Mars and you are doing this remote mission, like India just sent a spacecraft, right? Their round trip delay in light signal going to Mars and back is 20 minutes. Imagine controlling that spacecraft from Earth and making left-right turns. You couldn’t do it. You have to have computing on that spacecraft to do the local analytics at the edge and do the global things from Earth.
Anderson: So […] PRP on the spacecraft.
Banerjee: Absolutely. So what you’re talking about, Mark, is actually the concept of system of systems. You have to build system of systems, right? Those are little, little systems, right? And so what Dharmendra has done—I mean those are these tiny, very interesting cognitive synaptic-based stuff, but that by itself will not work. You have to connect those systems to other systems in a global network. So basically this whole area of edge analytics with some computing in the cloud, even those things will require cognitive things, but at a higher level. Like we should probably take this offline. Sorry.
Meyers: The work that was done earlier, which you say failed, in neural networks, back 20 years ago, some of it may survive. And one of the things that was most troubling to me was that by the design of those neural nets, they would give you the right answer—you’d do a compute and they’d give you the right answer. They couldn’t tell you how they got to that answer. And I think we’re going to come to a time where see patterns through these setups, these systems we’re describing, and we’ll have to have trust relationship so that we know, okay, you’re giving me the answer, steer left, or do this, or here’s the truth, but I can’t tell you why it’s the truth, right? We won’t know why it’s the truth, and we’ll have to have a pretty interesting conversation with our systems about trust, because it may never be obvious us. Our science may not why is that true yet. We’ll be learning in a way that we’ve never learned before. And so it’ll be very, very difficult for human beings I think to work with these systems.
Karabell: I mean a different aspect of trust which occurs in all this, particularly when it has the home implications, not necessarily the corporate ones, is how do we—we can hardly grapple with what we consider privacy that we allow to be invaded because it’s convenient versus privacy that we want to protect because we feel threatened. And privacy is an odd, it’s kind of a synecdoche of people’s concern about will they have control, can their information be used against them. You know, we’re all very comfortable revealing an immense amount of information if we feel it makes our lives more fluid and convenient, whether that’s us as consumers or us as citizens. It may have a little less implications when it’s GE and Boeing trying to figure out how much fuel or maintenance they need for a craft, although even there there’s issues of control and information. I mean, have any of you thought about what the—are we even really beginning—given how unprepared we are for our current level of information being used or not used by those who we want to use or not use it, are we in any way, do we have any system of preparation for what some of what all of you have been talking about would lead to?
Banerjee: I can take a crack at—
Karabell: Kelly, I’m interested in your take.
Banerjee: Okay, sorry.
Karabell: I mean both of you, just in the sense of what do you do with that?
Liang: Go ahead.
Banerjee: So who owns the data is going to be a very, very fundamental problem as we on, right? So you just go and buy a Nest thermostat for your home. It looks so awesome, it’s allowing you to remotely control your temperature, it’s “Hey, turn on the temperature, increase the heat when I come home from work, and I come home from work at 9:00.” Now the fact that you have just sent this signal from your iPhone to turn on your thing, somebody, in this case Google, knows about your work patterns. So just because it says Nest thermostat, is it Nest’s data or is it your data, Mark? So these fundamental questions are going to be asked, so those of you who are interested in this big data issue, there’s a professor at MIT, Sandy Pentland, who is kind of talking about a word called “the new deal,” the new deal for data. Very, very—and we actually are working with—so Mike Sutcliff, who is my colleague from Accenture, he runs Accenture Digital. We are funding some of the work at MIT. So these kind of issues about data privacy, about who owns the data, right? It’s an ABB transformer, but the transformer is sitting in my site, at PG&E, right? Can ABB just do all kinds of things? No. At some point, these kind of questions about Chancellor Merkel saying, “Hey, I want my own Internet,” these are issues of Internet governance that is going to come in and deal with it. Who owns the data? So those of you interested should read a very interesting article by Sandy Pentland from MIT.
Liang: I guess my short answer is we always take the perspective that the customer, the consumer owns their data. They can give permission of that data to their service providers, but at the end of the day you can control what you give. And it may be limited. It can be limited by duration, limited by individuals. And so I don’t see that tension, quite frankly, of who ultimately owns that data, because the consumer will ultimately own that data. Now, the data may flow through multiple networks, and so there needs to be connectivity and dialogue to manage who gets access to the data, but at the end of the day it’s got to be the consumer that owns that data and determines who has access to it.
Banerjee: Absolutely. So the whole concept of new deal for data is this. The consumer owns the data and I give permission to Nest or Google or whatever. So if the permission is an anonymized thing, if Google can say on the average, among the 100 million people in the US, 90% people come home at 6:00, 20 percent come at 9:00, that’s fine. That’s anonymized information. If it says Prith came home at 9:54 p.m., that’s private data. I don’t want that to be revealed to Uber to send a car—no, seriously, you could potentially, all the cases that she is talking about can be enabled. If from my iPhone I enable this, an Uber car can just land out there. I can have airlines too. So that part of who owns the data is a really, really interesting thing that will come about in this data and data platforms, and we think at Accenture, where we guide our clients, there’s going to be very interesting platforms that will evolve. I mean, I think you talked about platforms, where it’s very industry specific platforms and those will be the control points and who would own the data, and essentially consumers will own their data, but they’ll give permissions on an anonymized or individual way for people to monitor the data.
Karabell: Obviously, though, it gets increasingly complicated whether you have opt-in or opt-out systems, and then of course the consequences of not only allowing that data to be used—like, sure, you can opt-out, “I don’t want anyone to use my data,” but then will you then be able to use the whole series of services that everyone else is using, the non-use of them seeming to be a disadvantage. It gets unfortunately less simple than purely “I want my information used this way.” But let’s go here.
Tucker: Lew Tucker, Cisco Systems, and I ran all like Sun’s Java, Sun.com, Sun.com, large websites. You all know that in fact it’s a fallacy that the user owns the data because of request to remove all of that data is not possible. You can never remove it from all of your backups and everything else like that. You might want protect it. But that’s not what I—I wanted to get to one of the questions around are we looking in a world, in your view, that we’re going to have smart sensors, truly distributed intelligence out there, we’re talking to services, Prith, as you sort of said, which means—or dumb sensors that are just sensing data and streaming it back or whatever? Because there’s very different models, computational models, that would be employed depending upon which way we went. I mean, if we knew from back in the mainframe and client server days that we would have these kind of intelligence here, we wouldn’t have built a whole lot of huge centralized databases and large ERP systems and everything else. It would have been designed very differently. So in your view, are we going in a direction that is going to have—you know, Berkeley got Smart Dust. We can have a lot of intelligence very much embedded in the devices, and the dumb ones will simply have proxies located nearby so that the centralized cloud systems aren’t going to be absorbing enormous amounts of data.
Karabell: Actually, before you guys get to—Charles, from Equinix perspective, when you think about your three to five-year plan of what you’re going to need to build, what assumptions are you offering on that question?
Meyers: Well, I’m not sure if those would necessarily hit on that, because I think that the answer is likely to be both, and it will be application dependent as to which makes more sense. I do think though that, in an environment where it is not as—you know, it is dumb sensors, if you will, that collect information, transmit it back, I think that it all depends on what the bandwidth environment is, what the compute environment is, and how it can support it. My sense is, personally, there will be more of that, and I think that our assumptions about how our infrastructure evolves I think has to accommodate both scenarios but begins to put people into an effective proximity with one another where they can share information more effectively.
Nwokorie: Ije Nwokorie from Wolff Olins. My question is probably different. It’s around the social benefit of this. So if you think about the first, the sort of the knowledge phase where the benefit was this massive democratization of knowledge, that sort of everybody, a six-year-old girl in Kenya has access to the same information that I have, and maybe the social graph stage was around this incredible ability for us all to self-organize, and we can see that happening in the Arab Spring and stuff like that. What is it in this phase that is not just the sort of the rich world solution to middleclass problems of the home and intelligent living? What’s the big social benefit of this stuff?
Banerjee: Sorry I just jump in. So there’s a World Economic Forum project that we are leading—I’ll kind of talk about it tomorrow in the Internet Governance session. The World Economic Forum actually has—we have proposed to the WEF that we will look at the social, business, and economic impacts of the Industrial Internet, and for that project we are actually addressing this size issue. We brought together a bunch of companies who kind of advised us. So the simple thing is that there is a lot of waste that happens in the world. Yes, there is hunger in the world, but there’s a ton of food waste that is going on, and if you could have real smart sensors and so on, so forth, you could actually prevent that food waste. So some of the things that we have talked about is what kind of IoT technology, what can the developed world do in terms of through these sensors help reduce waste and sort of improve the operational efficiency in the developing world? Things like smart water, smart—so I can talk to you offline, but tremendous social and economic benefits will be unleashed by the Industrial Internet.
Liang: There are examples in the utilities world where now if you allowed your utility company to be able to control when to turn on your washer, dryer, dishwasher, when the grid has excess capacity, and you just allow them to make that decision for you, and so the utility company can say, “You know what, at peak times I’m not going to run that dishwasher.” But then the user benefits in that there’s reduced cost to them, and of course it benefits the whole ecosystem, with utilities being able to manage outages and how power is consumed. There are obviously mass economic and social green living benefits. There are cities, for example, there’s a city in Spain that they’re now—what they’re doing is they’ve built sensors all around the city that let city workers know when there’s waste that needs to be picked up, so that cities can design systems, logistics, supply chain, transportation systems that are based on the intelligence of where there’s waste, where there’s people, where there’s excess parking, where there isn’t, and make that city run smoother and faster. Now, all of that stuff needs to be quantified, and firms like Accenture will go out and do that analysis, but I think there are clear social benefits that will emerge over time as neighborhoods get connected, cities get connected, countries get connected.
Meyers: Well, as you pass the mic, I guess one other comment I would make is, as was talking about yesterday in a panel, you know, in the end you can’t expect the technology to achieve the social benefits on its own. If the participants don’t sort of strive for the social benefits, it won’t be realized. But I do think that the last couple days is a good indication that there’s a heightening of awareness around the problems of inequality, of various other problems that have enormous social impacts and I think impacts around the long-term stability of the world in which we live. And I think things like this represent opportunities for us to have levers at our disposal once that awareness reaches a flashpoint that can translate into massive benefits, but it will take a will first.
Alsbury: Seth Alsbury with Target, part of the new enterprise growth initiatives team there. And one of the areas we’re looking at is Internet of Things, of course. Since Target is I believe positioned well to help bring these smart devices to the masses, along with education, context on how they work, possibly some integration, and sign up for some of these services, wondering what the next steps are as an industry to crack some of these things like standards, like the data management, etcetera? Like who’s driving that currently? What does that group need to look like to get this stuff sorted out as an industry?
Liang: I would just maybe answer that question slightly differently, because I think as we have more consumer awareness and demand, naturally these issues, these friction points between—largely some commercial issues primarily, because technically I’m confident, I think we can solve it. Commercial issues will go away and will get resolved because consumers are going to demand a very seamless experience. They’re not going to care that they can only hook up a few devices at home. They want to be able to go buy hundreds of devices at Target and know that when they take it home, it just works. So I think as an industry, rather than starting at it and looking at it from a technology standpoint, how do we solve the technology interoperability questions, I actually think that the bigger challenge right now ahead of us is actually solving for consumer awareness and demand for the technology.
Banerjee: There is a technical answer too, in addition to those things. So we talked about standards. There’s three kinds of standard things that you should be aware of, if you don’t know that. At the lowest level, for the consumer type of things, Qualcomm is leading an effort called the ALLSYN Alliance, A-L-L-S-Y-N, and there’s companies like Samsung and others, and LG and all these sort of refrigerator companies connected to IoT, they are part of the ALLSYN Alliance. Broadcom and Intel, they are supporting a team called the Open Internet Consortium called OIC. And then there is this bigger thing called the Industrial Internet Consortium, which was started by AT&T, Cisco, GE, IBM, etcetera, and Accenture’s a member there. That is about 90 companies that are all driving standards and interoperability and test beds in this area. So IIC, OIC, and ALLSYN are three things that you should be familiar with.
Karabell: We are being given the time to go, time to go. But thank you so much for participating. This was eye opening and only the beginning of a series of discussions that I’m sure we’re going to be having in much greater depth and much greater complexity year by year by year. So thanks very much.