When Digital, Physical Worlds Merge

Semiconductor Engineering sat down with Simon Segars, ARM’s CEO, and , managing partner of Lanza techVentures, to talk about changes in the IoT, self-driving vehicles, cloud-based health monitoring, and the impact of machine learning. What follows are excerpts of this conversation.

SE: Several years ago the IoT was just a vision. It’s now evolving into something that is real. What’s changed from your standpoint?

Segars: There’s a lot of experimentation going on. People are starting to experiment with different devices and different ways of sensing the world around them, and starting to get creative about how the data can be used to provide benefits, whether they are consumers or the industry. We’re still in the very early stage of this technology, there’s a long way to go, and a lot more benefit that can be pulled from IoT.

Lanza: The things that IoT refers to have been here for a long time. We just did not have a way to classify them in a nice community that IoT, as a term, does. The importance and impact of IoT on society is just the beginning. It’s absolutely nothing compared to what it’s going to be in 5 years, and that’s completely different from what it’s going to be in 20 years.

SE: Are you getting glimpses now of what this will look like, taking it out several years in a variety of different industries—medical, industrial, consumer, automotive?

Lanza: When we started years ago, the Internet was the Internet of computers. We were dreaming that we could connect millions of computers, via cables, at that time. Then, suddenly, the whole thing became the Internet of people. We connected billions of people. Now we are talking about the Internet of things. We now can connect trillions of things. But the entire community of trillion of things doesn’t have rules or behaviors that it would accept or not accept. There’s an incredible set of things that we’d love to change in order for all of these things to be connected in an efficient and positive way. I can think of positive and very negative ways to connect things. That’s what we don’t see yet. It’s not going to be 20 years from now. It will be a few years from now that we’re going to start to see all the phenomena that we just did not expect from these connections.

Segars: That’s a key point. There are unexpected surprises that come from connecting all these devices together. In the early days of the Internet, it was about connecting computers and about sharing academic research or publicizing information about your business. In the early days of the Internet, nobody anticipated the giant social media or ecommerce companies that leveraged these technologies. IoT is very similar. Right now, there are cameras everywhere, sensors on things, starting to take the data and access that data, then we’ll start seeing more innovation. To me that will be one of the most fascinating things to watch.

SE: The 800-pound gorilla in the room is security. But instead of one gorilla, with the IoT there are a bunch of them all networked together. How do we solve this?

Lanza: The only way we’ll be able to solve it is to isolate the challenges by segments. There will be challenges in the medical segment, in the automotive segment, and challenges in all these different segments. Experts would be able to understand how attacks might happen and what are the sensitive things to protect. Every single segment is going to have a different way to deal with this.

Segars: I don’t think you can guarantee security in the short term. It’s more of a case of managing security in the short term. One of the challenges is that every time you connect something to the Internet, you risk creating another hole, another entry point for somebody with nefarious intentions that can gain access to the network and access data. The data that we are talking about here is very very personal, so it’s going to be something that will be taken very seriously. It’s something we need to come to grips with as we roll out all these connected devices.

SE: ARM has been very active in security for years. Do you feel like you made a dent?

Segars: What we’re really trying to do is create building blocks to enable IoT to be managed—to manage the security perimeter. We have made progress on how to keep secrets away from non-secure parts of the design. We’re trying to create something in the structure so that security can become a managed problem and one that is local.

SE: Lucio, from your standpoint, as you invest in companies, is anybody willing to pay for this? Is there an upside financially to security?

Lanza: It depends on the risk they perceive. Normally, that depends on how much they’ve been hurt yesterday. Eventually people will end up understanding they do have to do something. It’s not necessary that they personally are attacked, but that someone similar to them is attacked. We’ll see this very, very soon. Damage already has been done to several corporations, much of which has not been publicized, by just attacking security from the phone. This is something that’s obviously a big issue. Corporations understand that. Do they publicize it? No. They publicize positives, not negatives. The beauty, from a technical point of view, is this is a problem that you don’t just solve. You solve it, they attack it. You solve it again, they attack again, and then you’ll solve it again. So it will be something that requires a lot of attention, intelligence and intellectual capacity.

SE: Some of these designs, though, are expected to last 10 to 15 years. Is there enough that we can build into these designs that we can be sure that they either can be updated or secure for that long?

Segars: You build some flexibility into the design and you can do that by having separate updates and building flexibility into the hardware itself. You have to take an approach that in the future the nature of the attack is going to change. The nature of the response has to be able to be changed through the device’s lifetime. One thing that is really important is that you take the human element out of managing security. That especially applies to anything consumer-related, and that alone improves security enormously. Fundamentally, you think about the design of the device changing through its lifetime to respond to an ever-changing security threat.

Lanza: The interesting thing, and ARM knows this very well, is that when they implemented the first security on the chip, I was amazed that they were doing it. This was 1999 or so. Bringing security on the device is obviously that something is going to happen. To your question on the point of whether people are really interested and going to pay for this or not, in the end we will install this in the IP. They will receive it whether they understand it or not. The other side of the equation that’s going to be interesting in security is where does the information from the device go? Now you have the cloud and it is public information.

SE: What happens when we have trillions of things, not billions, and those trillions of things are not just connected to people like they are today. Instead, they will be connected to other things and sharing information across many different places, some of which you don’t know about. Who wins and who loses?

Lanza: I look at the human being and when I project the human being X years from now, when all the information about us happens to be in the cloud. It’s not just the information that I send to the cloud. It’s also information about my heartbeat, my sweat, my activities yesterday and two years ago, my emotions of when I was in a meeting, and information about when I move my hands. Everything is up there in the cloud. All of these things are very difficult to protect. You yourself will have this cloud, and that will be your cloud. And people will be extracting information when they need and want it. This image is scary. I can see how we can protect the device, but I can’t see how we are going to be able to protect your ‘cloud-self.’

Segars: There’s a lot of data being gathered about us constantly over time. We’re on the brink of really having to face up to that because the IoT extrapolates that enormously. That’s why we have regulation coming in, where there are rules about how companies can manage data and explain what they are using the data for. Ultimately, everybody gathering data will have to take responsibility for it. There’s just too much to lose. There are enormous benefits to the IoT, but there are some really scary downsides as well. As a community, we have to take it very seriously.

SE: So we could all disappear if there is a power outage or computer glitch?

Segars: There are backups in lots of places, so we’re probably good.

SE: When you look at all these IoT devices, how do you determine which one is better than another one? What should we look for in these devices? Is it coming down to an Amazon or Yelp review, or something else?

Lanza: That depends a lot from field to field. There are a few fields on everyone’s thoughts these days. Everybody tells me how three years from my car will be self-driving. The point is that for me to get to the point where I trust the car to drive is going to take a long time. The first thing I’m going to think is, ‘Well, who else knows how to brake my car?’ ‘What about stopping it from braking?’ These are the ones that will stop people from using it. Even if there are 100 people who really have a very good experience with self-driving cars, it only takes one person being hacked and it gets a bad mark. On the medical side, it’s even more challenging and interesting. The car is nothing compared to the complexity of the human body and brain. If you look at it from the human or medical point of view, what do you do so you don’t have to get sick and go to the hospital? From that point of view, with everybody being so different, the amount of information we have is unbelievable but the benefit is immediate.

SE: Simon, has medical taken off from your side? We’ve seen lots of talk about it coming next year, but that was 20 years ago.

Segars: We’ve come a long way. You can already upload data via your smartphone to your doctor. That is helping people make fewer visits to the hospitals already. We have a huge way to go, and there are potentially huge benefits to consumers. Health care is going to become a more acute issue as society ages, and with the cost of health care rising.

SE: What role do you think that machine learning and natural language will play in the IoT?

Lanza: Twenty years from now, the knowledge will be completely different. When I started working with computers, I remember the first company I was working in had an accounting department with 100 people. Did I ever think computers would change that? No, I didn’t. If I look at the medical side, it’s extremely interesting. What you are really doing is getting people to get information on themselves, and that’s where the potent element is—the acquisition of the information. Then, the really exciting stuff comes from the machine learning side. You’re going to have enormous amounts of information on yourself—so much you don’t even know what to do with it. You don’t know whether you should act on it or not. What you’re going to have is enough information that you can in effect get your data up in the cloud and you’ll have machine learning devices, this incredibly smart infrastructure, to be able to say, ‘Oh, that’s what happens with your sweat. There’s a 20% probability this is what the problem is. You should check that. And by the way, the fact that you have this change in your heartbeat is nothing to be concerned about.’ It will be an enormous amount of information that is based on more data than any doctor can ever learn in their lifetime. That is when the matching happens—IoT on one side gets more and more data, tied to the smartness of the machine-learning cloud. It will help you utilize this data and move you away from the accounting department. You won’t need an accounting department—computers will do enough.

Segars: Machine learning is the whole point of IoT. Without machine learning, the data is not very interesting. The insight of what we can get from the data is where the value is. It’s so much more data being generated that any human can sift through anyway. Back to the medical example, there’s more medical research published every day than anyone can ever read. The data generation is already outstripping the ability of humans to consume it. That is only going to get worse. You’re going to see machine learning algorithms running on the edge on the devices, working out what is useful versus useless data, sending it up to the cloud. There will be algorithms that lower the security risk for your data being broadcasted. You’ll see machine learning in the device, in the network, and up in the cloud.

SE: What becomes the most valuable commodity in the future—is it data? You certainly see that with companies like Google, Facebook, Amazon, and even IBM, trying to control the data.

Lanza: Data is a very valuable commodity, but the most valuable is the machine learning. That is the real big value. Machine learning is almost like suddenly as a human being you grow up and have more and more data and then the brain grows enough that you’re capable of managing all the data around yourself, as a human being of yesterday. You’re going to have an unbelievable amount of data. Now, the important thing is how do find out the data that’s important, how do you connect it and learn from the past, and predict the future. That’s all something that will happen with machine learning. The enormous contribution will come from human beings in making machines really learn, and in the end behave and work on the right things.

Segars: Everybody talks about being able to use the data, and the data being like oil was to the industrial revolution. It’s the raw material from which valuable information is created.

SE: One of the predictions of early Internet of Things age, companies would be born and die faster than ever before. In this new world of things, do you predict that will be same or will existing companies just get stronger?

Lanza: It depends on how you define a company. The reason people think that they won’t survive is because they don’t see all of the pieces. If you think about just the IoT, there are hundreds of millions of physical features and dimensions and physical phenomena that you’re trying to capture and make digital personalities. That won’t happen at XYZ Company, which is located behind my house. New companies will use tens or hundreds of thousands of designers somewhere in the world. They are going to be there designing something for a while as a part of the company. What is really happening is that the physical proximity will no longer be relevant. So the company that says everybody should go to work at 8 a.m. and leave at 5 p.m. is going to be the past. The future of a company is everywhere, and it will change over time. So you need to redefine what a company is.

Related Stories
Machine Learning Meets IC Design
There are multiple layers in which machine learning can help with the creation of semiconductors, but getting there is not as simple as for other application areas.
Safety Plus Security: Solutions And Methodologies
Part 2: Connecting to the Internet adds new demands for safety-critical markets.
CEO Outlook: Chip Design 2017
Political uncertainty, tempered optimism, continued consolidation, and concerns about capacity.