Skip to main content

The evolving network in the age of AI

Real Conversations podcast | S5 E22 | November 22, 2023


Nishant Batra is the Chief Strategy and Technology Officer (CSTO) with responsibility for corporate strategy, technology architecture and pioneering research at Nokia Bell Labs; Nokia’s information technology (IT) infrastructure and digitalization initiatives; centralized security domains; and Nokia’s venture capital activities.

AI is the top tech trend right now and Nokia Bell Labs predicted. Nishant Batra its Chief Strategy and Technology Officer (CSTO) discusses what this means, other big trends, and the the all-important role of the network.

Below is a transcript of this podcast. Some parts have been edited for clarity. 


Michael Hainsworth: AI will soon be embedded into everything we do, no matter what we do or where we are. But for the future of artificial intelligence to be realized, AI researchers are realizing a network that is high speed, with low latency, connected to a cloud continuum is critical. Nokia’s Chief Strategy and Technology Officer Nishant Batra spent last year building a plan to be that network, and he spent 2023 making that plan a reality.

Nishant Batra: The trends were still in a form of emergence. This year, we see that certain technology vectors are rather clear. We already claim that last year, artificial intelligence would be a big form of influence with respect to how the technology landscape evolves. But now, I mean, I don't think there's any doubt left to the fact that AI will be a massive influence on how this world evolves, not just the technology or the technology landscape, but probably every industry, every enterprise, every consumer is touched by AI in some format or other. So that change is rather apparent, and we're a little bit proud of having foreseen that coming and, partly because in Bell Labs, we've been trying to research AI for a good part of the last two decades, so there was some vision coming out of that.

You know, could we have predicted the transformer models would be so dominant? No, but we were playing with transformer models, so we saw that there's value to NLP. The other thing that has been nicely clear, and now we have changed the terminology of it, is cloud adoption. Cloud was critical to the whole evolution of how connectivity and information come together. Cloud is a very integral part of that and the whole phenomenon around cloud, it started with Enterprise IT. An Enterprise CIO would look at, okay, I've got all these data centers. What's the best format? Should I continue to own this infrastructure? Should I move to a hybrid with public cloud? I think that the whole phenomenon has matured. What we now call it is a cloud continuum, and this continuum spans across two axes. One axis is simply the divergence between public, private, and hybrid, and that's the choice every CIO makes even today. How much of infrastructure based on application sensitivity do they want to keep on-prem, and how much they want to keep on public cloud? I think there's a lot more trust in public cloud today than there used to be. Most are headed towards some kind of hybrid.

The more important continuum that I see, and that's what we have reflected in our technology strategy 2030 as well, is that there is a continuum from central cloud to regional cloud to metro cloud to edge cloud to far edge cloud to nano edge cloud to on-device cloud. There are, essentially, these three laws that come into play. There's the law of physics. An application owner wants to get access to information fast. That means latency is critical. There's a law of the land that means I don't want my data to transcend certain boundaries so that information is kept closer to the user than actually at a central cloud. And then there is the law of economics. The more that you have in the central cloud, the more it takes effort to fetch that information versus closer.

So those three come into play, and that's what has evolved in our thinking from last year to this year. And I know I'm taking a little longer on this answer because I want to build this upright. That whole aspect of cloud compute AI, it needs to be tied together with a fabric of connectivity. You and me are talking today; this call is going to the cloud and coming back. There is a last-mile access you have. There's a last-mile access I have. We're probably not connected to the same cloud, maybe. There is probably a cloud connectivity in the middle. That whole connectivity aspect carries tremendous importance because otherwise, we're basically a bunch of intelligence around the world not knowing what others are doing.

MH: Tell me about that connection, though, between the laws of physics and the laws of economics. It certainly seems to be, as we move forward, going to be a main driver to moving traffic, trying to balance those two competing laws, I suppose.

NB: Actually, I think they're, in this case, reinforcing and complementary. The law of physics is simply the further information is the longer it takes. The fastest one can travel, as far as we can see in connectivity, is the speed of light, and it's on an optical fiber, you can go quite fast with the speed of light. But the farther you are, the longer it takes. And then there is a lot of latency that is caused as a consequence of simply the stack, the telecom stack that we operate in. That we cannot escape. That's a matter of innovation, of thinning the stack. But the law of physics is if you're going to go further, you're going to take longer. So, keep information close if you want the experience to be better, and that's where the whole concept of edge cloud comes into play. There have been several businesses in this environment, called content distribution networks, that have evolved and are now coming to a very simple formation of what's called the edge cloud.

MH: Well, just that you did identify these eight core trends, and the top three are essentially AI, cloud, and connectivity. And what I'm fascinated by is how they're all reinforcing each other. We are not going to get more powerful AI unless we have more powerful connectivity, and the cloud is sort of in the middle, in between those two.

NB: We will likely get a lot more AI, but we will not be able to use it. We will be able to do a lot more in terms of inference models, in terms of training. Still, the usability the accessibility of the power of AI are only possible when there is an equally good connection to that. Yes, we can, of course say that there's a lot of embedded AI. And there is, but embedded AI is one aspect of it. All AI applications that are running on that cloud, when you have to access them, you have to then go through a layer of connectivity, or two, or three. If that doesn't improve, do we really harness  that power of AI and cloud appropriately? And the answer is no.

MH: I'm fascinated as well by one of those other eight core trends that I wonder if it doesn't get enough attention, and that is sustainability. There's a lot of talk about how much energy artificial intelligence will consume. What role does the network play in reducing that footprint?

NB: Actually, we did some analysis and simulation, we looked at network traffic that existed in 2019. 2019 is approximately the time when the first 5G networks were beginning to get rolled out, so call it still the age of 4G but the beginning of 5G, and then we did a simulation of what the traffic would look like in 2029. This That amount of traffic going up. And then, if we just do a linear extrapolation of the energy that is required to really deliver that amount of traffic, we will run out of energy in this world. And I'm exaggerating to make a point: you cannot allow the connectivity networks, the communication networks, to consume 27x amount of energy.

Therein comes the innovation, right? And the innovation is, and we've set ourselves an expectation that, despite this increase of 27 times, we should be able to bring the energy down versus what was being used in 2019 versus up. That means you're bringing it down, at least flat, if not down. And if you're bringing it down, one measure of success is you bring it down to half. And if you bring it down to half, then per-bit energy is down almost 50 times. That means 2%. Each bit should carry 2% of the energy that it used to carry in 2019. Imagine that curve of innovation that we have to go through in this industry.

MH: So then, how do you accomplish that? Is it just a function of innovating on the silicon and the software side of the equations?

NB: Largely those two help a lot. So, silicon is a very broad term. We see that Moore's law will continue to help, but with marginal returns, we're not getting 40 nanometers to 20 nanometers kind of return. They will continue to help. Then there are extremely good technologies that our ecosystem is building around chiplets, around increased performance of SerDes, which is connectivity within the chip, on-chip memory. There are a lot of subsystem elements that are coming into play that allow innovation and energy optimization, on the silicon side, so to speak but equally so on the software side, and on the software side, we can be a lot smarter.

AI plays a very good role here. If a certain geography is not experiencing the same amount of traffic flow as it should, it can then move to lower energy sources between terrestrial and non-terrestrial, micro, and femto. It can shut down sectors. It can shut down elements of processing and transceive and say, we don't need to use the energy now. And AI can make that self-deterministic action at that time. To me, it's an interplay of software, silicon, and system. And that interplay will allow us to innovate a lot in terms of getting to the targets that we all aspire to a sustainable world.

There is also an aspect of we will require better sources of energy. So, it's not just the demand side will get optimized. We will have to optimize the supply side as well—better sources of energy. We're already seeing tremendous innovation in terms of site solutions and moving to more green sources of energy. We're seeing that a lot of the telecom equipment is being powered by healthier sources from next-generation batteries as backup, and all of that will come to play. So, there is a demand side, software system and silicon. Then there is a supply side that must come together for us to meet these targets and, in my view, is a ‘no-choice’ journey we must go through.

MH: And it sounds like that's a partner relationship that needs to be established, maintained, and built. This isn't something that anyone telecommunications company or any one telecommunications company provider can do on their own.

NB: No, the full ecosystem. It would start from the service providers offering the service to consumers and enterprise. It will then come to the supplier ecosystem that they harness. Then it will come to supplier ecosystem we harness, and so on so forth. This will be a full ecosystem conversation, and it will have to happen as a joint exercise.

MH: You've talked in the past about the need to extend the scope of human possibility. What does that really mean, and how does 6G fit into that?

NB: I think what you're referring to is what I like to call the cyber-physical confluence.

MH: Are you talking about the singularity where we're all going to be just uploaded to machines?

NB: I almost would categorize it as that there will be a continuity between digital and physical. I don't see a future where, by the way, unlike some maybe far more acknowledged technocrats, AI will be a singularity. I don't see that to be honest. What I'm referring to is something more subtle and, more important and, more progressive and probably more positive. If you look at today, there's a certain physical attribute that we deal with. That physical attribute is that conversation you and me are having. That has already transcended. You and me are not in the same room. We're today doing this through a cloud continuum over a connectivity layer in different parts of the world. By the way, we don't even think of it. It's so natural to us, right? If you go back 20 years, this wasn't natural that there would be a video recording with somebody in two different cities.

MH: Yes, but Bell Labs back in the fifties and sixties was telling us video phones were, in fact, going to come.

NB: Yes. But the point is, as a consumer, there was a lot of novelty to this, right? But it's part of normal life now. Now, if you take this cyber-physical confluence to the next level, there are very glamorous examples of trying to operate an airplane or doing health surgery remotely. Okay, those are fine. But there are some very practical examples also where, if you look at a production floor, you can optimize the line capacity through connectivity when you're physically doing at a different location, and you're augmenting yourself in a digital reality, and you are executing a physical action in the digital world, and the factory experiences productivity. I think you will start to see a lot of these use cases through cyber-physical confluence. This will come through the evolution of connectivity, through compute, next-generation devices, and, most importantly, those specific applications that allow for that productivity to come through.

To me, Bell Labs has actually talked about it for a while, less publicly, and so there is this triangle of aspiration that we talk about. If the industrial world has to get better, then three attributes have to get better. Productivity, efficiency and safety. That means we should be able to do more with what we have. That means we should be able to do what we do today with less. And that means we should be able to do it safer. And if we actually somehow evolve that triangle, then we have done this right, and our cyber-physical research is oriented toward that improvement.

MH: How far off is that as a day-to-day life? Because I feel like we're already well on our way to that. You point out that this conversation here is something that would never have been considered 20 years ago as a reality, and now there's so much talk about the metaverse, and we're not talking consumer metaverse, we're talking industrial and enterprise, real applications for AR VR. Is this something that is in the distant future, or is this in the near future? Is this a pre-2030 conversation?

NB: Absolutely. It's a today conversation. This whole concept around human immersion through virtual or augmented or mixed reality to me is step two or three. Step one is to create a digital version, a digital twin of the physical realm. And if you look at the amount of digital twinning that is happening around the world right now, it is amazing. Without naming one of the top cloud providers in the world, their entire cloud network they have digital twinned, and they maintain it as a digital twin, in real-time. If you look at refineries, it is very common to have digital twins today because it's safer. It's very common to have digital twins of physical infrastructure like bridges and digital twins of cities. These digital twins exist today and will continue to, what they have to do is they have to increase in scale and improve in latency. And of course, this is driven through sensory input and telematics, which has been around, but the digital twin models are getting more complex, more sophisticated, and more real-and to me, that is already the creation of an alternative of physical reality into the digital.

Then the next question is how do you use that digital? And that's where, to your point, virtual reality or augmented reality comes in. So, I think, by 2030, this will be commonplace.

MH: Tell me then about that adoption and that evolution because you point out that businesses can't ignore emerging innovations or adopt them too slowly, but Enterprise has also become wary of that whole move fast and break things concept. So, what's the best way to approach that sort of triangle, as you point out, where we're more effective, more efficient, and we're safer as we adopt these next-generation technologies like AI and 6G?

NB: Yeah, I think it'd be good to look at enterprises and the industrial side versus the traditional CIO separately, right? The traditional CIO is looking for office connectivity, employee productivity, safety from a cyber perspective, applications which are more oriented towards, let's say, office use for collaboration, for ideation, for exchange, for audit, et cetera, versus an industrial application, which is looking at production environments of some physical nature. So, I look at them separately. I mean, that's why when we look at the metaverse, we call that either a consumer metaverse or an enterprise metaverse, or an industrial metaverse.

I moved to the valley about a year and a half back, and we hear a lot of buzzwords around the world, saying - hey, the valley's investing a lot today in metaverse. No, today, it's AI. It used to be metaverse, before that, it was Web3—all those buzzwords. And, in a way, that's true, but what I have learned here is almost every fund that is active has an eye out, not all, has an eye out to what the next evolution of the enterprise will look like, and they're trying to attack the problem in very different nuanced ways. And they're looking at how we can make that enterprise more productive. How can we make that office more productive? How can we make that factory more productive? At the end of the day... How can we make that supply chain more productive? That will create the productivity of this world. And that is what's happening all the time. And that's why when you hear of this decentralized Web3 or distributed metaverse or generative AI, they're all aspiring towards a more productive world. The industrial world is where you add safety on top.

MH: Before our time together is over, you touched on this as you mentioned that you moved to the valley about 18-months ago or so, and even earlier in the conversation, talking about the big thinkers in the space and how these are people who are talking about far-flung ideas that may be 50, a hundred years away or so, and I'm getting the impression that you're a little more grounded when you're thinking about the future. How do you keep your feet on the ground when everyone else is talking about robot-human hybrids and crazy things like that?

NB: I'm just not as rich, I guess, so I can afford to be – quote -- “normal.” Look, I'm not saying that they're necessarily wrong. It's easy to put an extremely high, far-out north star. I think that's fine, and nobody's going to know if you're going to be right or wrong; let's just put it this way. Unfortunately, when you do something like that, in my view, in a very humble way, you forget some of the journey to that and more than the journey it becomes... The lack of inclusion is what bothers me in those north stars.

So, when I look at my job, I'm privileged. I am representing one of the most important sectors that will digitalize the world, one of the most important companies in that sector, and one of the most important research labs in that sector. I find it to be my job that when we put something out there, it is relatively attainable. There are finite resources that are needed to get there, not infinite. That's why I'm saying I'm not that rich to just imagine everything is possible. It's not my position to say there will be life on Mars, and I'll put it there. That's not my job. It is my job that there is 6G out there, and everyone in this world has equal access to it.

<< Go to previous episode