Building the Wireless Future: Low-Power IoT, Edge Computing, and the End of the Gs

Welcome to Wavelengths, a podcast with Amphenol Broadband Solutions. What's going on, everyone? It's Daniel Litwin, the voice of b two b, and welcome to another episode of Wavelengths, an Amphenol broadband solutions podcast where we dig into all the major trends, technologies, the thought leaders, and the market movers in the wider world of broadband and telecom. I'm very excited about today's subject matter expert. Expert who's gonna be giving us a taste of that academic insight, which we're bringing to the show for, I think, maybe the first or maybe second time. Alas, we haven't done enough of these, and so I am grateful to be here with today's guest. But before we introduce him, let's go ahead and make sure that you're all caught up on previous episodes. We've got lots of thought leadership and lots of market analysis, on the podcast. So if you've missed previous episodes, make sure that you're going to our website, Amphenol Broadband dot com for previous episodes. You can also find, of course, more solutions, services, and other thought leading resources on the industry. You can also subscribe to wavelengths on Apple Podcasts and Spotify. Right? So just go ahead and hit subscribe, and you'll have a full catalog of previous episodes as well as notifications when we drop new ones. Alright. So let's go ahead and jump in, folks. I wanna paint a picture here for you. Imagine a world where your autonomous car negotiates traffic in real time. Your factory floor operates and orchestrates thousands of smart sensors with zero lag. A remote surgery feels as natural as an in person procedure. Right? Some of that may seem still a little fantastical, but that's the promise of tomorrow's hyperconnected networks, the infrastructure that they're supporting. And honestly, the research isn't all that far off. We're getting closer and closer every day to this kind of advanced and mature hyperconnected ecosystem of IoT devices and the, you know, connectivity infrastructure and backbone supporting them. Right? So really what we need to solve for today to achieve that future is looking at and solving today's toughest integration puzzles. Right? This moment feels especially critical too as we have this discussion. Recently, we saw twelve major European operators, including Vodafone and Deutsche Telekom. They urge regulators to free the entire upper six gigahertz band, right, or risk trailing the US in the six g race. We see here in the US, the USDA's reconnect program has surpassed one billion in rural fiber backhaul investments. Big. There's a lot of investments going into fiber infrastructure right now, which we've talked a lot about on the show. You'll wanna catch up on some of our hot takes there. But again, this underscores fiber remains the lifeline for dense wireless deployments. And then we see, for example, at NVIDIA's GTC conference not that long ago, we had T Mobile, Cisco, Meter, and others teaming up on AI native six g stacks that are built on their proprietary platform. This is signaling that intelligence is gonna be baked into every layer of the network. So there are some timely recaps here for you to to help frame up our conversation. But more importantly, what we wanna discuss today with our guest is how do we weave these various threads. Right? Fiber, spectrum, AI, edge compute, right? And the IoT ecosystem that all of that is enabling. How do we weave these threads into a fabric that can actually carry tomorrow's applications forward? Right? How do we solve some of these integration puzzles? Make sure everything talks to each other the right way. So I'm pleased to welcome on today's episode, mister Swarun Kumar. He's professor of electrical and computer engineering at Carnegie Mellon University and director of the YTech Lab. His team is actively working on research that is redefining wireless protocols from the ground up. They're working on everything from resilient edge architectures to AI driven resource orchestration. You name it, we're gonna dig in today, and I'm excited to have his expertise on the podcast. Professor Kumar, welcome. How are you doing today? Thank you, Daniel. I'm doing well. Thank you so much for having me. Absolutely. Absolutely. This is gonna be a good one. We're definitely gonna get in the weeds today. What I wanna do first is let our audience get a bit more of an elevator pitch on your background here. Could you, give our audience some insight into how you landed at leading the WITECH lab and some of the research you've been doing, right? What are some of your research areas of focus and expertise and how is that gonna color our conversation today? Of course. So I lead the Emerging Wireless Technologies Lab or WITECH lab here at Carnegie Mellon. Previous and prior to that, I was a PhD student at MIT where I worked broadly on WiFi, cellular, next generation wireless. And so this was a natural evolution for me when I started my academic career about a decade ago at this point. It was very exciting times for wireless. We had gone, I think if you think about the start of the millennium to the mid 2010s, we'd gone from virtually no smartphones to a world where we had billions, and we are now at a world where there are more phones than people. It's quite exciting times to be in. And my research back then was about building faster Wi Fi, building faster cellular. What does it take to go from three gs to four gs? Now we're talking about four gs, five gs, six gs. The Gs are expanding. My research at Carnegie Mellon has broadly been about faster connectivity. A lot of it in the context of the Internet of Things, something that you referenced yourself just a while ago. What does it take to connect these tinier devices that you What does it take to connect everyday objects around you to the Internet? And then you have different constraints, like energy constraints. You cannot assume reliable availability of power. If I want to connect my T shirt to the Internet, I don't want to have to charge it every day. So what does it take to build low power wireless systems? What does it take to build systems that are connected to the cellular networks, but don't have a battery at all? And what does that look like? And okay, if these sorts of networks exist, what can you do with them? I think you also referenced the numerous sci fi esque applications that can be enabled if this goes on. What are those sensing applications? Could you track location of devices? Could you sense the environment? Could you create situational awareness? Could you help fuel autonomous vehicles? Could you help augmented reality applications? That's another stream of research that's been going on in my group, not just understanding how to improve wireless from a communications perspective, but also sense the world around us. Let's dig into some of that research here a little bit, because I think this is going to open up a lot of threads of conversation. What y'all are pioneering there is pretty, I mean, pretty industry shaking, to put it lightly. So let's dig in here. How is your team working to redefine wireless protocol design? Right? Dig into some of the recent areas of focus, some of the things you've discovered, again, whether this is AI driven spectrum access, new link layer abstractions, I mean, whatever it is, what breakthrough are you most excited about right now and what's guiding some of your research? A few different directions. I think one of them that's very exciting to me, just thinking about how to scale the internet of things. So the objective here is take any object of interest, and from my perspective, slap on a very low cost RF interface, a tiny battery. Imagine the old school, you know, coin sized batteries that you might find in an old school watch, right? And then could a device of that form factor, of that power, last ten years and still communicate a small amount of data to a cellular base station? Now what does that look like? So this is the vision for what cellular IoT will look like. There's a whole range of protocols that have existed over the years that are broadly called low power wide array networking protocols, things from narrowband IoT in the cellular world. There have been the equivalence of this in the unlicensed domain like LoRa and the like. And the objective here is connected everything. And now this opens up a big can of worms if you think about it. First, how do you make sure these devices don't simply drain energy and are still able to create useful applications? Second, and I know this is a recurrent theme in this podcast series, which is scale. When you have a huge number of these devices, what is the pressure that that creates on the base stations? What is the pressure that creates on the backhaul? As well as broader questions like, okay, if this exists, why should it exist? What are the applications that really care about connected everything? What can I do with it now that I know that these devices are at scale? So this is one sort of grand vision, both in terms of improving power and energy efficiencies of these systems, as well as thinking about the new application that these systems can enable. And if you had to, you know, paint a picture of some of those applications, I mean, once we have, these systems operating at scale and they're able to operate efficiently and manage all these different nodes and properly feed power and connectivity to those different nodes in a right sized way, what applications become more possible? Yeah, there are many. Let me just pick a few from a few different verticals. I think the most classic application that I think is already happening in some small ways is just simple gatherings of sensor data. So imagine from your home, from your Nest thermostat to other things that you might find around in your home, such as your electricity meter or water meter, there's already a lot of progress towards making these wirelessly connected. So imagine the next meter at your home or the next device at your home not requiring, I mean, wiring or in person readings, just, you know, these are organically internet connected devices from the get go. So literally IoT devices from the get go. But there are grander versions of this. So imagine applications in the healthcare domain, not just the regular old wearables that you might have, but completely designed from the ground up systems for on body sensing. So in our lab, we're taking this a step further. Obviously we're researchers, so we are to think ten years ahead. So we've been developing essentially on body tattoos that can sense, for example, muscle fatigue and so on. And these are again wirelessly connected to infrastructure. We've built systems that are wirelessly connected smart pills. So literally pills that you consume down your throat and they can sense inflammations of the throat, the esophagus. And again, they're wirelessly connected. The idea being for these kinds of systems, instead of doing invasive procedures, you can't imagine a doctor coming to your home every day to check your digestive health, but you can imagine swallowing a pill every day and these are wirelessly connected. So there are numerous applications in the healthcare domain. If you now no longer have essentially your medical tests, at least some fractions of your medical tests are replaced by wirelessly connected sensed devices. There are many, many more in transportation. I think that's something that's so often mentioned, from smart traffic lights, smart road signs, to smart self parking assistance systems. Many, many more. List can go on and on. I can overwhelm you pretty quickly. So I'm happy to go to any of these, whatever is more exciting for you. No, I mean, are all really engaging. Let's stick with healthcare for a second because I think of all of the applications you're bringing up, new healthcare innovations obviously come with their own layer of red tape and safety considerations, of course. And maybe the time to roll out is a little slower in healthcare too. And this brings me to sort of a larger question, which is on how you're bridging lab innovation to some of these real world networks, right, which often hinges on the right kind of collaboration model where your research is both influencing these new use cases, but it's also anticipating and responding to the real immediate needs that these industries are feeling. So, I'm curious kinda how y'all approach some of your partnership structures in this research, right, whether that's open test beds, co innovation centers, IP sharing, you know, sort of like, active investments from different groups or industry associations, whatever it might be, what have you found as some of the most, you know, effective strategy for moving wide tech prototypes into, you know, actually more viable products and commercialized ecosystems? Like how do you approach that and develop and then execute on those relationships? Yep, I think that's a broader question that not just connects to healthcare, but also a lot of other domains. But let me take the healthcare question first and then broaden it a little bit. I completely agree with you that healthcare innovation operates at a different timescale than innovation in almost any other field, just because of the various hoops to pass through. That said, the way we've approached it is we don't do it alone. We are technologists, we're not doctors. So we partner with doctors. Close to CMU is the University of Pittsburgh Medical Center, one of the renowned medical centers on this side of the country. And we have experts that range from a wide range of domains. I've talked about the collaboration with experts in muscle health already, but the most recent fun collaboration we had is with a dental team and the dental school was interested in dental sensing. And so we built this smart sensing toothbrush for them. A lot of these requirements really emerge from the ground up. We go ahead, we are technologists, right? So we go ahead and talk to the application stakeholders. For example, we talk to doctors about the problems that they have, the technological bottlenecks that they have, and this is how we start. The way we innovate in the medical space is quite interesting. We start with a very simple pilot studies. These are all obviously IRB approved within our respective institutions. And we have participants where we demonstrate something at small scale. The fortunate and good thing about wireless research is it's not as invasive as, let's say, a new medical treatment where you have the safety test with, let's say, giving you a new medicine. There's a whole range of different steps there. Whereas if you're just talking about a sensing system or a scanning system that is not necessarily going into your body or something that's just on the surface of your body, the amount of red tape that you have to encounter is significantly lower for obvious reasons. But I think the real challenge is sort of getting across that hump, which is going from something we show in the research lab to something we show in pilot studies, to something that we can then commercialize. And there we work with external partners. And that is a whole other enterprise. A lot of health related innovations in wireless that have not had as significant a bottleneck are those that are doing not necessarily diagnostics for healthcare, but broader fitness sensing. So that has a lot less directive. For example, if you're just interested in activity monitoring and the like, it becomes much easier to do that at scale. For broader industry solutions beyond healthcare, we partnered with a variety of people from Google to companies like Bosch and even more I'll give you one unique example. We had one with a company called Bridgestone, which is a tire manufacturer, and they were interested in sensing the health of tires using radar, which is also a wireless sensing system. They were interested very broadly. They just wanted some sensor that can tell you a tire if it's worn out, where a tire can inform you if it's worn out, if it's abrased or not. And we built this tiny radar based sensing system for them. There it's kind of the other way. We already have a commercial partner who's ready, willing to deploy something and they just want new solutions that they wanna try out and test. So this creates the kind of short circuits, the whole thing, because we're able to make impact on the industry much faster. Hearing that, you know, this research is already driving some of these next gen use cases is very encouraging. You know, to take it back to my intro, all this kinda like, woah, next level application environment. It's not really that far off, I mean, from from what you're describing, which is exciting. And I think it shows that, there's an understanding from the industries that you're, you know, researching around and creating solutions for that it's in their best interest to collaborate with top researchers to, help support the development of y'all's research at at the, you know, YTEC lab, for example, because it's in their best interest. Right? It it will motivate and is already, you know, creating solutions for very real problems, whether it's in health care, whether it's in, you know, industrial manufacturing, transportation, you name it. Let's dig in a little bit more into some of the interplay here between fiber and wireless, because I think this is an important part of y'all's research, right, sort of one of the underlying factors that y'all have to solve for to help enable a lot of these use cases. So as we gear up for ultra dense six g, and we have this like six g race across the world, it's obviously no longer enough to treat fiber and wireless as separate silos that develop independently. Because the demands on both are starting to increase and they need to be right sized to maximize each other. Right? So what are some concrete architectures, control plane integrations, or sort of approaches that you all are researching and have found to be viable or that you're excited about that you think will better let fiber and wireless evolve together as a unified fabric? Give us your thoughts there. Yeah, great question. I think Here's my perspective of this, right? And just wearing my systems hat, it's so easy to focus on, I mean, as a consumer, I think most people who might be watching this podcast are just interact with wireless as a consumer. And you don't think about the enormous layers of infrastructure that goes on at the backend in connecting you to, let's say, a data center or somebody else on the other side. And from a systems perspective, I think it's important to think of this as, imagine this as plumbing, right? You want to go and widen the narrowest pipe for enough water to flow through network. And fiber often is the bottleneck. In some cases, wireless is the bottleneck. So you need to evaluate your system holistically and expand where that bottleneck exists. And so I think it's extremely important for you to focus on every aspect of the system, find out where those bottlenecks are and expand. Networks are scaling in a massive way. And so it's important both to Now your gut feeling might be, Hey, five gs, six gs, so on and so forth, it's all about throwing more spectrum, throwing more bandwidth at the wireless operators. Now the wireless operators are, even if you throw all the bandwidth in the world at them and then you keep the fiber backhaul the same, it's like trying to squeeze the ocean into a tiny narrow pipe. That just doesn't make sense. So you need to think about the system holistically. And second, I think there's another aspect that, I mean, was then a term that you just mentioned, ultra dense. And this is where I think the need for fiber is underappreciated. And I'll tell you a story, at least in my book, how important fiber is. A lot of what we do, what we have to do for demonstrating almost any research that we do, I'm just wearing my researcher hat, not even talking about anything the industry does, is set up test beds. So go on the rooftops of Carnegie Mellon and you'll find a bunch of antennas, like prototype six gs, seven gs, whatever you want to call it, antennas for our own systems here on campus. Connect a few things. And then we try to scale that to some of the neighboring regions around Pittsburgh just to scale up our test The first bottleneck that we face is just access to a reliable backhaul. And if, let's say, new funding agency wants to sponsor a larger scale effort for us, the first question they ask us is, do you have the backhaul needs? Because there's no point in the base station getting access to enough data if the backhaul cannot support it. And I think another bottleneck that we face is precisely what you said, the interface between that antenna and the backhaul. And I'll tell you where we face challenges as researchers, which I think is something that transfers to anybody who's innovating at a very fast pace in the industry as well, which is if you plan to deploy something and you call that, let's say, six gs now and deploy it, a year from now your competitor has done something better and something smarter, right? And now you need to keep pace. And this is something that is not very hard in other domains. If you go to the AI world or if you go to the software world, broadly speaking, they innovate all the time. What Google search looks like right now is very different from what it looked like, forget a year ago, like a month ago or a day ago even. And they innovate in rapid timescales because it's just a software upgrade under the hood. Now, I think for a long time, there's been this vision of can six gs to whatever the, you know, five gs to six gs or six gs to seven gs or whatnot be just a simple software upgrade? My answer to that is yes and no. Yes, in the sense that we want rapid upgrades to the system, at least from a security point of view, if not anything else, to happen. But we can't do that if we don't have the hardware support to enable it. That means we need to make sure that when that software upgrades happen, nothing breaks, which means the interfaces between the wireless and the backhaul, the fiber is well figured out, carefully ironed out. And we also have to make sure that the system as a whole scales. So if tomorrow you have a software upgrade that supports twice as many users, your hardware can keep up. So a holistic and careful thought process for how to make a software evolvable future G stack has to be envisioned. I don't think we're quite there yet, partly because there are so many different people developing these things, they're just not talking to each other enough. I'm curious about some of the, know, well, let's be real, right? Every network layer has its own headaches and then those just get magnified as you describe the situation that you just described. Right? As these two ecosystems have to develop in tandem, you start to realize the pain points of, yeah, mean, we wanna upgrade our wireless connectivity infrastructure, it isn't always as simple as just a software upgrade as much as things are, you know, getting more capable on that front as well, right, with the ability to leverage existing fiber infrastructure and just continue to maximize, the wireless connectivity on top of it. But regardless, again, every network layer does have its headaches, but some of these pain points really only surface when you try to marry, right, wireless and fiber more closely. You try to kind of wed them, and and wedge them together a little more closely. So I'm curious what are your labs top three wireless system challenges today? If you had to summarize them more. If there aren't three, you know, just give some of the ones that come to mind immediately. And how do those collide with or align with some similar challenges with the realities of fiber backhaul? I think the top one that comes to mind that straight away interacts with fiber is scale. I think that's something that we've been seeing for networks across generations over the years now. I'm sure if you take the quantum of data traffic that's going on and let's say cellular networks as an example, as a whole now versus even a year ago versus five years ago or ten years ago, orders of magnitude expansion. And so when you have that sort of scaling, every element of the system has to keep pace with that level of scaling. And that's a challenge. And so I put under the broad scaling umbrella, everything from interference to enabling diversity of services, to enabling reliability, robustness, all of these while the system scales. And so that I think is the bigger challenge. One of the aspects that my lab looks at is energy. There is energy consumption both at the end device that you want to minimize. There's energy consumption at the infrastructure as well that you wanna minimize. One of the things that I remember once talking to someone whose primary job was just setting up these future G base stations, right? And I asked them, what is the most surprising thing you learn about like setting up these base stations? And they were like, the cooling that you need, right? It's just, you don't think about that, right? You think about a base station as just wirelessly connected infrastructure. You don't think about it as ultimately there's a bunch of compute infrastructure that's heating up and you need to cool it down. And then that's where the energy consumption is. And so if you think about the operating expense costs for a lot of these operators, cellular operators, it is energy bills. And so then if you're sending more data, sorry, more energy bills, right? And that increases their operating expense in addition to all of the capital expenditure from fiber and all of that, right? So how that, how do you address that problem both from the operator perspective, but also as you or I, the consumer, who then has to worry about, okay, how long can this call last? Or how long does this video call last because my phone's battery is gonna run out? So this problem actually works both ways across the board. And third, I wouldn't call it a challenge, I would call it more as an opportunity, which is why are we doing this in the first place? I think this is one of the biggest questions that not just the industry, its consumers are asking of us. So here's how I would put it. Phone plans aren't getting any less expensive. If not, they're getting more expensive. And now your consumer is asking you, Do I really need this extra gigabits per second or extra megabit per second? A. And B, if I get this extra megabit per second, what can you give me? What is the application? Oh, okay. I know ten years, twenty years from now, there's going to be these robotic cars that are flying around and there's going to be this, you know, robotic surgery that's going on. But today, next year, two years from now, what is it that this higher connectivity can deliver for me, the consumer, that justifies me paying that extra dollar on my phone plan? And I think that's a question we have to urgently address. And I feel like there's this disconnect. You talked about disconnects between fiber and wireless. I think there's a deeper disconnect between two different industries, which is the application developers and the infrastructure developers. I'm bucketing everyone from the operators to the fiber to wireless researchers like us. We think infrastructure without thinking what is this infrastructure for. Because that's sort of what we do. We just feel better is always universally the right thing to do. But then the question to ask is, where are the next future apps going to be? And do they justify this investment? And a different way of putting it is, can we tailor hardware investments, our infrastructure investments in a way that actually meets those future demands? And I think that's another research question from my book because I want this industry to actually go to gigabits and terabits and exabits per second, right? And my question is, what are those applications that matter that really make these useful for the consumer? I want to follow-up on that communication gap. As a research lab and the fact that you all are facilitating a lot of collaboration between subject matter experts in different industries, some of the, you know, users, I guess you could say, of these applications or, you know, the entities who would be deploying them, and then yourselves, the infrastructure researchers. How are y'all working to try to bridge some of that communication gap between the folks that are developing the applications and the folks that are developing the infrastructure. If that is one of the big layers that is, you know, impeding or at least slowing down the ability for, the use cases in the infrastructure to align more closely and develop in tandem, right? What's y'all's approach to bridging that gap and what have you seen work? Yeah, I think a lot of it is making conversations happen. Like I said, the first, I think as a researcher, it's often tempting to just go after, okay, some metrics, okay, throughput, scale, performance, Good metrics, they're good metrics. What are the three numbers, right, that you care about the most? But if you ask, if you go, sort of wear a different hat, which is, I think this is one of the advantages of being a wireless researcher who mostly looks at user interaction. I'm one step closer to the user than fiber, right? In the sense that you think of yourself as a user too. What would I want my phone to do tomorrow that it cannot do today? Okay, if it's medical applications or if it's transportation applications, what would I want as a consumer of transportation, right? We all healthcare, which we all are, I wished my network enabled, right? And now you can imagine only so much. So you go ahead and talk to these stakeholders. I think this is one of the unique advantages of being in academia, that is you get to talk a lot. You get to talk a lot to your students until they're bored, but you also get to talk a lot to stakeholders who come in, experts in medicine, to experts in transportation. In Carnegie Mellon, we have centers for transportation, we have centers on neural health, we have centers on a whole range of topics. And so those bring in together experts who then deal with their own pain points. And I think the way I think I found the problem to be interesting is, interesting problems to emerge is approaching them with humility. As a classic example of this, Carnegie Mellon has this campus in Africa, in Rwanda, the CMU Africa campus. And we had a seminar at one point where the CMU Africa folks came in here and presented their work. And I found a lot of it, actually, the majority of their faculty work on the digital divide, right? Which is how connectivity is such an important element to deliver services to rural regions. Very simple things like a farmer, a rural farmer, knowing when to come and bring their produce to the market. And they don't have a good way to get that because just forget anything else. The wireless infrastructure simply doesn't exist. So then how do you enable it? And so these sorts of conversations are when you realize, okay, the problems that we are used to are very different from the problems elsewhere. And the only way you truly understand them is to work with local stakeholders to really understand what matters and what does not. Like there's no point building, thinking, building an Uber or building a market app ecosystem if that underlying infrastructure just doesn't exist. And so understanding where the problems are important. And I think one of the things we're doing with one of the moonshot initiatives we have here called Compute and Communication for All is trying to build that infrastructure. So sometimes the problem here is we understand what the problems are and we are doing something else. Let's say we're building the smart app for rural health in Africa, and then we realize, okay, the network infrastructure doesn't exist. So what's the point of building the app if the network infrastructure doesn't exist? I think having those conversations really helps in tailoring your system to make the maximum impact. This is incredibly insightful. Thank you again for everything that you're, laying down here for us on the podcast. I've got just a couple more questions here, before we wrap things up, but let's talk about more devices. Right? All of this infrastructure is supporting, an increasingly active and dynamic IoT ecosystem. We could look at the wider one, or we could even hone in in each industry. Each industry is even seeing a more complex and dynamic IoT ecosystem. The devices are speaking to each other more. Devices are increasingly capable of processing data at the edge, right, and being intelligent nodes themselves. We're seeing these be, you know, industrial deployments. We're seeing them be consumer, facing deployments and everything in between. So as more intelligence moves to the edge specifically, right, you know, we're seeing AI supported edge compute now, really widen the possibilities of these IoT networks being more capable, you know, intelligence gatherers and and processors essentially. Right? So as more of that intelligence moves to the edge, aggregation points for fiber infrastructure also need to do more now than just pass traffic and just consolidate fiber. Right? So I'm curious your thoughts on how fiber aggregation, has to adapt, right, to support tomorrow's and even today's more context aware AI driven devices. Is there an issue there that has to be addressed, right? What are your thoughts on the connective tissue between those two forces? Yes. And I think there's a grand opportunity here for people building infrastructure. But first, I think it's important to explain why this is happening. I know people are so excited about edge, edge, edge. There's all this, you know, earlier it used to be the cloud and now the edge and now there's fog and everything in between. So it helps to remind ourselves why this is happening and should this be happening as well. It all boils down to, I think, Einstein, right? Speed of light as a limit, as a universal limit. If I want my, let's say search engine result to show up within a finite amount of time, and that request has to go, let's say from Pittsburgh here, where I am to the Bay Area and come back around, just that round trip time because of speed of light is just going to be way too much. You're several milliseconds too late already for me to get that response. So this is where the edge helps, right? If I had mini Google or whoever, right, right next to me, that's a shortcut that makes my life a lot easier. I'm gonna get a quicker response than having to go all the way somewhere far away. And so this trend is not new. This has been happening, one step at a time, right from content distribution networks for like decades now. It's just that now it's reached an extreme level where the computer is kind of part in your device, in your home, part in your tiny cell tower and part in your city as well. And I think it's a natural progression of things. I think the opportunity here is for infrastructure developers to rethink themselves. Because for the most part, the objective function for, let's say, anyone developing wireless, from wireless to fiber to anything in between, we've been like pipes. We are ultimately developing the plumbing for the water that is bits and bytes on the internet. We've been glorified plumbers all along, right? But now if there's computers everywhere, not just inside the data centers of the big tech companies, but also alongside our infrastructure, alongside base stations, alongside routers and access points everywhere, and they're developing meaningful services. Now, I think it's not wise anymore to think of these companies as networking pipes, but as networking pipe and some intelligence embedded in what these boxes are delivering. Think at least my tiny complaint with that is that the mindset of the industry has still been pretty much, Oh yeah, we are we have pipes plus some computers. That's it. We're leasing out space. We are leasing out resources. That's what we've always been doing versus thinking, Hey, wait a minute, we have all this compute infrastructure right here, why can't we use it to deliver something interesting? What is it that we can deliver as a value add that generates more revenue for us? And so there's this recent trend, at least in the research world, that's I think taking some traction in the industry as well on broadly called joint sensing and communication, which is don't think of yourself as just a communication pipe. Think of yourself as something bigger. Think of all of the applications you can deliver, just for example, signal data, using data analytics, the data that's already flowing through your network. And then what is it that you can enable using that? And I think for me, this combination of AI and edge and all of that, it would be a shame if that was just, Oh, let's make Google and Microsoft of these other Meta and all of these other folks, let them do more AI on our stuff and we'll build more things for them to do stuff. We should do stuff. We can do stuff that they cannot because we have access to metadata that they do not. And I think that opportunity is a serious blind spot for, I think the infrastructure industry as a whole. They're sitting on a gold mine, they just don't know it. Or maybe they knew it, but they're not willing to mine. Love that analogy there. That's a good one to end on there. All right, Professor Kumar, last question here for you, and thank you again for all your insights today. This has been enlightening. I think our audience has a lot to chew on here, and hopefully some action points to apply themselves or some challenges that they can reflect on and start to address. Because as we've covered here on the podcast, a lot of these major challenges that are reshaping the larger telecom industry, you know, broadband, fiber, wireless, everything, their team efforts. And they often have a lot of different stakeholders with a lot of different interests. They're not necessarily competing, but the pressures on all those different interests are different. And so there has to be active collaboration, active communication, like you were saying earlier. Right? Infrastructure has to be speaking with application developers. You know, fiber has to be collaborating with wireless. The researchers have to be working with the subject matter experts. And so the more we can encourage that kind of collaboration, the better. So hopefully this podcast has given you all some more of that, right, and given y'all the tools to help tackle some of these challenges. Alright. Professor Kumar, last question here for you. We're gonna crystal ball it a little bit. You know, we don't have to go ballistic here, but looking ahead, like, let's say another ten years, twenty, thirty five communications architecture, we could have networks centralized around digital twins, let's say. We could have more distributed and decentralized AI meshes of networks. We could be looking at seven g, eight g, who knows at this point. But from a more practical perspective, and informed by your research and the the kind of future that you're actually actively developing with your own research and your team's work, what vision, right, or hybrid vision for communications architecture in the next ten years do you believe is gonna prevail? Right? And what are some of the key drivers that are influencing your take here? I am going to make a prediction in that the Gs are going to die. I think The death of the Gs. You heard it here first, Bose. Yeah. That is, I think you kind of see this in almost any trend. After a while they run, you'd use for any technology, they stop issuing version numbers. After a point it becomes meaningless they rebrand, right? So the same thing is going to happen to the Gs. I can't imagine twenty gs, twenty one gs. Instead, I think what's going to happen in the future is is a softwareization, is that your G has changed from yesterday to today. Your wireless network and infrastructure and the services that it's providing is changing constantly. Security updates are happening constantly. System improvements are happening constantly. You don't hear about Google search version one and version two and version three. The same thing's gonna happen to infrastructure too. I think it's gonna happen, of course, for logistical and physical reasons. It's gonna happen a lot more frequency to the software fabric. That's gonna be constantly upgraded. The hardware fabric is going to improve and upgrade and have to keep pace as well, but it's going to be very heterogeneous. Because I wish I could say there's going be one technology. Oh, everything is going to be just fiber. No, there'll be remote regions where you need to have some wireless backhauls too. The infrastructure is gonna continue to be heterogeneous. It's gonna continue to be dynamic and evolving, but the software is gonna rapidly and rapidly evolve. I think that's the only way to sustain a competitive industry. Otherwise, I wish we are not talking about twenty one gs or twenty two gs a decade or two from now. What a great prediction, because I think it speaks to, you know, what you're imagining is the maturity of the ecosystem. I mean, once once this interplay between wireless, fiber, and the applications, right, are all more in sync, There don't need to be these huge, you know, ten year waves of infrastructure deployments and ten year waves of, you know, reworking to level up to, you know, four gs to five gs or something, right? We will have an ecosystem that is so in sync that improvements, like you said, happen daily. And I think the research you're doing is helping lay that groundwork and create more of that needed communication between disparate parties that's gonna make that level of of rhythm, you know, of all ten fingers playing the piano, right, actually happen. So, professor Kumar, thank you so much for your analysis, your insights, and your hot takes today. This has been great. Again, folks, we've been chatting with Professor Swarun Kumar, professor of electrical and computer engineering at Carnegie Mellon University, director of the YTech Lab. Professor Kumar, if folks want to tap into some of y'all's research at Y Tech or look into some of your various published insights, how can they learn more about you and y'all's work? Great. They can reach my lab at ytechlab dot com or just search for me and you'll find me. Perfect. You're famous enough. Yeah. Right? Folks will know where to find you. I like it. All right, professor Kumar, thank you again. This has been great and looking forward to more. Thank you so much, Daniel, for having me everyone, for tuning in to today's episode of Wavelengths, an Amphenol Broadband Solutions podcast. If you like what you heard and saw today and you wanna tap into more of these kinds of dynamic and actionable insights for the industry, make sure you're heading to our website, Amphenol Broadband dot com, and make sure that you're subscribing on Apple Podcasts and Spotify for all the previous episodes and notifications when we drop new ones. And check out our LinkedIn too. We're gonna be doing more on Amphenol's LinkedIn here soon, dropping short form snippets, hot takes, and discussion starters. You're gonna wanna get involved. Alright, folks. I'm Daniel Litwin, the voice of b two b, and we'll catch you on the next episode of Wavelengths.