In The Arena by TechArena - Advancement of the Network with Roy Chua

Episode Date: March 26, 2024

TechArena host Allyson Klein chats with Avidthink Principal Roy Chua about advancement of the network across 5G, edge, AI integration and more as the two share insights from MWC24.

Welcome to the Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome to the Tech Arena. My name is Alison Klein, and today I am really delighted to be joined by Roy Chua, principal and founder of AvidThing. Welcome to the program, Roy. Hey, Alison. Good to be here and good to see you. I guess the post-MWC craziness. Yeah, we just came back and survived another year of MWC. And I know that you were reporting from MWC and as was I. I really wanted to have a chat with you because I love your insights about the network. What was your biggest takeaway in terms of the network arena and what you saw in Barcelona in

terms of the vibrancy of the communications industry. Yeah, I think the fact that we had 101,000 people show up, which was more than the 95,000 that GSMA expected, I think that in itself, I think tells us that there is still a lot of activity around telco. And, you know, despite all the challenges that telcos face around the world, the reality is people still do care. There's still a lot of investments. And I think there is an ongoing move to try to modernize the telecom industry.

So I would say a couple of things that sort of hit me. One, on the telco side, I don't think anyone could have avoided it. There was AI, ML, Gen AI everywhere. So let's get that out of the way. That was one. The second is this notion of telcos wanting to monet for programmability, for better integration with applications, and also a hope and the promise of monetization, which they've been seeking around 5G. So I think that was the other element that we saw around telco. And then the third element that we were seeing is the fact that communications is a lot more diverse. or the combination of multi-network or multi-path

to different devices and provide coverage in rural areas and other places where you might not easily run fiber. So I saw a good amount of that. And then I think the other element that's promising is private wireless or private 5G for industrial applications, for other rural applications, for more creative deployments, educational institutions, and so on and so forth. And coupled with edge computing.

So looking at private wireless with edge computing and seeing some aspects of that, especially as it complements Wi-Fi or replaces Wi-Fi, depending on what use cases you're looking at. So a lot of those elements across the board. And then in general, a desire to seek new applications and new ways of monetizing the 5G infrastructure that's been built out over the last couple of years. And that continues. I don't think there is a magic bullet, a silver bullet.

I don't think that's even a bronze or tin bullet. We're still working at it. I have so much to unpack now that you've given me those highlights. I mean, I almost feel like the interview is done. I'm just kidding. So let's just talk and start with the resiliency of the telcos. You know, I think that there's been talk for years with the rise of the cloud and the large cloud service providers that the telcos are going to struggle to keep up and to keep pace with the delivery of technology. Yet the MWC was a vibrant conference with tons of discussion about deployments and new

opportunities. What do you think? You've been in this space for a long time. What do you think maintains the resilience of the telcos in terms of their unique ability to provide services to both consumers as well as enterprises and what, I think, points to the importance of the communications infrastructure and the increasing importance of communications regardless. And so they've been around for some time. They continue to, I would use the word, innovate very carefully. They continue to progress, at least, by the very least. So I think it's key that telcos exist for communications and communications are becoming more critical. So I think that there's two elements in that.

Now, telcos are usually viewed as national infrastructure or critical infrastructure across the board. And so there is an incentive for governments and regulators to obviously keep them around or keep them healthy or at least barely healthy in any sense, right? I think some of that keeps the telcos going. But the reality is that some, not all telcos, but many telcos still continue to innovate. They continue to modernize.

And when push comes to shove, I think many of them will react and they eventually will be forced into action. And so I think we're seeing some of that. And to your point, there was this talk about hyperscalers and clouds and OTTs coming in and destroying the telcos and hollowing them out. And I think the reality is that I'm seeing not many telcos get with it. We don't see this hyperscaler telco tension very much on the show flow anymore. In fact, we see more collaboration. So we see the telcos, by and large, all embracing

some from a hyperscaler, some from a hybrid infrastructure, where the public clouds are part of it, multiple public clouds at that, and then building out new services in a more creative way using this hybrid infrastructure. So seeing a lot of that. And one example that I think points to all these things going on is seeing more tier one telcos launching mobile infrastructure, speaking of 5G, on hybrid infrastructure that includes a public cloud component, a private cloud component, disaggregation, CICD sort of cloud practices across the board.

And I saw that with TELUS and with NTT. These are tier ones. And I think NTT has 87 million subscribers. So again, it's a very sizable tier one that's deploying a modern hybrid cloud infrastructure with disaggregation, with modern practices in a potentially very large network. So I think that gives me hope that many of the telcos worldwide

are in fact moving. I mean, again, there's a lot of challenges. Yes, ARPU is declining. Yes, a lot of them are challenged. Look at what's happening in Euro. But I think they continue to survive. They continue to, I wouldn't go as far to use it with Thrive yet, but certainly they are surviving. They are changing. They are adapting. And when forced to, I think many of them will. You know, one of the things that I was thinking about is just communications infrastructure management is really complex and hard and very different than managing large scale data centers. And I think that that hybrid that you speak of really speaks to that, right?

That there is an opportunity for collaboration. You know, when I look at 5G, one of the things that I wanted to talk to you about is the innovation to embrace the full standard. And, you know, when we talk about 5G, and we've been talking about it for years, some of the things that we've talked about is cloud native networks, network slicing. You brought up private 5G and edge deployments, the virtualization of the RAN. These are some of the topics that I had wanted to go to MWC and see where we were as an industry with the deployments of these technologies. And I did see some progress. I'm curious as what you saw, and if you feel like we're maybe getting over that hump of, or, you know, the trough of

disillusionment, if you will, associated with some of these technology maturities, that telcos are adopting them, that enterprises are doing things with private 5G or private mobile networks, et cetera? So I think the answer is yes, but it's a cautious yes, right? So I think let's start with sort of the macro 5G virtualization, the cloud RAN, the segregation, open RAN in sort of the extreme case. So I think it's happening. I think when someone like, say, a TALIS is telling you that, you know, they are replacing their base stations and 50% of those were traditional sort of RAN and 50% is going to be open RAN, sort of new RAN, right, with the various providers and vendors that they have. I think that speaks to the level of comfort that they have. And what they're telling me is that 98% of the KPIs

for open RAN with the vendors that they've chosen actually meet or exceed those of traditional RAN. So I think that's promising. So I think some of that is overcoming the challenges. And when someone as conservative as NTT Docomo decides that they can deploy open RAN- Yeah, that was pretty cool to see that. I mean, yeah, 87 million, right?

Even in the hybrid cloud architecture with a public cloud provider and a private cloud. I think that says that, I think, you know, the tide is turning. Again, it's not immediate. It's not like we're saying like in 2024, everything's going to be switched and there's no challenges. There are still challenges. There are still deployment challenges. There are still deployment challenges. There are still management challenges.

And there's still going to be massive MIMO, you know, sort of performance challenges for sort of large-scale deployment and large capacity. But I think the tide is slowly shifting and turning. So I think on the macro side, 5G, it's happening. But bear in mind that a lot of the telcos are still hurting from that big capital investment and trying to get 5G on the new radio side out there. And some of them are holding back that 5G SA transition for the core, standalone core, because they don't know if that brings them new monetization opportunities,

increase in ARPU or new profitability. So I think there's this cautious optimism where they're looking at these things, but they're worried that they could overinvest without the corresponding increase in revenue. So I think that there's that still sort of holding them back. On the private wireless side, yeah, we've said for a couple of years now that we're going to see this drive towards private wireless adoption. And the reality due to many, many reasons, part of it was supply chain, part of it was maturity,

part of it was the user equipment not quite being there. And complexity of deployment was one of the big factors. Regardless, private wireless hasn't taken off as quickly as I think most of us have hoped. And the reality now is that it's taking hold. It's a little slower on the uptake, but many of the demonstrations I saw with the server vendors, with a lot of new players coming to the private wireless industry, gives me hope. There's a bunch of sort of Taiwanese OEM, ODMs that are coming with cheaper solutions, easier to manage solutions that could sort of break open the market. Likewise, with some of the incumbents in the space,

I think we're seeing more deployments. But again, the uptake is not as fast as we thought, but industrial use cases are leading the pack. The traditional transportation ports and all those use cases are leading the pack. Even in educational institutions, I'm hearing a lot more about private 5G deployment across universities at large scale and using that as a critical infrastructure for new services in parallel with Wi-Fi. So it's happening. It's just not

happening as fast as I think we had hoped originally. I think that that's probably indicative of the industry that we're dealing with too. I mean, I think that the telcos tend to be cautious about what they're deploying. They've certainly invested a tremendous amount in 5G, and that gets into the monetization question. But before we leave private 5G behind, I think the other thing that was interesting to hear about at MWC is some of the policy work that's been done around dynamic allocation of spectrum, especially in the Biden administration, but other countries expected to follow suit that will open up the barriers of private 5G

spectrum allocation in a way that people can actually use the technology. I think it's going to be interesting to track that. From an Open RAN standpoint, I think that, you know, the Tokamo announcement was fantastic. I also just think that I'm seeing continued innovation from silicon to systems to software and O-RAN. So the vendor community is speaking very strongly that they still see a lot of TAM in this space, to me at least. And I'm really anxious to see if we start seeing a hockey stick of deployments there. When you look at, I have to talk about AI. I know you said it's there. I was really surprised because ChatGPT came out in November of 2022, and I don't remember AI being such a topic at MWC 2023, but it hit me in the face at this event. AI was everywhere.

And actually, I found the use cases really practical in a lot of cases of where you could apply AI to manage the network and then manage customer engagement. Obviously, the easy one is, you know, customer support features being supported by some sort of LLM. Okay, cool. That makes sense. But even AI to do some automation of network functions seems to be, at least in my mind, the more substantive opportunity of AI. What did you take away? And do you have a sense of how mature the efforts are in this space? Yeah, so you're right. There are sort of easy picking.

So I think, you know, in terms of AI and ML, so the telco industry and the NEPs, the network equipment providers and, you know, the software providers have been using elements of AI and ML for many, many years now, right? So if you look at optimization algorithms and traditional AI and machine learning, right, in terms of the old CNNs, convolutional neural networks, you know, RNNs and LSTM, those methods have been used within network management optimization for

some time now. Even some of the sentiment analysis and customer, sort of customer support, next best action in terms of customer, you know, back and forth. I think those existed even sort of pre-Gen AI. Now, Gen AI came along and I think exploded that, you know, sort of the whole market sort of exploded around that. And so you're absolutely right. I saw the same thing as you, that there are sort of very obvious deployments of Gen AI, customer service, where it's aiding the customer support rep in terms of summarizing what's going on, even providing sort of next best

action with better outcomes than sort of traditional AI ML that they've used. So I think that part is very clear. All the network equipment providers and the solution providers, OSS, BSS are building AI into their platforms using the data that they have. And of course, you have Salesforce and ServiceNow that are trying to break in the telco market, doing the same thing. So I think we're seeing the same thing across those. And that makes sense, right?

Those use cases, summarization and suggestions and all that. Next, you know, I think that makes sense. On the network management side, I think the optimization element is still early, especially with Gen AI. With existing AI, sure. Those algorithms existed, they've been tested. Gen AI, I think that is still in process. Troubleshooting, I think that makes sense.

I think what I'm seeing in terms of root cause is more around root cause suggestions where you're looking at the network logs and suggesting areas to dig into as to what might be the cause of trouble or looking at events and summarizing those or alerting. I think that's in progress at this point in time. But I think that's the next domino to fall right after customer service. And then the more complex things where intent-based network configuration, network design and

planning, I think that needs to come later. And in fact, I was chatting with some of the folks there around how Gen AI could be used for planning. I think the agreement with experts is that I don't think we're comfortable enough to have Gen AI do intent to direct network configuration. I think that's going to be an interim step in between where there's a model-based element where intent generates a model, the model can be verified, and then model to configuration.

I think we'll see that, but I think that is going to come a little later. I mean, parallel marketing and sales will keep using Gen AI to generate new offers, generate your own personalized plan or the right images to incense me to buy something. I think that parallel track's ongoing as well. But yeah, Gen AI definitely did have a lot of impact, saw a lot of demos, although at some point they all blurred and looked very similar from vendor to vendor. I mean, you could sort of guess, yes, I'm going to show that and I'm going to show that. And it became sort of a blur after a while.

But it was interesting. It was interesting for sure. I think readiness, we've got to be careful. What showed me that we still have to be careful around Gen AI was seeing a couple of demos and using Gen AI, and they said, yeah,

this is what did they say? It was guardrail. It was like protected. So you can only answer very specific questions. I'm like, really? Sounded like a challenge to me. So I'm like, so I go in there and I break the problem. Change it to do something else.

So they said, hey, Roy, it's just a demo. I understand it's just a demo. But what it says to me is that, yes, it's a demo. But whatever framework we're using doesn't have robust guardrails. Right, exactly. If we had a robust guardrail framework, even if it's a demo, if you constrain it, it shouldn't break out of the constraints regardless. So it says to me we're still in the early days, very promising, very enterprising use cases out of the equipment providers and the ISVs and the hyperscalers. But I think we just need to be aware that the whole framework is still early, immature. We

got to be careful about where we place it. That's all. That makes a lot of sense. I did walk through a lot of demos where I felt like, okay, your company had a meeting and you knew you needed an LLM demo in your booth. So here we are. And it did feel very repetitive after a while. Yes. But. Yes. Repetitive and sometimes contrived.

Exactly. Really? That's a use case? Well, yes. Yeah. But it's going to be interesting to put a pin in that one and watch as it matures in the industry for sure. Because it will. And it's going to go into really interesting spaces. So my next question is an odd one.

I always think that MWC has the wacky and wonderful that you find because there's so many people showing so many things and you're just like, oh my gosh, this use case is just lights my creativity off. So I'm going to ask you, did you see anything wacky and wonderful at this year's MWC? I didn't get to see much, unfortunately. I mean, the thing is that I always like to walk the booths just to see what the crazy set of demos are.

This year, I didn't get to walk the booth very much. It was all the meetings and everything was going on. There was something weird I saw, but I can't quite remember it. It was, I think on the last day. And it was one of the Chinese manufacturers has something, some it'll come to me, but, but not, not up to. Okay. I'm going to give you a couple of them. I'll help you out.

So one of them that I saw was physio. They are out of Alicante, Spain, and they're actually doing a Gen AI physician assist, which is not, which is not wacky and wonderful, but what's wacky and wonderful about it is I asked them, are any of you technologists? And they aren't, they're physiotherapists and they're using existing off the shelf LLMs to train a model based on their patient data. And I thought that that was interesting because it tells you the opportunity for innovation across line of business to use these tools without actually knowing how to code and create incredible solutions for the marketplace. So that was one interesting one. The second was one that was about the future of film.

I talked to a woman who has nothing to do with networks. So I'm just going to talk to you about it for a second. Speaking at the event about the future of film and she has a company called artefacto also out of spain that is doing um you know basically the intersect between ai and the creativity of film and i thought that was a really interesting topic what i think the reason why it stayed with me was because the network examples were all so similar that this one was just like, okay, now my mind is going in some interesting spaces.

The third one is the standardization of AI as it relates to 6G. And I'm just going to bridge into our final topic with this one, which is... Sure, sure. I did talk to a couple of people about 6G. We're still debating whether 5G will be fully deployed. And there seems to be a confluence of AI and broad proliferation of AI and networks with 6G for some folks.

There seems to be, you know, oh, yeah, AI is going to be used for 5G, but standardization will come with 6G. That's kind of the line that I got from a few people. So one question for you is, where do you think 6G is? How far off in the horizon should we consider this technology? And do you buy that there's going to be standardization of AI in 6G, or do you think that the genie's out of the bottle and we're going to be having broad proliferation of AI with 5G networks? Yeah, so I think the reality is, so I think a couple of things,

which is that, yeah, 6G is supposed to bring even higher data rates, right? Terabits per second, lower latency, massive connectivity. And you're like, wait, we had that in 5G. Well, next scale up, right?

Auto magnitude type thing. But 6G also has energy efficiency and you're right. There is talk about using AI in 6G standardizing network performance efficiency, and you're right. There is talk about using AI in 6G, standardizing network performance efficiency, intelligent automation, stuff like that. So that's going to happen.

But I think the reality is this, right? Every G is sort of iterative. It builds on the previous one, and the learnings that are coming from the previous generations will feed into the next generation. That's the advantage of all the releases, you know, 18, 19, and so on and so forth. I think the reality is that AI will hit 5G networking

and AI ML will be deployed and will learn. And whatever we'll learn probably will feed into the specifications on the 6G side. I think it's still too early yet to sort of definitively lock things down. And part of that is because AI ML, just because of the amount of investments that we're seeing in this sector,

innovations that drive, you know, like, look, it's every week, literally, right? I mean, two weeks ago, we heard about large language model Gemini Pro 1.5 from Google with a million tokens that you can access. And I just tried it out. I mean, I got access last week and I was dying to try it out.

So over the weekend when I got home, first thing I did was, you know, whip it up and throw in a context of about 400,000 tokens just to see what it could do. And it was pretty impressive, so I can say that. And then this morning, Cloud 3 was announced by Entropic, right? So given that kind of rapid changes and the fact that open source models are coming in and you've got a grok, you know, team,

not the ex-grok from Elon Musk, but the grok with a Q, you know, from the former developers of the TPU, the Tensor Processing Units at Google. If you look at what's happening on the hardware, you know, from the former developers of the TPU, the Tensor Processing Units at Google. If you look at what's happening on the hardware and software side, it's just so rapid. It's just so rapid. And I think the reality is that we're going to have to sort all those things out

before we can standardize on them, just because the rate of change is on, I mean, that's like a daily, weekly basis. And I don't think we will be able to sort this out. So I don't think it's going to slow down for a little while. And I think there's no way you're going to standardize any of that. I mean, the reality is we're going to try to catch up with 5G as it stands. And as we learn about the impact and section between the AI, ML, Gen AI on 5G, that will make its way into the appropriate release definition for 6G.

Thank you so much for being on the program and sharing your insights. I always learn stuff when I talk to you, Roy. So thank you. Well, likewise. I am looking forward to hearing more about your reporting from the network, Edge, and beyond. We didn't even really get into the Edge of this podcast, but we're out of time. But I know that folks who are listening

are going to want to keep following you as well, as well as engaging. Where can they find you? Yeah, so we're in two places. One is abitthink.com, where that's our website. And then the other one is where we publish our multi-vendor reports. That's nextgeninfra.io. So nextgeninfra.io is

where we publish our multi-vendor reports with a media partner that we have. So that's where you can find our reports and you can figure out where we live. All right. Well, thanks so much for being on the show today. It was really great to catch up with you. Well, thank you, Alison. Thanks for the opportunity. Always a pleasure. Thanks for joining the Tech Arena. Subscribe and engage at our website, thetecharena.net. All content is copyright by the Tech Arena. Thank you.