Ray Le Maistre, TelecomTV (00:12):
Hello, you are watching telecom tv. I'm Rayla Mara, editorial director at telecom tv. And today's discussion looks at some of the points raised in our recent DSP leaders report, AI's impact on the Radio access network. Now, earlier this year we conducted a survey with the mobile network operator community and asked a series of questions related to the use of AI in the ran. And we received responses from 58 executives and used those results to produce a DSP Leaders report, which can be downloaded for free from the telecom TV website where you can find it in our reports section. And as part of our DSP Leaders programming, we've organized this associated panel discussion to examine some of the key survey results and the issues raised in the report. And I'm delighted to say that joining me today on the program are Viraj Abhayawardhana, director of Mobile Access Strategy at Liberty Global. Paul Miller, chief Technology Officer at Wind River and Francis Haysom, principal Analyst at Appledore Research. Hello everyone. Good to see you all and thanks very much for taking the time to join us today. So let's start by getting some big picture observations about the use of AI in the Ran Viraj, maybe we can start with you. This is a very big topic in mobile operator circles right now. What's the thinking at Liberty Global about the use of AI in the radio access network?

Viraj Abhayawardhana, Liberty Global (01:51):
Hi Ray. Well firstly, thanks for having me. So as you know, Liberty Global, we have a number of different operators across Europe and they're all very different. They're all operating different markets, but of course very, very competitive markets. And what that has allowed these over the years, they have adopted different techniques, different vendors. So the use of AI in particularly the RAN, will have to be suited to the relevant operator. But we at LG as been consider AI as a game changer. So at LG we have set up this central program to drive AI or adoption of ai, not just the ran but across the board across various different elements. But in the RAN per se, we have looked into the whole network lifecycle from planning and design to deployment and commission right down to operations and maintenance. And through that what we have tried to identify is what are the key areas, what are the primary areas where we can use AI now so that it can deliver real benefits?

(03:10):
And there are four areas that we are focusing on. One is network planning, so essentially looking at various different parameters to optimize the planning elements of the network. Then there's a network optimizing where we optimize how you deploy the network elements such as, for example, small cells. Thirdly is energy efficiency. That's like any other operator. That's a primary focus for us and we are deploying many features that can optimize the energy use of energy in the radio access network. And finally is the network monitoring and maintenance. So particularly on the predictive maintenance and predictive AI driven predictive trouble ticketing. So that's one of the four areas that we are really focused on in deploying ai. So from an LG perspective, our mission is to drive thought leadership, bring best practices and share across the op course. So that's what we are focused on right now.

Ray Le Maistre, TelecomTV (04:20):
Okay, thanks Viraj. Great overview of what Liberty Global is doing. So that's the operator perspective. Paul, let's come to you if we can and find out how Wind River is looking at the use of AI in the ran.

Paul Miller, Wind River (04:37):
Thanks Ray, it's great to be here. A fun discussion today talking about ai. Obviously AI is really at an early adopter stage from a technology curve perspective, but we're already starting to see very interesting use cases for it. In the RAN in particular the ability to do dynamic beam forming, dynamic power control, manage energy much more intelligently and efficiently by basically taking advantage of live data from the network and letting the AI and machine learning algorithms dynamically control things is generally viewed as one of the greatest applications for AI As we move forward. Our company though, we're involved more in the infrastructure in the open ran architecture for the RAN when the RAN becomes fully software defined and running on virtual environments and we're seeing a few changes there in open ran. This is a complex multi-vendor environment where you have a hardware layer generally caught server platforms and then a virtualization technology such as provided by Wind River often based on Kubernetes and the like.

(05:38):
And then the application layer comes in and that creates for the service provider a complex environment to integrate and deploy. And when you scale that to tens of thousands of sites as we have in production, it becomes very difficult for a human being to understand and operate such a diverse and distributed architecture. We say AI as a really powerful tool to help assist in the automation of operations within the service provider, whether it be predictive maintenance, predictive outage avoidance, root cause correlation, anomaly detection or even just ease of use. In fact, we have a demonstration on telecom TV advanced AI for open ran that allows you to see how an AI, much like you would if you've used chat GPT and conversed with it, how you can do this similar thing and have an automated system dynamically fire the API calls into the network to collect information and present it to you. So in summary, we're seeing AI used extensively across the radio functions, energy management, beam forming, and as well as the operations challenge that you have for such a complex network. So it's pretty exciting work in front of us for a new technology such as ai.

Ray Le Maistre, TelecomTV (06:53):
Okay, great. Thanks very much Paul and Francis. I guess this is something that Appledore research has been looking at for some time.

Francis Haysom, Appledore Research (07:04):
Absolutely Ray, I think we see AI as actually really a key enabler of what we think it has to be a really important transformation. The route to autonomy and automation. The network requires the AI as Paul just sort of said that it's a whole network problem. The scale is too big, increasingly too big if we want to do that type of autonomy and automation for the wider network. In terms of the question more generally, I think it's a degree of yes and no. I think we are making a lot of progress in terms of use of AI, in terms of supporting the mechanization of a lot of the way that we currently go about running the network, be that root cause analysis, proactive identification of faults, et cetera. We're fundamentally not changing actually the process that we have in the network. And a lot of the really interesting parts of AI can be about this sort of going beyond just the way we do it at the moment.

(08:14):
Ai interestingly enough, visual recognition for example, it really succeeded when we stopped trying to say how AI works based on how we thought human vision worked. And I think a lot of the opportunities in AI is how we use AI to solve a problem. We don't quite understand at the moment a much more whole network solution and tied up in that is a little bit of warning. A lot of what we're talking about particularly in the RAN is very tied to the some self-optimizing networks or self-organizing networks. If you look at that, lots of very powerful ideas that were there at the front, but if we really look at where it's sort of got to, it's very much still a kind of sell site by cell site optimization. It isn't this sort of whole network optimization and that's potentially where we need to be looking in the future.

Ray Le Maistre, TelecomTV (09:11):
Okay. Alright, thanks everyone. Well let's look now at some specific points raised in the report. We asked the mobile operator executives, is it too early to properly understand how AI can impact RAN operations? And that's kind of what Francis was talking about there really and we can see actually that the results as shown in the graphic here that a significant 74% of our mobile operator respondents believe it is not too early to understand the impact of AI on the ran well mean that suggests there's a great deal of confidence in the operator community about AI's capabilities. Paul, do you think they're right to be that confident?

Paul Miller, Wind River (10:02):
I think it's still early days, right? We've seen an incredible maturation lately of this type of technology with things like chat GPT and the Metis LAMA model and all of the industry activity that's happening globally around AI is quite impressive. At the same time it's early to figure out how it will be applied to telecommunications networks. We do have some visibility into that as we've discussed. Whether it be dynamically managing radios with beam forming and power control and network slicing, there are things that we can obviously apply it to. And then of course as I had mentioned on the operation side, helping people run such a complex network can really benefit from ai but it is really early days. So I do think there's confidence and that people understand how to apply the technology but still early days and we'll see it change over the next several years.

Ray Le Maistre, TelecomTV (10:50):
Okay, thanks Paul. And Francis, maybe this does come back to your point and I know you've been looking at virtual RAN and open ran for a while as well. Can we truly understand right now what kind of applications might be running on RAN intelligent controllers in the coming years or do have the operators really figure this out already?

Francis Haysom, Appledore Research (11:15):
I think we're a long way from that at the moment. I think there's a much more fundamental thing. It's very easy to answer a question about ai that's no criticism of the survey I should say, but AI is not a problem that Telco is trying to solve. It has current problems it's trying to solve, which AI may or may not be useful in terms of solving. But I think in terms of, for example the R in OPEN ran as an example, a lot more about what we can achieve there comes down to what does the telco want to do and what does it want to do differently? What are the problems it wants to solve rather than the existing problems it has. It's those new problems or thinking about the problem differently that I think will bring AI to be a much more sort of fundamental shift in the network and may also tie into much more fundamental basis on which things technologies like open ran can actually differentiate themselves from existing forms of ran.

Ray Le Maistre, TelecomTV (12:17):
Okay, yeah, great points. Let's move on. Now a lot of mobile operators are saying they're already putting AI to work in their mobile access networks, but also they're moving and deploying at very different paces. So we asked our respondents by when do you think AI will positively impact RAN operations for most mobile network operators? And the key to this question is the reference to most mobile operators. We weren't specifically asking our respondents about when they'll be impacted. Now as you can see from this graphic, the majority expect that positive impact to be recognized during the next three years, 22% in 2025 and 38% in 2026 to 2027 while 16% of our respondents expect the positive impact to kick in across the mobile operator community from 2028 or later. So the consensus is that it's going to take a while to have a real impact across the board. Francis a year is a very long time in ai. It would seem. Do you think it's going to take that long for the majority of mobile network operators to feel a positive impact from the use of AI in the,

Francis Haysom, Appledore Research (13:45):
The observation I would make is that everybody kind of sees this as a medium term goal and I think that somewhat ties to the sort of, yes it's important, but I'm not quite sure what I'm using it for. I think that the thing, if the danger of being in slightly on repeat, I think it's real business problems that will drive much more immediate ai, but this is a problem I need to solve Now whether it's Paul's, I need more spectrum, I need to be much more efficient at allocating that spectrum. I can't afford spectrum, you're going to deploy AI LX solutions very quickly if that will enable you to be much more proactive in that sort of area. So I think it will ultimately come down to there will be some very specific short-term gains that AI can play a major role in probably longer term in terms of maybe changing the organization, changing the way in which organizations run that's far more likely to be in the medium future.

Ray Le Maistre, TelecomTV (14:50):
Okay, thanks Francis. And Paul, do you think the poll results, the survey results accurately reflect how the impact of AI is going to slowly spread out across the operator community? We often hear from the tier one operators and the group operators, but there's hundreds out there at the end of the day, aren't there?

Paul Miller, Wind River (15:14):
Yeah, I think in a way it's a bit of a fun question for the telecom industry. You correctly noted that a year in the time of AI is an incredibly long duration. But think of the telecom operators and their history for adopting new technology a year is a blink of the eye, right? It takes them often quite an uncomfortably long time to adopt new technology and that has certainly been changing as we've adopted NFV in the core become more software defined there. As we look at open ran at the far edge, obviously the edge is going to be a plagiar, a major location for AI inference engines and this sort of thing. But to see a one to three year response in your survey for either already using or adopting quickly, that's an incredible rate of adoption for the telecom industry. That's incredibly fast. And so I look at that very optimistically that I think that the technology is being looked at very seriously by the service providers and being adopted very quickly.

Ray Le Maistre, TelecomTV (16:10):
And I guess there's a sense with AI that nobody really wants to get left behind. So there seems to be, it's not a panic so much, but certainly the acceleration button on r and d and those associated processes does seem to have been ramped up a bit within the telecom operator community of late. Okay, thanks for those insights. There's also quite a bit of discussion right now about in-house telco AI developments. So we asked our mobile operator respondents from where should mobile network operators source the AI tools that they will use for RAN optimization. And as you can see from the chart, the responses were really spread across the board a real mix between in-house, the traditional RAN vendors and other vendors with quite a few of our respondents. Still unsure about the best source for this viraj to discuss this. Maybe we can start with you this time. What's the liberty global approach to sourcing AI enabled RAN optimization tools? Where are you going to get these kind of capabilities?

Viraj Abhayawardhana, Liberty Global (17:27):
So Francis said about real business problems and for us obviously energy reduction is a critical business objective. So from LG perspective, a lot of our core have adopted what's called managed services. So traditionally they've looked into vendors to network operations and management. They have used self organizing networks as a feature or a capability for a number of years now. So it was natural for us to gravitate towards that as how you can optimize and introduce AI into those modules. So self organizing network, as you know, comes in various forms and various features. There's self configuration such as plug and play and then self configuration, which is about how you optimize the network like load balancing and thirdly self feeling as well. So all of those are in summary, there are a number of features already deployed and they're delivering benefit as we speak now, but we have done some proof of concepts and we've done some trials about how do you enhance that with ai.

(18:49):
I'll give you a couple of examples. So for example, specifically in terms of energy saving, the AI introduction of AI has shown that you can predict more. So for example, you could predict maybe an other in advance about what are the network load that might be and therefore you can take actions much more earlier. And secondly is clustering. So if you have a certain policy of let's say shutting off carriers, optimizing that cluster is absolutely paramount important because that affects their customer experience. So balancing that and balancing between energy efficiency and customer experience, AI has shown that there is some benefits even now. So there's some really interesting trials and proof of concept we are doing and in one case we are actually even trying to introduce gen AI as well. So that's where our focus has been about what tools that we are going to use for adoption of ai and it seems to be initially primarily through use of so and introduction of AI within that.

Ray Le Maistre, TelecomTV (20:07):
Okay, thanks Viraj and Paul to come to you next. It's natural isn't it, for the mobile operators to engage with their existing RAN suppliers for these kind of capabilities. But as things move on, as architectures, network architectures move on, I guess that's where the engagements begin to change. Would you say that's accurate?

Paul Miller, Wind River (20:37):
Yeah, I think so. I think we'll see AI over time pervasively spread throughout the network. Typically the RAN providers and somebody mentioned the Rick earlier, things like the SMO, these are going to be landing zones for AI technology to help manage the radio infrastructure in the ran. That's natural, but it doesn't stop there. If you think about a telecommunications network deployment, there are hundreds if not thousands of different device systems that are deployed out there. The IMS core, the 5G core, the far edge RAN workloads, the infrastructure that everything runs on, there's tremendous opportunity for AI to hook into all of these systems and help automate and manage them as well as to leverage them for enhanced energy consumption and performance. So I think we'll see it certainly start at the RAN because there's tremendous energy savings and optimizations that can happen there and that will obviously be provided by the RAND providers, but you'll see it pervasively throughout the rest of the network. It's just too powerful a tool not to use there.

Ray Le Maistre, TelecomTV (21:36):
Yeah, sure, absolutely. And Francis, how are you seeing these engagements developed? Do you think? I mean your examination of the open ran sector, obviously you concluded and have continued to conclude that a lot of engagement despite a potentially broader ecosystem is going to come back to a small handful of companies. But how do you feel about the engagement around AI optimization tools?

Francis Haysom, Appledore Research (22:07):
I think all of these are valid ways in which a telco CSP can engage with ai. I think this really comes back to the business problem that the business is trying to solve. And there's three words I would use there are trust, their ownership and their innovation. Your views on each of those three will determine which route you take. If you are looking for ownership, ownership of the whole problem, you are very likely to go to your existing suppliers. They can own the whole RAM problem, they can put it together, they can put an insurance package around it in terms of when things go wrong. And that's a very valid role that many CSPs are looking for in which AI can be very well brought in by those integrated suppliers. If you are looking for trust, you're going to look at different aspects of trust and who do you trust to deliver to you what your history with is. And finally, innovation. If you are looking for the standardized solution that everybody's getting, then go with your network equipment provider. But if you have a very clear view of what you are trying to do in your ran, how you're trying to innovate in your ran, trying to do things differently in the ran, that's your opportunity to either own that problem yourself or to look for best of breed suppliers to supply that for you.

Ray Le Maistre, TelecomTV (23:31):
Great. Thanks very much Francis. So we're coming to the end of our discussion now, but let's end with a key takeaway from each of our expert panelists. Francis, we'll come to you first. What's your key takeaway either from the report or from today's discussions?

Francis Haysom, Appledore Research (23:51):
I think we're very much still at the beginning of the use of AI in the ran. I think if you looked at, it was quite interesting if you looked at, there was actually quite a small number of providers that were saying, I'm really, this is absolutely critical for me. And there were quite a few quite important responses there. And if to take my Brit hat on is if somebody says it's quite interesting, that doesn't mean it's really interesting. So I think we're still at an early stage. I think we've got to be slightly skeptical about this is a magic wand and I will reiterate it still needs it's, AI is not the most important thing. It's what do telcos, CSPs want to do as businesses that will drive the adoption. It will make things happen very quickly or it will make things happen very slowly.

Ray Le Maistre, TelecomTV (24:51):
Okay, thank you Francis. Paul, what's your big takeaway from this report, from this discussion?

Paul Miller, Wind River (24:58):
Well, I think it was a really excellent survey and there's a couple of interesting data points in there that we didn't discuss that really struck me as being interesting. One is the applicability of AI to open ran. Obviously open ran as a technology for RAN is very interesting, but really what it is, is the deployment of a virtual cloud throughout the entire service providers network. What that means is obviously an easier way to deploy AI functions. If you have bespoke hardware implementations in your network and you look to deploy ai, especially at the far edge, that becomes quite challenging. So the software defined environment is one that we have to pursue if we want to take advantage of ai. The other thing I thought interesting in the survey was the discussion about the far edge. And we've seen AI used both in the core and the edge.

(25:46):
As I had mentioned before on telecom tv, we have a demonstration advanced AI automation for open ran. That's really a core function where we take a large language model and have it learn the APIs and help manage the network from a core central location. But really the intelligent edge is where the large volume of AI is going to be deployed, and that's instead of a learning function, that's an inference function for ai. This is where the action happens to collect data and act upon it. And this is where all the RAN sites are. So I think it's going to be really interesting to see the results of this survey kind of point to the future as to where this is going to happen in a software-defined way and where it will happen in the network architecture in the core versus the edge. So really excellent survey data.

Ray Le Maistre, TelecomTV (26:29):
Okay, great. Thanks very much. Paul. Viraj, final word to you, what's your key takeaway for our telecom TV audience?

Viraj Abhayawardhana, Liberty Global (26:39):
So i echo Francis's comments there. So we are barely scratching the surface and the early results has been extremely promising, but we just have to identify which type of AI is good for and where. So that's the focus from us is identifying what are the primary areas and what are the ones that I give the best return on investment. So that's what LG is focused on.

Ray Le Maistre, TelecomTV (27:08):
Okay, excellent. Well great. Thanks everybody. Well that's the end of today's DSP Leaders Report panel discussion. Remember the report we've been discussing, the impact of AI on the radio access network is available for anyone to download right now from the report section of the telecom TV website. And I urge you to check it out because as Paul mentioned, there's a lot in that report that we haven't had time to discuss in detail today. I'd like to thank our industry expert speakers again for joining us today and sharing their insights and to you for watching. Thank you and goodbye.