Tony Poulos, TelecomTV (00:16):
Hi, Tony Poulos here at DSP Leaders World Forum in Windsor. We have an extra version of extra shot coming right up and today I have with me Robert Curran, who's a consulting analyst at Appledore Research. Welcome Robert Lee Myall, who's the Chief Executive Officer at Neos Networks. And last, but certainly not least, Udayan Mukherjee who's a senior fellow network and edge group at Intel Corporation. That's a very long title. You must be very important.

Udayan Mukherjee, Intel Corporation (00:47):
I don't know about that.

Tony Poulos, TelecomTV (00:48):
Listen, the first thing that the session we just sat through was talking about the AI native telco leveraging AI through the network. One of the things that came up, one of the first comments that was made by Gabriela, one of the panelists that you were with was AI for networks or networks for ai. Which Lee, would you like to jump in on that one?

Lee Myall, Neos Networks (01:09):
I will jump in on that one. Sort of both really, if you live in my world, so Neos Networks is a fiber network operator, so we sit on both sides of that equation. I think though the point we've all got to stick to with AI is what are we looking to achieve through it? So for me, I think it's got to be about the customer and improving the customer experience. So actually that will cover both of those, both sides of that equation. And I thought it was really interesting in the session earlier about how in that session, in fact the two before that, if I remember it was two before that where panelists were saying think about what you are trying to achieve. And I think customer centricity came through on both occasions. So AI for networks, networks for ai, ultimately cheeses, it sounds AI for customers I think is where we've got to be.

Tony Poulos, TelecomTV (02:03):
Do you want to add to that?

Udayan Mukherjee, Intel Corporation (02:04):
Yeah, there are some difference though. AI for network is how actually going to use AI to improve the network. All you hear that AI native ran others, that's AI for network. So there's layer one, layer two, scheduling, telco, oss, bss, how can I use AI to improve either KPI or return on investment or something? Yeah, network for AI is slightly different. Now doing all this, if you're running some distributed ai, edge versus cloud, how do you achieve the network plays a role there. All the, so ai, you talked about the cluster of AI service networking plays a big role there. So there is the network and that a lot of operators have started thinking, Hey, can I monetize that part because I'm a network guy? So that's the other angle of it.

Tony Poulos, TelecomTV (02:55):
Well there's one more angle is are our networks capable of coping with the extra traffic that's going to be generated by all this AI work that's going around this?

Udayan Mukherjee, Intel Corporation (03:03):
Absolutely as a network for ai, how the network can get scale up to actually address all these AI use cases.

Robert Curran, Appledore Research (03:10):
Tony, can I just add a couple of points onto that? The AI for network and network for ai, it makes it seem like there's a kind of equivalence between these two things, but it's not true because obviously AI for networks who've been doing that for a long time, and again, some of the panelists brought that up. So you talk to some of the networks people and they say, why are you fussing about ai? We've been doing it for 10 years and longer. I think the networks for ai, the nearest kind of comparison I can think of is the way in which networks have become re-engineered for video when we start introduce video into both fixed networks and mobile networks. Massive change in how the networks. So that's really what we're trying to understand as an industry is are we going to see that scale of change, that scale of impact on high networks are being used? And that's where we need to put the context in. So it's not just about messing our prepositions, it's really about asking that question at Lee's point about at the end of the day, what do customers get out of this? So that's got to be the north star for this, but two quite distinct things.

Tony Poulos, TelecomTV (04:07):
So there was another challenge that came out of that was how do you prioritize use cases for ai? And it's okay, we're all talking about AI and it came up in the conversation, there's so many opportunities with ai, where do you start? How do you do the prioritization?

Lee Myall, Neos Networks (04:20):
I think, and we touched on it earlier, I think there's the ensure that you've got the beginnings of a strategy here, the ground plan, but start with what you can actually get off the ground early. And I think there are plenty of early wins that are easy to get off the ground, easy to agree to, easy to budget for in your organization to, earlier we talked about if you are handling structured data either internally or it's inbound maybe from your partners, then that structured data lends itself well to applying some kind of innovation around AI to handle it more effectively, more efficiently. No rehandling, no re-keying. Take errors out of the equation. So have a go at those

Tony Poulos, TelecomTV (05:00):
That's like network nirvana, what you just said.

Lee Myall, Neos Networks (05:03):
Well we're heading for that I think. But the point being I thought earlier was that get going with something you can do tomorrow literally and don't be afraid to give it a go. Make sure that it's not something that'll burn the house down obviously, but it will be containable if it goes a bit wrong whilst you're working on something a lot larger and a lot more ground up here. Just get going, get moving.

Tony Poulos, TelecomTV (05:27):
So give us some examples of that low hanging fruit. Yeah,

Udayan Mukherjee, Intel Corporation (05:30):
So we talked about the power management.

Udayan Mukherjee, Intel Corporation (05:33):
Every multi-core silicon platform has got power stakes. You leave it to OS and OS will do it in that granular level, but AI at the application level can make it more granular. That may saves money because you're not drawing the power when you don't need it. That's a very simple example. That's one thing that all of these ran vendors like Samsung, Ericssons of the world are very much at. And then they hold these customer support tickets, the debugging aspects of it that we talked about. That's another area. So instead of jumping straight to channel estimation, MIMO, where still that debate, am I getting anything better? Is is improving? Am I getting two DB or one db? We don't know that yet. There a lot of work going on, but this customer support, the debugging and power management resources. There are many of those things that you can do.

Tony Poulos, TelecomTV (06:26):
Gabriela made a really interesting comment. I was a bit shaken by, she said, we really need to do more direct deployment of AI use cases rather than dilly ding. So fail quickly, fail short, whatever. But is that realistic? Because the telcos and the CSPs and dsp, we've been very cautious over the years. We don't like to throw something onto the network without testing it. Robert, are you finding that that's the case with ai?

Robert Curran, Appledore Research (06:49):
Yeah, I think TELUS, they want to get the hands dirty on this stuff and so they are being quite, I think, creative in looking at possible use cases. I think to go back to your earlier question, where does it matter when you've got to look at where you're spending time and money and effort already, I think there was obviously the whole kind of customer experience area is an obvious area where you can have some impact and you can see a pretty quick return on that. It's things you're already doing, but you can do them quicker and faster and save money. I think when you get into the networks area, things get a bit more complex, a bit more refined, a bit more. There might be something that happens once a month, but the cost of that thing happening is very high, so it may not be a frequent thing, but the impact is high.

Robert Curran, Appledore Research (07:29):
So the return of using AI to improve it could be very high for you. So you've got to have an understanding of what you're already doing. Where's your time and effort already going in, where are your resources and then what are the critical impacts you're trying to avoid? I think one of the speakers in the last panel made a very interesting point about telcos don't live in a vacuum, that the world outside is moving on around them. So there are cases, plenty cases particularly around security, cybersecurity, where you need AI just to match what's already happening out there in the wider network. That's something you don't control, but ai, various forms of AI can help you to address those issues outside of what you're already doing as a network.

Lee Myall, Neos Networks (08:03):
I think there's another aspect here as well. There's the cultural aspect. There was some mention in the various panels about people getting on board with the change and so on. I think in telcos you've still got a lot of cultural change to drive. On the ops side, it was transformed with digital transformation, DevOps, et cetera. That's been going on for years and that has led to cloud native

Tony Poulos, TelecomTV (08:28):
Yeah.

Lee Myall, Neos Networks (08:28):
AI native and cloud native for that matter in the telco environment. We've got to basically create a more homogenous landscape between technology and operations and bring them together because they've quite often been pretty separate even within the same business. And that's where you need to drive that cultural change, create very flat landscapes, even with data homogeneity, a flat landscape technology and operations and you really create an environment where the AI can do its thing much more pervasively.

Tony Poulos, TelecomTV (09:00):
I think that cultural challenge comes up all the time and it'll take us a while to address it. As telcos have been around a long time, it's a big organization to change around

Lee Myall, Neos Networks (09:09):
Even longer than I've been alive.

Tony Poulos, TelecomTV (09:10):
Exactly. Listen, we've reached it about time. Thank you very much for being with me today.

Lee Myall, Neos Networks (09:14):
Thank you.

Tony Poulos, TelecomTV (09:15):
And for those folks at home that are watching, don't forget our next session coming up is all about the green network. Oh my goodness. Driving extra energy efficiency from network core to edge. You'll be there for sure for that one. I know. Udayan, thank you for being with us and see you for the next version of Extra Shot.