Guy Daniels, TelecomTV (00:07):
Okay everyone, thank you very much. We have our last little session now where I'm delighted to be joined by Andrew Coward, who is the general manager of software networking at IBM. And it's our third insight of the day. But we're going to go off on a bit of a tangent I think, Andrew, aren't we? I think we should. Yeah, I think we should because this is the last session. So first of all, lemme just ask you how are you as IBM, how are you helping telcos with their AI requirements? Let's just get this question first and establish this. What exactly are you doing that makes Telco's AI strategies successful?

Andrew Coward, IBM (00:49):
Well, guy, I'm so glad you asked.

Guy Daniels, TelecomTV (00:52):
It just came to me flash of inspiration.

Andrew Coward, IBM (00:56):
So I joined IBM four years ago with the intent of rebooting IBM's networking business. And since then we have now acquired five different companies. We have seven products in market really going after. I'd see the real key problems or challenges that we see are not just in Telco, but in networking more broadly. The first problem around data collection, which we've spent a lot of time talking about today, how do we get the right data in real time and focus on the right things, namely applications, the things that people use. We haven't spent that much time talking about customers today, but it turns out that they're the ones who complain about applications, not about networks. So can we tie those two things together? So that's been a lot of our focus Secondary, the automation. You heard from Vest a little bit earlier today from a company called Client we acquired back in March this year.

(01:52):
How do we drive automation and make that low code, no-code, easy to consume and just make it very fast as well. Can you solve that problem? And then linking those two things together, of course ai. So working with IBM research, how do we take some of the new models that are being built? For example, I wonder how many people in the room actually know that LLMs are not good for time series data? That's a really interesting thing, right? So we are using LMS for all this. We haven't talked about the fact that most of the data that telcos have is actually time series. So IBM research came out with a small model based on time series called A Tiny Time Mixer back in March. We pushed that into open source. It's been downloaded 500,000 times, so not just a telco focus, but anything time series. It turns out a lot of things that are time series focused. And so what we're doing is taking that and a lot of other technologies around AI and using them together with the data we're collecting together with the automation we're building to drive a complete end-to-end capability to deliver the automation and the orchestration that's necessary.

Guy Daniels, TelecomTV (02:59):
This is December. It is the season of Dickens, and you and I spoke earlier in one of the coffee breaks about how perhaps we are conjuring up the ghost of Christmas past here. I

Andrew Coward, IBM (03:10):
Think we are.

Guy Daniels, TelecomTV (03:11):
Yeah. Do you really think this is coming back to haunt us?

Andrew Coward, IBM (03:15):
Well, I mean we talked about the ghost of Christmas future earlier just a little bit. I mean if you think about the past sessions that we've had in this very room, we have talked a lot about the inability to collect the right kinds of data, the inability to automate, and it's not as if AI is this magic wand that's going to solve all these things. And so I kind of feel like where, I guess the ghost of Christmas past materializes technical debt when they show up on January 1st and that technical debt is that we still need to make sure that we have built the right collection mechanisms. And some telcos have done much better than others on this. I'd say where most telcos have done very poorly is automation. We've had this really big push over the last decade on orchestration.

(04:01):
I have very frustrating conversations with telcos frankly around redoing an RFP that they may have issued four or five years ago because they never made a decision and they never moved it forward. And yes, we've got some new technologies in place to go after those problems, but this is an unmet requirement. You can't have AI orchestrate a network that can't be orchestrated. They're not magic. And so that push for APIs and a, I find A-P-I-F-I, not just the outbound, the kind of public face, but the internals from the cell phone network, the sailor, the backhaul, et cetera, all has to get done.

Guy Daniels, TelecomTV (04:40):
There's another AI discussion going on. Unfortunately it's not here, but there are scores and scores of young enthusiastic developers and it's a massive community if you follow any of them. You've got discussion sites. They're really doing a lot of innovation, real cutting edge. We haven't really spoken about any of the innovation that we're seeing there today in this context. We touched on voice this morning. I mentioned realtime, API. We could have talked about built new and browser based apps and browser based software that's going to kill off apps. And you talk about no-code, low-code, there's an awful lot going on. So why aren't we talking about this? Why aren't we bringing these into telecoms?

Andrew Coward, IBM (05:25):
I think that's very interesting. We mentioned the word revolution a couple of times and I remember that the Gil Scott Heron song, the Revolution will not be televised. And I kind of feel like we're not seeing because we're coming from top down in telco. We're trying to figure out what the strategy is for AI and pushing that down. We are not understanding what's actually happening at the coalface or the code face, if you like. And so much revolution is taking place with how developers within our organizations are actually wanting to code what tools they're finding, what things they're doing. Famously, how many organizations in this room have shut down the use of open AI until you could get your hands on what was going on? I promise you that they off your developers are all off finding their own tool sets and code and methods because they're out there. You plan to rewrite Spotify this weekend?

Guy Daniels, TelecomTV (06:19):
No, I'm just going to rewrite the telecom TV website because the panel's very easy. It'll take me about two minutes and I'll have a fully functioning website. The stacks out there, it's easily accessible on my browser using chromium. It is incredible what can be achieved. And a lot of this is, yeah, it's alright. It's just show and tell some demo stage, early demo stage. But this is going to work its way into commercial products and production. And

Andrew Coward, IBM (06:45):
It is interesting. If you think about IBM's history, we have, I dunno how many billions of lines of cobalt that's still out there and maybe Fortran out there and everything else. So we're being asked by our customers, interestingly, don't convert that to Java or some other language, but can you just document it for us? And so there's this kind of huge process of, I have no idea what this code does. It was written in 1968. Now how do we make it understandable so we can decide what to do with it? So there's some actual baby steps that our customers are taking to use AI in code. And then you've got the kind of full, how do I just rewrite something completely or just give it a few prompts and do that? One of the interesting things we've been doing with the plant technology we acquired is actually with low-code, no-code. Very quickly cycling through how do you tell it to just write something for you? Write me an API that takes a firewall rule and publishes it to five different types of firewalls I have in my network and for it to take not much longer than I just spoke at to actually go deliver that. So it's going from one extreme to the other if you like. And I think different organizations are going to be in different places around how quickly they adopt this.

Guy Daniels, TelecomTV (08:01):
It goes back to what you just said about Spotify because if you Google it, it's on YouTube, there's demos, walkthroughs on YouTube and the prompt is it's a language prompt. It's make me a Spotify club enter and there it is. It's quite

Andrew Coward, IBM (08:14):
Incredible. Exactly it like how much the worry is. Of course when you've done that, if you have to go edit it or make changes and you're not a coder, are we going to create a new set of technical debt around the inability to go essentially maintain these products because the code is non intelligible, right?

Guy Daniels, TelecomTV (08:35):
So Andrew, what should we as an industry be focusing on? What should we be covering? What should we be going into more detail on within AI in the coming year and when we get back together and we will meet again in a year's time for another type an ai, what is it we should be talking about where we be ideally because we don't want to repeat ourselves. We haven't got time to repeat ourselves, right?

Andrew Coward, IBM (09:00):
Well let's talk about the AI models. I think we've kind of been a little, what's the word, bipolar on what AI models, if we have one big one for everything, do we have lots of little ones from different vendors? I think that will shake out somewhat through the next year. We shouldn't discount startups from coming along with models. We haven't talked about startups much today. And if all the innovation is happening in that space, if I think about the growth of my own business, it's because I've been acquiring startups into IBM because they've got the juice if you like. They've got the magic that we need. And so I think that's which models play out. My personal view is that we will need different models for different things and just the idea of saying I need something, it may not be a different model, but it'll certainly be tuned. That's different. Radio versus backhaul versus core versus ip. I think we'll need all of those. And so the agents, as we've been talking about will materialize into how do they resolve for the different elements of the network and then how do they feed up. So again, we haven't talked about it. Hierarchy of agents and hierarchy of Vals will absolutely play out and I think that's going to be interesting.

Guy Daniels, TelecomTV (10:13):
Well, obviously we will have an entire day of AgTech AI next year, so you guarantee that that's a given. But anyone got any other ideas, just get in touch with us please, because Ray and I are always eager to talk to you and listen to what you have to say. For now, I've got a bit of housekeeping, but for now Andrew, stay away, Rob. But thank you very much indeed and go. Okay. So that's all the time we have this year. Thank you also very much for being with us today and contributing. It's been very lively. We do appreciate contributions. As Ray said, we didn't do any of our prearranged questions. They all came from the audience and from the discussions we had during the breaks, we must do it again. Let's move it to the big room next time should we talking to which tomorrow we are moving to the big room.

(11:08):
So if any of you're with us tomorrow, it's the great Telco debate back for, it's the 11th edition. It's I believe 11th edition. So we enter our 10th year of the great Telco debate. Same building, just a different room, the big room. So please join us for that if you can. If can't we are live streaming tomorrow's festivities, so you can catch us online or watch it on demand in a few days time. Now those of you who are coming with us on the big red bus, I have been informed the big red bus is stuck in traffic, which is why I'm talking very slowly, slowly.

(11:48):
Tell us about what you're doing at IBM Andrew. It's on its way and it's going to be arriving in approximately five or 10 minutes time outside, but it will take you five or 10 minutes to pack your bags and go downstairs and make your way out. So it should be there at about quarter past five. If not, just hang around in the freezing cold. Hopefully it's not raining. Don't miss the bus as it says here on my little list. Alternatively, we have cordoned off an area of our local pub, which is just a few minutes away in I think that direction, which is called the Telegram. Telegram. Telegram. There you go. If you're coming with us on the bus, the bus will drop off. Its passengers there at approximately six 30. So if you go there now, you'll miss the rush. So you'll have lots of space. That's it from us. I can't direct this out any longer. So thank you to all of our sponsors because you really make this possible to all our delegates, to our speakers, to all of you. Thank you so much indeed. Thank you.