MWC25 | Andrew Coward, IBM

Clips
More from MWC25

Andrew Coward

GM, Software Defined Networking

IBM

play_circle_outline Navigating the Varied AI Landscape: Engine Models, Real Products, and Comprehensive Solutions

play_circle_outline Maximizing Efficiency: Real-Time AI Applications and Small Models in Network Operations

play_circle_outline Maximizing AI Model Accuracy: Time-Series Data and Multiple Sources for Contextual Understanding

play_circle_outline The complex nature of network infrastructure and the need for tailored solutions

play_circle_outline Maximizing Growth: Acquisition Tactics for Low-Code Firms & AI Integration in Telco Networks

Info
Transcript

Andrew Coward, IBM

Andrew Coward

GM, Software Defined Networking IBM

Andrew Coward, General Manager of Software Networking at IBM, joins theCUBE's Savannah Peterson on the final day of MWC25 Barcelona for an insightful discussion on advancements in AI within networking. Coward shares his expertise in tackling industry challenges, such as enhancing closed-loop automation and transitioning from GUI to sophisticated AI solutions. Analysts from theCUBE, including Bob LaLiberte and Dave Vellante, contribute to the engaging dialogue, ensuring a deep dive into these transformative technologies.

Key takeaways from the conversat... Read more

explore Keep Exploring

What are the different approaches companies are taking when it comes to AI implementation? add

What is IBM excited about in terms of artificial intelligence models? add

What are some new models, such as Tiny Time Mixer, that can be used with time-sequenced and time-stamped data to make predictions in various fields? add

What is the challenge with the increasing complexity of technology and the need for customization in today's market? add

What is the reason for attending this show? add

bolt Powered by CUBE AI

Andrew Coward, IBM

search

Savannah Peterson

>> Good afternoon networking fam and welcome back to Barcelona, Spain. We are here for our last segment of our four days of live coverage here on theCUBE. My name is Savannah Peterson. Very excited to be here with Cube Alumni and VIP, Andrew from IBM today. Andrew, thanks for coming to hang out.

>> Love to and the last show of the day. I mean, it's a great honor, it is.

Savannah Peterson

>> Well, and you are a great guest to have in this slot. This is becoming a tradition for you and I. We sat down last year as well and I am excited to catch up where we are now and where we're going next because I know you've always got some hot takes for me. Last year, you and I talked about the over-hyped state of AI, and I think we had both felt like there was little AI stickers put on the booths all over the place kind of at the last minute, just to stay relevant in the conversation. How is it different now this year?

>> Yeah, I remember coming to the show really frustrated. It was like-

Savannah Peterson

>> Right. I felt the same way....

>> this is not good. I mean, we're seeing products this year. It's great. It's great to watch, and I've been kind of trying to categorize what you see at the show in terms of AI and try to bucketize it, to kind of makes sense of it. And I think it's kind really interesting. There's groups of companies here, cloud providers, people, including IBM, providing models, and I think a model as an engine. If I give you an engine, you're going to have to build your own car. You're have to bring data to it. You're going to have to make it ... That's one bucket. I think that's kind of interesting, and a lot of companies are doing that. They're building their own solutions, they're training the models and so on. There's a second bucket, which is companies that are doing these kind of vertically integrated stacks. They make base stations or routers and switches, and they're saying, "Buy AI in my complete stack." And there's a third set of vendors, and a much smaller set, who are trying to build a horizontal solution, and I put IBM in this bucket too, where we're trying to make AI work across a network and not just in one vertical stack or in a place where you have to build your own car. So those vendors are kind of building cars for the customers that you can buy and then go deploy them and not have to worry about the training or the-

Savannah Peterson

>> Go drive them on multiple tracks, essentially....

>> or hallucinations or anything like that, because all been sorted out. So that's where I think the industry needs to get to. I would say that at IBM, obviously, but it's an important thing of actually solving the serious problems that sit behind it instead of saying, "Well, that's up to the customer to sort out," right?

Savannah Peterson

>> Yeah. I can imagine these are some pretty complex problems. Are there any applications that you've seen that are in the wild, actually working right now, that you're particularly excited about?

>> Well, in the wild, yes. Excited about? Uh.

Savannah Peterson

>> Please elaborate. Let's hear it.

>> A lot of the things you see around here in the show, a chat interface has been placed on a GUI interface. And so what that means is you can use the AI-

Savannah Peterson

>> I'm a little fatigued with you on this one, but yeah, keep going. I'm less impressed. Let's put it that way.

Savannah Peterson

>> Exactly. And it's kind of cool. You can call up a report, you can do this and this, but we didn't invent AI to replace the ... well, we didn't only invent AI to replace the GUI interface with yet just another way of doing things, right? That's not what it's about. So the excitement, I think, is what is probably the next phase of it is, as the AI gets used in real-time to get what the industry needs, which is closed loop automation, remediation, prediction, all of the things that will drive the efficiency out of it. Now, I'm not saying that a little GUI typing into a chat bot isn't going to get some efficiency, but that's maybe 5%, not the 90% that we want to get to.

Savannah Peterson

>> Well, and it's not the most impressive form factor, honestly. I mean, the analogy I keep using that I got from an old boss of mine was when we got the internet, the first thing we did was write letters, but we called it email. And I feel like that your example with the GUI there or with the chat bot, it's kind of the same thing. We have AI and now we're kind of just doing the same thing we've always done rather than solving new problems and taking on different challenges. Yes, I mean, I know companies are thinking about that. You're probably seeing a lot of that.

>> And then the other thing we're excited about at IBM is small models, not big models.

Savannah Peterson

>> Yes. Louder for the folks in the back, please.

>> I know it's-

Savannah Peterson

>> It's not all about LLMs.

>> Right. And LLMs, yeah, they're not a fix-all for all of this, right?

Savannah Peterson

>> No.

>> And there's no reason why an LLM for a telco should be able to give you a plan about how to launch a moon mission. This is not a thing for us, right?

Savannah Peterson

>> Yeah. We don't need to over-engineer it, essentially. And you need to efficiently engineer it or build it out and train it and do everything.

>> That's exactly right. So, for example, many people don't understand that LLMs really don't understand time-series data because it doesn't understand the difference between what happened yesterday versus what happened 10 years ago. And almost all data that we get out of networks is time-sequenced and time-stamped. That's important to us as telco folks, right?

Savannah Peterson

>> Yeah.

>> So there's new models and IBM's built one called Tiny Time Mixer. There are others. Very small parameter, million parameters, and they understand time. And what that means, importantly, is you can use them to make predictions, like what happens next? So the excitement for us is where we can take network data and then we can apply it to weather information or TV schedule, and then we can make predictions about what's likely to happen, or we can understand why something happened in a larger context of more than just this data set. That's kind of fascinating to me.

Savannah Peterson

>> And what's the advantage of being able to do that real-time and faster?

>> Let's give you an example. So if you think about what happens during a Super Bowl is a good example.

Savannah Peterson

>> Yeah.

>> Half-time, people stop streaming. They go make a cup of coffee, pour themselves another drink, go get some snacks. The streaming, the number of streams just drops off precipitously for that period of time. Now, if you're looking at this network going, "Why has the traffic dropped so much? There must be a problem." It's a false positive. If you have the TV schedule-

Savannah Peterson

>> Outlier, obviously. Yeah, yeah.

>> That's right. So it's that predictive thing that stops things happening. Then the other thing we're thinking about is take the weather information. Well, as you see a storm coming in, what's it going to do to your network? Are you going to start moving things around? Are there things that you can predict in the last three storms happened that you want to mitigate as a consequence of that and how those get predicted and built into the model? So it's very interesting how multiple data sources, time series data sources, then get brought into this.

Savannah Peterson

>> Well, it adds context.

>> It does.

Savannah Peterson

>> It's essentially giving you another access of information.

>> That's right.

Savannah Peterson

>> To a degree. I see it as a little Z axis. When you're talking to customers and you're out there with your community, do you feel like folks are a little LLM crazy or are they very open when you start talking about the impact small models can have?

>> I think people have realized the cost of running large models and the cost of training large models-

Savannah Peterson

>> Yes, actually, outstanding point, Andrew. Outstanding point.

>> It's like isn't there a cheaper way of doing this? And what we are seeing is the democratization of AI. It's almost free to put data in and run it against AI model, but if you need to train it, that's the expensive bit. Well, now the training piece is coming down massively in costs, and that's why I go back to the engine versus car analogy. Why are we expecting telcos to build their own cars for this? It's not like we ask them to build their own radios or build their own routers.

Savannah Peterson

>> And they don't need to, and that's why partners like you, I can imagine, are so critical.

>> That's right. Provide the car. You might say there's a choice of engines. You want to use IBM's Granite, do you want to use Llama? Do you want to use one of the other ones? So you might just like a car manufacturer, right? I'm going to run with this analogy.

Savannah Peterson

>> Please.

>> We have different colors, right?

Savannah Peterson

>> I'm here for it. I'm here for it. Let's go, Andrew.

>> So, I mean, different colors of car and then different sophistications of the car depending on your network. And, of course, there might be things that you might want to do that are unique. Almost every telco infrastructure is very special, and that's just the nature of it. But one of the things that I think we are really, really focused on is applying to the complexity of networks. With every new generation, with every G, with every three ... I call it a Noah's Ark problem. They've got two of everything. Two vendors in every single place. It's also why the vertical stack vendors and the folks who are adding AI to that, that's just actually making the world more complicated, not more simple. Because now if you want to look at, say, a radio vendor, you've got to use their AI stack to understand that. That's a completely different world to the one that's for the next vendor in radio or in the router network or something.

Savannah Peterson

>> Right. It's not a one-size-fits-all.

Savannah Peterson

>> So it doesn't help with the ... It's not one-size-fits-all, but it also doesn't help with the remediation of problems across the network, because you haven't got that simplicity. So being able to take those models and then build a car, but also enable it to go to different places depending on the customer, that's really what we're delivering at IBM.

Savannah Peterson

>> I love that. This car analogy is going to serve us well for a long time. Well, I look forward to when I have you on next year and we bring back the car analogy and you can tell me exactly where we're driving.

>> Exactly.

Savannah Peterson

>> So talk to me a little bit about how you manage that complexity from a solutions perspective. Like you said, if all these telcos are different and unique, is it bespoke every time? Are you using some of the same engine components? What's that like?

>> Well, yeah. First of all, it's no secret. One of the reasons I love coming to this show is that I'm looking for the next company to acquire.

Savannah Peterson

>> Yes. Ooh. Yeah.

>> And in the network space, we've now done five acquisitions with IBM. The last one we did was a company called Pliant, it was just after Mobile Congress last year. And one of the reasons we acquired that company was a low-code, no-code framework for working with many, many different vendors and driving the automation to that. So meaning that you could basically use the technology with something like 2000 different products that were already integrated and we can add the next one within a couple of day windows. So what that's meant is, we've given the AI teeth to actually go make configuration changes and actually do stuff within the network infrastructure. Without that, AI is kind of pretty useless if it just tells you you can do stuff and doesn't actually go do it. If you allow me with another analogy, we had the car analogy, I have a genie analogy for you, which I-

Savannah Peterson

>> Let's go.

>> So, I mean, a lot of people think about AI as being magic, and so think about that as a genie. Now, if you were to meet a genie and the genie offers to do things for you, there's a couple of qualifying questions you might want to ask your genie first. Well, there're two things. First of all, is the genie omniscient? Does it understand everything about your network or the world in order to be able to grant your wish? That's the kind of first thing. And the second one, of course, is it has to be omnipotent. It has to be able to have the power to actually go make that change. So just by saying, okay, we think you should do these types of things, maybe you should want to be typing this stuff into your ... I mean, that just doesn't work. So by taking the omniscience on the omnipotence, you give the genie power. And if you don't have that, you do not have a genie, you just have a malevolent spirit.

Savannah Peterson

>> Or a hallucination.

>> Or hallucination.

Savannah Peterson

>> Maybe pun intended. But yeah, I like that analogy. I was mentioning in one of our other interviews earlier this week, feels like there's ... you mentioned magic. It feels like there's magic fairy dust, little IBM magic fairy dust all throughout the show floor. You can see your partners. You can see your collaborations. You can see a lot of things running with our magic analogy, now that we're here. If you could wave your magic genie wand and all of a sudden the market that you're trying to reach or the companies you might be looking at to acquire, knew something and were able to change something, what would that be?

>> Yeah, I don't know. I think for us it's the ability to actually get to the end problem of solving automation end-to-end. Customers are going to our stand and they're saying, "I want to save 30% of my operational costs. How do I do that? Where do I go with this?" And you kind go through the use cases, reducing the time it takes to resolve an incident. IBM's customers do not like being on the front page of the Wall Street Journal because there's been an outage.

Savannah Peterson

>> I'm shocked.

>> I mean, that's career ending, right?

Savannah Peterson

>> Correct.

>> And so getting that resolution down to seconds or minutes, versus hours or, in some cases, days, I mean, that's a massive credibility thing for our customers, plus-

Savannah Peterson

>> Absolutely. I mean, they need to hold that trust with their customers.

>> And so I think that's the destination. That's my wish. Not just a wish. It's what we're building, too. And then the second thing is, can you increase the timing between these incidents to a much bigger level so that your predictive engines, the thing that you saying is ... One of the things, for example, we found, we're starting to predict when hardware's going to fail in the network.

Savannah Peterson

>> That is so cool and critical. Okay, so how are you doing that?

>> And we didn't expect ... that wasn't one of the things we looked at.

Savannah Peterson

>> But it's such an important choke point.

>> It's about the interception of data. So we see, for example, packet drops or we see the DB level of signal strength go down, and then we look at the line voltage of the power supplies and we see variations. And then when you look at the historic data and you make predictions on the historic data and say, "Did this actually happen?" You start to understand that you're getting failures. Now, that creates a really interesting challenge, which you've never had before, which is do you send that card, that line card that's about to fail back to supplier and say it's about to fail? Because they'll test it and say, "No, it's perfectly fine." Are we creating a new supply chain as a consequence of this, where you've got just-in-time supply parts that are being delivered, because you can predict when things are going to fail? I think that's how AI is kind of changing the industry when we start to affect that.

Savannah Peterson

>> It sounds like a simple thing just in terms of the complexity of what you just said, but it's actually a really big thing and a critical thing. There's so much data. I mean, I'm preaching to the choir, I'm sure, with you, but in case the audience isn't aware. I mean, device failure, across the board within these systems, it can ruin things. It can pull things down. You don't know where it is. You don't know what's busted. To be able to prevent that from happening, ever, gives you such a consistent experience.

>> Right. Right. And, again, the operational cost savings, the 30% that that customer want to get back. I want to get back to point you made is the why question, right? It's funny, because we're doing these demos all week, and as soon as we put the AI up into the magic point, it's like how did it know that? Most people that come here are deeply technical and quite a lot of them are deep radio engineers. And it's like, well, we took this KPI and we took this KPI and we took this KPI and this is what we saw and this is how we ... And they suddenly realize that actually building that car themselves is too hard because there's just too much, there's too many different models, too many different things you have to put together for that, right? And I think that's the kind of a hard realization that you just don't take all this data out of a big LLM and hope for the best.

Savannah Peterson

>> No, absolutely not. And I think that's a really important point. I got to ask you, have you found any companies out here on the show floor that you're curious about acquiring, perhaps, since you mentioned that?

>> Yeah. I mean, obviously, everybody comes here, so it's a great place. I mean, for us, I think a lot of it's around the types of data that we're bringing in to process and manage, and there's so many things we have to think about. Configuration as code. So when a device is configured, there's code that's there, but it's now important in the AI model. The data that ... all of the different sources of data. So I mentioned about bringing weather in, and so, for example, with our own products, we actually see internet weather. Internet weather is a thing.

Savannah Peterson

>> Okay, you got to explain that. What is internet weather?

>> So which telco in which city with which content distribution network is not looking great right now? Where is it raining, if you like?

Savannah Peterson

>> Yeah.

>> And so this is really important because-

Savannah Peterson

>> Interesting....

>> if you are broadcasting a sports event, you might choose to use a different content distribution network to push that content out toward if you know that it's problematic in one area. So having that sensing information then becomes really important. And if we see that across streaming, media, gaming, all those things, then we've got that additional input. It's where it really comes to play, right? I've sat in call centers of telcos. It's fascinating. And things have changed more recently, but I've seen phone calls coming with customers saying, "I can't post on Facebook right now." And the person on the phone pulls out their phone, pulls up Facebook and says, "Well, it works for me right now."

Savannah Peterson

>> Classic.

>> So that is not what we should be doing, right?

Savannah Peterson

>> No, no. Yeah, it sounds like a really good QA process there.

>> Exactly. So having the tools to actually map an application and understand, well, actually we know that this particular application is not doing great in Chicago right now, and then-

Savannah Peterson

>> It's a little snowy in the internet weather in Chicago right now.

Savannah Peterson

>> So that gives us an understanding, maybe it's not my network and maybe it's something external and those become important attributes.

Savannah Peterson

>> It gives ... what I'm hearing throughout this entire conversation is we're really driving towards a place of visibility, a place of the ability to predict what's going to happen next and to have more cohesive, end-to-end experiences without interruption?

>> That's right. And that's where we're driving to, right? I heard-

Savannah Peterson

>> That's where the car's headed?

>> Right. If you remember the show Silicon Valley, the very last episode-

Savannah Peterson

>> I actually live in the show every day of my life, but yes.

>> Well, the very last one was set in ... it looked like Burning Man. I don't think it was supposed to be Burning Man. And they built the network for the event and it was failing. They were going to go home the next day. And overnight the network had reconfigured itself and it worked perfectly, and they were terrified that they'd now got something that they didn't understand how it worked.

Savannah Peterson

>> Right. Oh yeah, yeah, yeah. It went ahead and did something.

>> It was a brilliant kind of end to the series. So as we think about AI from an IBM perspective, understanding the why and the responsibility that comes with it to be able to say, "We know this happened for these reasons, and this is the answer," right? There is no magic in it. There can be no magic in it. It has to be explainable all the way through.

Savannah Peterson

>> Yes, it can feel magic, but has an explanation and-

>> Exactly.

Savannah Peterson

>> Yeah. And has a process that's repeatable, or at least you're able to figure out. I wasn't expecting us to talk about Burning Man. Have you ever been to Burning Man?

>> I haven't. It sounds like an amazing event.

Savannah Peterson

>> I'm a big burner. Shout out to my Burning Man family just because this is a unique opportunity. All right, Andrew, since this is our tradition now to have this chat, at least, here once a year, I'm curious to ask you, what do you hope to be able to say next year that you can't yet say today?

>> I want to be talking about the products that we've built and we've launched through the course of the year, and that's what's exciting. It's actually, to the frustration I had last year, the frustration I have now is I've got a lot of stuff that they want to talk about and share, and we're just round the corner from being able to do that. So that's what I hope to be able to be able to .

Savannah Peterson

>> Well, I can't wait for you to share it all with us next time. Thank you for always making the time for us.

>> I love being on the show and I want to continue tradition. It's amazing. Thank you so much.

Savannah Peterson

>> I know. Yes, but we'll definitely continue this tradition. We're having a little bit too much fun not to. Thank all of you for tuning in to our four days of live coverage here. Got to give a real big shout out right now to our fabulous production team, the Henderson Brothers, to Andrew, to Ken, to Frank, who's doing the sales, to Bob LaLiberte and Dave Vellante who are on this desk as fellow analysts with me, to all of the AR folks out there who curate these fantastic guests for us. I am so grateful. I feel privileged to learn from the smartest people on earth, and I hope that you all have enjoyed this as much as we have. Signing off for the last time from Barcelona, Spain, in 2025, at Mobile World Congress. My name's Savannah Peterson. You're watching theCUBE, the leading source for enterprise tech news.