 And good morning and welcome back to Barcelona. We're here at Mobile World Congress on day four of our fantastic wall-to-wall coverage. I'm joined by my fantastic co-hosts and the co-founders of theCUBE, Dave Vellante and John Furrier. Gentlemen, thank you so much for being here. We have a very special guest with us this morning as well. Andrew from IBM, we had cocktails in a castle this week. We did, didn't we? That was amazing, wasn't it? It really was amazing. How's the rest of the week been for you? It's such a frenetic show. It's so intense. But actually, I'm kind of a little bit frustrated, if I'm honest, about some of the messaging and the things that have kind of been said at the show. We're in this kind of massive hype cycle at the moment around AI. And I want to get to that. What kind of hype cycle could you, I don't know what you're talking about. I'm here to get things off my chest if that's okay. All right, share. Please share. Please share. You're in a safe space. No one's listening. You can say whatever you want. Really? Unfettered opinion. Yeah, so what are you frustrated with? What's the big, what's the big... So, there's this kind of magical one we're waving over the industry, right? Which is AI is going to fix all this stuff that's broken. It's going to radically reduce the cost of operations. It's going to make all our experiences as mobile phone users, smartphone users, so much better. And yeah, I think there's a really big reality gap between the promise that's been told and where we are as an industry and what needs to happen to actually realize that outcome. Let's talk about that, yeah. Yeah, let's step into the gap. We've been on the edge, we're on the gap. So, tell us more about that. What's missing? So, if you think about the simple things that we would want to do with AI that would really change our world experience, right? I want to know why my social media post yesterday didn't make it. Right. Right. Spray my life. It didn't work. Now, you could argue that was because if my content was really poor and it didn't, but more likely, I'd say more likely because I think I like my content. The network was a problem. No bias involved. The network was a problem. So, I would say into the little AI chatbot, right? Why was my Facebook experience terrible yesterday? Now, you think about what's necessary to get to the answer for that. Yeah. Right? And if you ask telcos, they'll say, we've got so much data, we've got, right. So, my simple question, without a chatbot, can you answer that question right now? And the answer is they can. The problem is it takes them about four hours to do it. Right. And they may not have a lot of data. It's a little too much latency there. Yeah, there's a lot of, right, and the reason is, there's a lot of things. Was it the radio? Was it my wifi? Was it the phone? Was actually the connection to the social media site broken at that point in time in my town, which can happen, right? So, there's so many different areas where that could play out. So, that's kind of one, let's park that problem, right? And say, do they have enough data? And the second part of it is, if I want to make it actually have teeth and do things, if I say, I would like to increase capacity to the stadium for FC Barcelona by 20,000 spectators and make sure that all of the public transport's covered. And I want to do that for four hours, right? Again, if that's what I'm typing in, what has to happen today, that might be a three, four week process for most telcos today, because they don't have the orchestration across their systems to go do that, right? And so, that's a ticket that goes to the transport guy, a ticket that goes to the radio guy. So, you can imagine how manual that process really is today, right? And so, Andrew, this is a good point. So, we were just talking on our intro this morning about the one trend at the show, is that you have the telecom, fixed wireless, all the systems are great. It's the IT enterprise integration challenge that's going on right now. So, not let's say in the AI hype, there's a bottleneck around that transition and everyone's working hard to integrate telco into the enterprise and 5G is right front and center. Private 5G is just kind of like not ready for prime time or is it? What's your view on IT integration with telcom? No, I think the IT piece is the piece that everybody hates doing because it's hard. If you ask a telco, go put up another 1,000 towers, they know exactly how to do that, right? If you say, could you make these two systems work together and put an API in front of it? Ah, that's already, yeah. So, a lot of the work at IBM is kind of exactly that, right? It's like, how do we make sure they've got the right data, they've got the right access to the data, they've got the right orchestration tools in place, and so all that prep work that's necessary to make all this stuff come together is the hard work behind the scenes, right? But Andrew, I'm hearing, if I hear it correctly, tell me if I got this right, it's not necessarily a data quality problem, it's facile access to that data and the lack of coherence in a timely manner and it's automation from an orchestration. That's exactly right. And so, getting all these different data sources together is a kind of fundamental. We want to be able to do a kind of mother of all queries, if you like, database queries that go across all these things so that you can answer those questions. Now, the tools that come together for AI are very good at that and making associations, but you have to have all that brought in, right? Is there, when you think about the big data days, when data was just, and when it was, the data quality was a nightmare, but organizations built teams with data pipelines and data engineers and quality engineers and analysts and now they've grown those pipelines and there's a lot of pain there, but the data quality got much, much better. I would imagine the IT shops within Telcos probably did the same thing, but it sounds like there's this other bastion where that process hasn't occurred. Is there an analog here that, does that have to occur where you have to have people rolling up their sleeves and data scientists getting the data in order, or is it the case that the data quality is actually okay, it's just there's not a clear way to get to it? Well, right, I think the data world, I think they have done a lot of work on that, and it's more the union of things, right? So, how do you associate X with Y, and then labeling? So, take a simple thing like an address, you find a field called address, right? Now, if you're not a network guy, you would say that's probably the first line of a street address, you're a network person, of course that's the IP address, right? So, being able to use AI to kind of go in and essentially relabel the data is a use case. So, having, if you like, got the data and reasonably well-formed, we can use AI to, let's say, curate it into a better place and then make it suitable so we can do compares, right? Because if I do compares about addresses and I'm not even talking the same type of address, then we're off to something completely wrong. Well, that's interesting because AI should do that much better than previous, so if you would, previously you would use really mundane mathematical approaches to solve that categorization problem, like support vector machines and semantic indexing and stuff that, yeah, right. But if I can then now apply AI to do that work, that's a game changer. That can do that, right? And to give you an example, in a non-network space, if you run a quick-serve restaurant chain and you want to know how close are my suppliers to all of my quick-serve restaurants and how long will it take them to drive a truck from the warehouse to my store? That's a really interesting problem that, by bringing very disparate data sets together, you can actually start solving. And to my point about my social media posts, that's exactly the same kinds of disparate sources that have to come together to do this. So you're a data guy or you're a networking guy? That's, it was interesting, right? Because IBM obviously has a lot of analytics chops and you've got a lot of tooling here, so I'm curious as to how, you're talking all about data and automation, how do you interact with your data colleagues? It's funny, isn't it? So I've been at IBM for three years. I joined as a network guy. I think I've just absorbed all this data stuff in the process, but. Well, your title is General Manager of Software Networking. Is that, what is that? Is it data networking? Yeah, what does that mean? Yeah, I was just like, well, I mean. Tell us what you do. So that's a really, it's a really interesting world we live in, right? Because if you're going to start a new business as we did and get IBM refocused on networking, you obviously don't want to do and solve problems that already been solved really well, right? So why would you go out and build a new router or a switch or something that's just already out there? Particularly in this world where you point all the data and all the AI and all the programs are in the cloud. So as I joined IBM, we really wanted to solve a problem that we were almost creating for ourselves, which was as we put our customers on the hybrid cloud journey and moved all of their systems and everything into and between the clouds, we found that we broke the network in the process. And the network got broken because it was no longer under the control of the enterprise, right? It was no longer this physical thing that you could buy dark fiber with and put a Cisco or a GINP router in. It was all virtual, right? So really our point in being and focusing on networking business is to bring control back to this very disparate world that we now live in. Every time I'd meet. It feels like that could be a metaphor for a lot of things. It does. The thing is it's a feature, it's a feature, not a bug by the way. So hybrid has a lot of disparate components. You got to go public or across public networks, not going to have end to end control. You're going to need SLAs. You're going to need to have reliability. You do. And it's also about talent, right? So now I need an expert who understands AWS and Azure and Google, and they're all very, very different worlds. And so we want to have a kind of normalized experience for our customers. Whereas they deploy to the cloud, it doesn't matter which cloud they go to, the networking looks the same, the data is managed the same way and the outcomes are the same, right? It's control and visibility. Is that right? And then so where do you get the visibility? What part of IBM or is that a partner ecosystem? Yeah, so since I joined, we've actually done four acquisitions now. A couple of them are focused on visibility at an app level and at the network level. Take us through those. So for example, we acquired a company called Sev1. Sev1 were predominant authority on performance information for how applications actually run across the network. To answer that question, is it the Wi-Fi that's broken? Is it the cloud device? Is it the application itself and be able to, or is it something up in the cloud and be able to put that story together and then essentially democratize that data. One of the things that the heads of networks will tell you whenever a problem happens, they can't solve the problem because the phone keeps ringing and some executives somewhere saying, why can't I do that, so why? So to democratize that, here's all the applications, here's how they're sitting right now. This is the performance of them, and if they're down or they're broken, then this is what we're doing about it. And to have that data in real time, to the earlier conversation, before you get to the AI piece of it, is vastly important. And then secondly, to use some embedded AI technology to be able to predict what's going to happen and what's going to break. So observability sits under your umbrella? Observability of network sits in my team. Yeah, that's a great, yeah. So, Andrew, you told us what you're frustrated about. What are you excited about? Is anything surprised you this week, any customer conversations you're having that are inspiring? I think, you know, telco-wise at the show, the other kind of other topic that's been happening is around APIs. Yeah. And around the API find, is that word, can I use that? There's now. Of the networks themselves. And that to me is a really important step that I think finally we're taking seriously as an industry. Because unless and until we do that, we don't, to my software networking title, we don't have a fully programmable infrastructure in network. I can't call a network and say, I want this amount of bandwidth, I want this type of service and I want it for, so my other example of turn on capacity for 20,000 spectators, well, how am I going to do that for each of the elements that are in the network? It's going to be an API. Right? Absolutely. And so I am more hopeful this year and we've seen a vast amount of progress have been made, announcements and so on. I still think there's a long way to go, but we're starting to see those APIs trickle out in the industry and that's really going to help reset what we do. Well, I want to bring up the API piece because we've been asking ourselves, where are the developers at? So the app, what's the killer app for 5G? That's the question, next question. Two, where are the developers? Are they cloud developers or are they pre-existing developers? So if the telco has more APIs. Yeah. Where, who's building what on what? So how do you guys see that? What's the vision of how you see the app developer market because there's not a lot of DevOps being discussed here. I mean, cloud's just the toe in the water for telcos. It's a good call out there, John. Where is the, where are the developers? Well, wait, so they're out there. Well, let's start with the basics, right? If you think about how the hyperscaders built their infrastructure, they built and consumed APIs for their own use and then they published them and then the developers came. So they could do anything inside their world by using APIs. So all their developers could assemble. So as we've acquired companies inside IBM as an example, one of the most important things that we force acquired companies to do, if you like, is to open up their infrastructure from an API perspective so that I can take those things and use them as components and build new products and new services. Yeah. Right. Customize the ecosystem. It also means I can be wrong about a specific target market, but I can be right broadly, and if I have the APIs, I can adjust and go off and say I want to be in this market. And I'll give you an example, right? We launched the- Or agility, yeah. Right, we launched a global load balancer service in the cloud two weeks ago. And we did that because we said we've got all the APIs, we've got the structure for what we want to do. I want to adjust it this way. Now, to your question around, well, so are the other developers out there? Well, if you think about it when AWS first published a set of APIs and they come use this, was there a developer community? And there wasn't, right? But what happened was, people said, wouldn't it, if it's easier, smarter, cheaper to do it this way, then I'll do it. Well, that's that point. That's kind of the solution for everything else. The developers were out there and they were saying, wow, my choice is build a data center and use the cloud. So boom, they go there. Better mouse track. Okay, they go there. Here, the developers are like, okay, I got Lama open source. So they see him out there, but they're not thinking, oh, I'm going to go program on AT&T or Verizon. They're like, they're not thinking, like, that's my home. So you say, okay, we heard Qualcomm, the same thing, but they're out there. So they're going to build apps. They got to run on something. Yeah, so like. Well, right, but to that AWS analogy, right, build your own network, or go into the cloud and do it, right? It's that. Now go back to my network managers have lost control of their network. If I can now go build myself a global network and infrastructure through the use of APIs of maybe not just one, but multiple telcos around the world. And by the way, we'll come back to that issue in a second. What it means is that I don't have to worry about whether that infrastructure is going to perform. I don't have to go out and form contracts with every single one of those SLAs and then manage them. You're risking yourself essentially, yeah. Because that's right. So that opening up, you know, every enterprise on the planet wants to think that that global network they're renting is under their control and management. And I think that's the destination of where these APIs go because it should enable you to be able to do that. So we're just seeing the bit parts of it. So while I can go to ATT, Verizon and say, I want this circuit with this API and use this amount of bandwidth, that doesn't sound particularly interesting. Go stitch it together globally then. I think it helps. And let me ask you a question. So okay, we love APIs, so we love that. Good comment. So the question is what's the relation in your opinion? What's the relation between APIs and foundation models? LLMs is just large language model, but you've got graphics, multimodals, so we'll say foundation AI models. What's the relationship? Are APIs going to be the plumbing for models? What's your vision on this? I regard the APIs as the teeth actually for AI. Let me explain that. Because they can't actually do anything. They're powerless unless they can get to the data as we discuss or program the infrastructure behind it. So what that means is, in Intelco land, all those chatbots are going to do is simple questions, like why Andrew's bill is $20 more this month, right? That's, I mean, do I need AI to do that? I'm not sure. Versus find out why my experience was terrible, right? And that is how you bring the APIs together with AI, because that's the teeth that they use. APIs are eating AI. Take away from this. API, no, I love that, because it's true. You need the APIs to program the DevOps, program the resources, and it's generative, so things are happening, they're running, they're generating, which is assembling resource or managing resource. Wow, you have had so many fantastic sound bites, Andrew. I'm so glad we got you on the show this morning. We could talk forever, but unfortunately, we're going to have to wrap this one up. Last question for you. What do you hope you can tell us when you're sitting on this lovely desk with us next year? I would like to tell you about all the fantastic customers that are starting to use this technology, the APIs, the AI, and how it's actually coming together, the use cases, and how it's actually worked, and maybe a little bit sadden us about what didn't work. I mean, that's always how much, how we learn, right? Yes, hopefully less frustration for you next time we get the opportunity to chat. Thank you again for being with us. John, Dave, thank you both for being on my slides, and thank you for tuning in from wherever you are consuming this fantastic content. My name is Savannah Peterson, live here in Barcelona at MWC. You're watching theCUBE, the leading source for enterprise tech coverage.