AI and the Mainframe – The Open Mainframe EP 25

Hi everyone. Welcome to the Open Mainframe. I'm Alan Shimel, CEO of Techstrong Group, and we are the, the co-producers of this monthly video series that we call the Open Mainframe.

Our co-producer partners on the Open Mainframe are the good folks from the Open Mainframe Project with the Linux Foundation. And together, we, we produce, as I mentioned, monthly this video series where we talk about relevant topics in the mainframe. A lot of them regarding, you know, open source and, and, and stuff like that.

But really just anything that's relevant around education, technology, topics of interest to our mainframe audience. And that's kind of exactly what we have today. We're gonna be discussing AI and the mainframe, because why not AI is being discussed about everything else.

Before we dive into the topic and, and a little bit more info on that though, I want to introduce you to our great panel for today's event. Uh, I'm gonna introduce them by name and ask them to tell you a little bit about themselves. So let's get started with that, I'm gonna ask, uh, Mike Michael Curry, if you wouldn't mind, Michael, welcome.

And if you wouldn't mind introducing yourself. Thanks. It's a pleasure to be here today.

Appreciate it. Um, my name is Michael Curry. I am the President of Data Modernization Business Unit at Rocket Software.

Uh, we focus on helping companies to modernize their mainframe investments. Uh, I've been, my entire career has really been in the data analytics and AI space, so I've been in this space for a long time. Uh, and AI has come a long way, really even in the last couple of years.

So we're really excited about it and looking forward to this discussion, actually. Absolutely. And what a great thing about this.

We've got a guy working for a mainframe company who's been working on AI and data analytics all these years. Timing is everything in life. Michael, thanks for joining us.

Um, next up I want to introduce you to our Golden Gopher alum here, ke, if I mess up her name, I apologize in advance. Keala Estrada. Moler Klu.

Hi, I'm Kely Estrada. Moler Elia. It's okay.

The third time's gonna be the charm. I know it. I hope so.

So, I'm a content marketing specialist with Broadcom Mainframe Software. I've been in the mainframe ecosystem since about 2017. I started with my hands on mainframe and, uh, covering mainframe trends and technologies as the managing editor for IBM Systems Magazine, mainframe edition.

Um, then I later became senior editor for Tech Channel, and now I'm over at Broadcom Mainframe Software. I've been here over a year, so most of my job is keeping my, my hand on the pulse of the mainframe ecosystem and hearing and learning what trends and technologies are really top of mind for the people and organizations and customers here in the ecosystem. That's great.

Thank you. Our third panel member today is Andrew Sika. And Andrew, if you wouldn't mind saying hello and introducing yourself.

All right. Thanks for having me first of all. Um, so I'm Andrew Sika.

I'm a senior technical staff member at IBM. I've been with IBM 24 years, the entire time I've worked on the mainframe, uh, from ZOS development now to ai. I currently lead what we call our AI on Z Solutions team, which is really focused on driving adoption of AI on the mainframe.

Um, my prior role, I had the pleasure of leading our software development efforts to support the Z 16 integrated Accelerator for ai, right? So bringing, uh, on ship AI acceleration to the marketplace. Uh, so again, uh, nice to meet all of you and, uh, looking forward to the discussion forward.

Andrew, thanks for joining us. And again, we're so lucky to have yet another AI expert on mainframe with 20 plus years of mainframe experience. Thank you.

Um, our last panel members', not really a panel member. He's the co-host along here with me, excuse me. He's also, when he's not doing this, the executive director of the Open Mainframe, uh, group and project at Linux Foundation, as well as an author, my friend John Murdoch.

John, welcome, Welcome, Alan. It's been too long. Yeah, it's probably been a month or more, right?

It, it has been well been, it has been. Well, now I'm excited about this 'cause I've had numerous people coming up to me and, and have been asked about what's going on with mainframe and ai, what opportunities are there? So, boy, this episode could not have come at a better time.

I am super excited to hear what some things are going on Unless you were in the path of a total eclipse. And I don't mean of the heart today. We recorded this and, and we're keeping John from his eclipse.

So we're gonna try time, try to be timely on this. But John, thanks for joining. It's always a pleasure to have you co-hosting with us, Eric.

So guys, I'm in technology 30 plus years. I've seen a lot of water under that bridge. I've seen a lot of hype.

I think we all have. I don't know if I've seen as much hype as we are seeing around ai, but quite frankly, I don't know if we've seen anything that has the potential to be so disruptive, and not just to the mainframe or technology, but to humanity. Or if you believe the hype as we are seeing with ai, everyone wants to know what, how is AI gonna affect this?

How's AI gonna affect that? So let's jump right into it. Andrew, I I'm not gonna pick on you, but Matt, you sound like the perfect person to kick this off here, you know, is is there a role for AI in modern mainframe?

Oh, ab absolutely a million percent. And there's been obviously a huge amount of hype around chat, GPT as you said, and, uh, the generative AI capabilities. But AI is more than just generative capabilities.

It can provide a great way to just help enterprises make better business decisions, especially if you think about the critical workloads that we have on the mainframe, as well as the immense amount of, again, mission critical data on the mainframe. It's just a, a tremendous way to extract additional business value right out of those assets and really help a company be just better innovate all around. Right.

Let, let's yield the onion back on that though. How, how do we do it? I mean, you know, so for instance, with Mainframe, we we're all hearing, you know, Nvidia, Nvidia is a true $2 trillion company.

Companies probably has a higher market cap than IBM or any of the companies aren't here these days, right? And, uh, for, I don't wanna call 'em a one trick pony, but a lot of their money is coming from these ai, you know, GPU processors, um, what, you know, what specific, we don't run as far as I know Nvidia chips on our z systems on our mainframes. So how, how's, how's mainframe, you know, does the, does the AI take place somewhere else?

Or how did the, where's, where's the peanut butter meet my chocolate here? That, that's a great question. And really, when we think about the way that AI works and the role that things like Nvidia accelerators play, especially in Disney industry, they, they obviously, they focus a, a great deal on the training of very complex models as well as now with large language models, right?

Just in general the training, fine tuning and scoring of the, or inference of those models. But when we look at the broader set of use cases, um, what we can see on the mainframe is, first of all, we do have now an on ship inference accelerator, right? That can really accelerate the mathematical operations used in machine learning and deep learning models.

And really what that enables us to do, and what that enables all of you out there to do, is not only to, to potentially train some types of models on the platform, right, on your ZOS or Linux on Z environments, but also to bring those models natively and deploy them right to your mainframe environment, right next to your critical business applications. And now with features like that on Ship Accelerator, we can score, right, or infer using those models at the speeds and feeds that are required by the most complex workloads in the world, right? So the high throughput, low latency, right?

Workloads that are like critical customers tend to run on the platform, and we can support huge amounts of requests per second that single digit millisecond latency for very complex models. So it's really, when we look at ZI think a lot of people out there have the perspective that, you know, it's a, I'll use the bad L word, right? A legacy platform when in fact, right on the platform, we do have the capabilities to support the most advanced workloads and to bring the most advanced analytics and AI right next to them, right?

Yep. So, so kind of dig into that and maybe open up to Michael and Kely as well. What, what are, like, tell me use cases, like what's a use case or two that each of you maybe are seeing that is, is tying mainframe and AI together?

Yeah. Kelly, do you wanna go first and I'll follow you after you? Yeah, you know, honestly, when I think about use cases, I think organizations who are looking to adopt AI within their business, it really depends, right?

And so, um, I wanna kind of challenge that question to say, even when you don't know the specific use cases, AI is one of those trends that is moving so rapidly that even if you don't look to explore and learn about this and explore how to train LLMs and see how they could maybe be applicable within your organization, um, you'll be following immensely behind if you don't start that process even before identifying those key use cases. And so that's kind of my insights there. And then I will toss it over to Michael and see maybe he'll talk about specific use cases.

I wanna come back to a point on yours, but I wanted to hear the other folks talk first. So go ahead. Yeah, so I think it's really interesting.

I think, um, when you think about it specifically on the mainframe, there's two ways to look at this, right? There's a huge amount of data that lives and is produced and is, you know, live on your mainframe and in the mainframe universe that today probably isn't being incorporated in the, in the AI initiatives that are happening in most businesses. It's because it's hard to get to, it's hard to make sure that that flows seamlessly into those systems.

It's, you know, a a lot of that stuff is happening in the cloud, and so getting all of that data synchronized with our, those AI initiatives is difficult. Um, so that's one element of it. So, and there's, that's any use case in the world, but if you think about it, if that's your core transactions running your business, that's the, the data that's your purchase orders and your customer transactions and all the things that are happening in your business in real time, that's the most valuable bus, uh, data within your business.

You wanna be running your AI against that data. And the faster you can run it, the closer to real time that can be, the more valuable your insights can be, and the more effective the actions that you take from that data can become. And that's why it's really interesting when you look at the tellem chip on, on the mainframes to be able to actually run those inferences directly on the mainframe.

The other side though is there's a ton of use cases just for the mainframe itself. So if you look at what's, what's really inhibited companies on the mainframe, it's largely been about skills and, and a little bit about cost, right? There's concerns about cost, there's certainly concerns about skills in a declining skills base.

And, and one of the great things about AI is it can help to bridge the gaps between the, the skills that are required to manage this. And like normal, you know, people that don't have a lot of experience on the mainframe, and also can be used to take a look at all of what's happening on the mainframe and recommend ways to reduce cost. Like, how do I make sure that I am doing the processing in a, in the most efficient way?

Or how do I make sure that I am, um, reorging a database if, uh, if I need to, uh, to improve my, my latency? Those types of things. So there's a lot of things in the mainframe world that can benefit from the application of ai, but when you think about the big opportunity, it's largely about the data that today lives on the mainframe, uh, that can be incorporated more seamlessly into the AI universe.

Yeah, Absolutely. And I can give, uh, I'll just give a few specific use cases that I know we're working on and I've been working on with some major clients, right? Just to give some examples.

And one of the things that my team does is we go and work with clients on, right, basically uncovering and unlocking the use cases they wanna focus on and how to get jump started on them. And we are seeing, like, as you would imagine, a lot of fraud detection, right? Fraud prevention use cases in particular, you know, if you're a, a bank for example, and you're processing your credit card authorizations on the mainframe and you're processing 15, 20,000 transactions per second, and you've got about five milliseconds of latency you can afford to introduce, let's say fraud detection, um, you know, within that transaction, then you're not calling off to the cloud in five milliseconds and getting response back.

That has to be done right on the system, very close to that business workload. And so we see a lot of fraud prevention, anti-money laundering, uh, loan approvals, but then we've also seen a lot of use cases around medical imaging, diagnostics, um, chatbot services, even in some cases, and other use cases you might not associate with, with Z and, you know, Z workloads and Z data. So I, I think the range of use cases, um, that you might see and, and that we've seen certainly is immensely broad, right?

Absolutely. John, you wanted to come back to something? Yeah, I did.

Um, you know, Keio, when you were making a lot of your comments there, um, it gave me a sort of thing of your, you know, you, you have a really unique position of looking at this also from a media perspective, and you have undoubtedly covered a lot of technology cycles, you know, from early to growth to the whatnot. Like when you're looking at some of, like ai, which I mean, I think to be fair, we are still, well, it's existed as a topic of years and years and years. It's, we're really, really just on the early cusp of it.

How is it from like a media, uh, perspective and just sort of like, like a, a communication perspective of like, how do you, like, how is it best to communicate this? How is it best for folks to think about this? Yeah, so I think, and I kind of compare it to other trends I've covered from a media perspective, and AI has actually been a lot different than other trends.

Like, like open source, like hybrid cloud, like sustainability, like quantum computing. AI seems to have evolved a lot more quickly. And I say that because you often, when you hear a trend coming about, right, you hear buzzwords.

And when you start to hear buzzwords online or you see them at conferences, or you see technical sessions at something like share, I keep an eye on all those things and I say, okay, this is something, this is a topic that I'm gonna have to cover later because it's starting to come up more and more. AI comes up with hot buzz terms like chat GPT all the time. I see it off also get, um, conflated like AI and generative ai, and it gets conflated with automation and machine learning, and people sort of lump them together into this one singular category, when in essence, each of those terms is very distinct and different.

So when I cover this from a media perspective, I first look at all the terms surrounding it, the hot topic terms and buzzwords that people associated with it, and the terms that people conflate with it. So my first step is differentiating those terms for my audience. My second step is making sure people fully understand the differences and nuances between them.

And then the question becomes, okay, why should this trend matter from what my organization, from a business perspective, what is this going to do for me? Because there's the technical side of it, the way the technology actually works, and then there's the business side of it. The business benefits that adopting that strategy is going to bring to your organization.

So covering it in the media means understanding the relationship between both sides, speaking with subject matter experts who know the way the, that AI and generat generative AI functions and is evolving and fully understanding what benefits adopting that strategy will offer. It's two separate sides of the coin, but you have to tie the two sides together, otherwise there's no communication at all. Um, and so that's really what I focus on when I write about generative ai, or if we do a video session about generative ai, it needs to bring those two sides together, and it also needs to cover the way the trend evolves over time.

Generative ai, for instance, even in the last year, like if you look at LLMs, we've seen a rapid evolution of open source LLMs instead of just custom l LM models, right? So you have to stay up to date and listen to other people and represent all of the different perspectives surrounding that trend. You know, on that note, If I were to do a survey of, you know, what, what technology stack or what platform is most agile, you know, to, to adopting new technologies, I'm not quite sure the mainframe would be at the top of the list.

And maybe, and you know what? Maybe that's a perception and not a reality, I think for sure. But, you know, we've seen challenges in, in adoption of new technologies in, in the mainframe, um, community when we see new IT trends popping up, right?

Uh, whether it be something like the cloud or, or j you know, a new language or what have you. Um, and it, and it's hard, right? Because the other thing is, guys, there's ai, which has the ability to change our civilization, and then there's things like hybrid cloud that US geeks get all excited about, but you probably don't see on Main Street, right?

And, um, AI is main street. What I'm wondering, have you guys encountered any resistance? Resistance is futile, of course, but any resistance, um, you know, from the community in terms of, is AI gonna be something that the mainframe community embraces and adopts and leverages as part of our modernization?

And I'll throw it to any of you. Uh, if you wanna chip in it, you know, start it off and then if you can chime in. Yeah, I could, I could start.

So I, I mean, I think, you know, in general there is a lot of, um, acceptance of the trend, right? I I, and maybe it's because it's moving faster than the prior trends. And, and I think to your statement about, you know, the innovation happening in a little slower, I'm not sure that's exactly the case, it's just that there's been such a difference in the technology bases that have been, you know, existing in the mainframe versus what's been in the cloud.

And a lot of that innovation has happened in the cloud over the last few years. But if you look at some of the, the changes that have happened in the mainframe to kind keep up with that, it's constantly been making sure that the stuff that's being done on the mainframe could be, can coexist in that hybrid world, right? Across both, both of those, uh, universes and AI specifically, I think there's a lot of hope that it can help address some of the challenges that have existed on the mainframe.

I, I mentioned it before, but the idea of having, um, you know, skills issues where you have the people that understand some of the, um, more complex programming environments in the mainframe or some of the, uh, how your systems work as an example on the mainframe and all the interactions across, um, different systems, that stuff is, is hard, right? And it takes specialized skills that have been learning this stuff for a long time in your specific environment. So what I've found is that most companies see AI as an opportunity to potentially help explain what's actually happening in those systems, as an example.

So being able to look at what's, what's going on within a, um, mainframe system environment, look at all the interactions across those systems, see places where systems might be bogging down. Uh, there might be errors, there might be other opportunities, and be able to use AI to explain, um, all of what's happening and give it, uh, advice or recommendations on how to address, uh, speed issues or, uh, throughput issues, whatever it happens to be. And so I think that acceptance is there because people see the opportunity and, you know, IIBM has done a great job with the mainframe and bringing the tell chip to the mainframe to allow those inferences to happen there so people don't have to move the data off to do the inferencing against those models.

Um, there's a lot of things that have been done and make it more comfortable for that population to really adopt those new technologies. I can also jump in here. Um, you know, Michael said AI offers a lot of hope to solve common challenges in the mainframe ecosystem.

And I think a lot of technology trends we've seen offers that. I think the reason AI has been seen as so mainstream though, is because things that are popular instantly are often quite polarizing to, and I don't mean that in a negative way. I think that AI brings up a lot of ethical dilemmas for people who don't fully understand it, and even for people who do fully understand it, right?

You have to look for ways to avoid, um, sensitivity around bias. When you're looking through input data, you have this underlying fear, well, what if AI replaces my job? Well, no, that's not gonna be the case.

There needs to be a working relationship between human technologist and AI within your organization, within your environment, wherever you put it, right? And so when you have polarizing opinions come up around trends like this, I think it improves and boosts the popularity of what you're talking about too. And that's what I'm seeing a lot more with AI compared to other trends I've covered in the past.

Yeah. Andrew, you want to chime in anything on that, or else we'll move on to the next kind of No, I think the only thing that I'll, that I'll add really is that, um, and both right tr agree completely tremendous really responses. Um, one of the points that was raised was around really the acceptance of AI on the mainframe.

And I think part of, one of the things that we see that's a, a certainly has been a, a slightly different on the mainframe versus other environments is, again, because of the critical workloads on the platform, because of some of the regulated industries, I think there is definitely a much greater focus on right explainability, right? And the other sort of governance pieces that were discussed around ai. So we, we do need to be able to understand why AI is making the decisions it makes, um, whether there's potentially bias weighing in on those decisions.

If it's doing something like, like let's say rejecting a credit card application or the business is making a decision to re reject a credit card application based on ai, right? Those are really, really important points that customers do care about that also helps them, right, start to trust the technology more as well, right? Having it really that explainability factor.

Yeah, it's an interesting, uh, point. I think one of the key things about what we're talking about here is that one of the resistances that has been in place is that people haven't really wanted to move a lot of the data off the mainframe because you, you move it from a place where it's very well governed to a place where maybe it isn't as governed or, you know, it's, you're not exactly sure where it's gonna go. I don't have the same kind of understanding of every step of what happens to it.

So one of the things that's happened with AI is that if, if you have to move all of the data to the cloud as an example and run everything there, and then your results are happening in the cloud, you sort of lost touch of that control point you have around the mainframe. So being able to do your inferencing, you know, train your model sure in the cloud, but actually do your referencing locally and keep that data in the place where it's trusted, uh, without having to worry about some of those things. That's another element of governance in addition to worrying about the, the accuracy of the model and, and how it's come to its conclusions.

So the mainframe becomes the place where you want to do a lot of this stuff. Uh, and I think that's an exciting, uh, opportunity for many companies to be able to actually run things natively there. So, um, here's the money question.

What are some of the AI use cases that are, are starting to already gain traction in the mainframe world? So I think we talked about a couple. Um, I really like the fraud use case.

That was a, that's a fantastic one. I think, um, one of the ones that we're seeing quite a bit is, uh, in the mainframe world there's a lot of, uh, unstructured information that's created and stored, right? If you think about, uh, output management that's coming out of your, you have your transactional systems and you take your off your J pool and you've stored archive, uh, the, the reports that come off of those systems, a lot of times those are customer statements.

Um, and we're finding a lot of banks that are wanting to be able to take their customer statements and enable them for generative ai, um, so that when a customer comes in and they wanna see a history of different transactions in their account or when did their, uh, interest rate change or whatever, they can very easily ask those questions against a corpus of statements that it, that go all the way back and not have to kind of sift through and find the July of 2020 statement and pull it up and sift through it on their own. They could just ask questions against it. So those types of use cases I think are very interesting and not something you might have thought of.

Um, and again, you don't necessarily wanna have to move all that information off into the cloud to do it. You wanna be able to keep that information nice and safe and governed where it is today, um, and, you know, still have the controls around access and the sensitive data that's embedded in those needs to be protected. So doing that natively on the mainframe is, is something that's, uh, really exciting.

We're seeing a lot of opportunity there. Yeah, Absolutely. Good.

I'll, I'll add to that. I think, you know, obviously again, we've mentioned fraud detection and some of those other use cases. I, I think in general, um, one of the places where we start is we're working with clients and, and where we, we see unintended gravitating towards, not surprisingly, are cases where there's a really good return on investment, right?

And so we sort of, when we're meeting with them and we're having these discussions, we're balancing, right, the feasibility aspect of it, meaning is this something that, that, you know, is reasonable to expect that we can do, um, that AI can do and, and that they can understand and explain, you know, uh, up against or, or alongside, right? What's the potential business opportunity, especially now in, in just in today's, you know, market and economy? Um, it's really hard for enterprises to justify new projects, right?

Um, there, there does need to be a real opportunity there. And I think one of the reasons why we've seen such great traction with AI on the platform is that a number of these use cases are really self-justifying. And tho those are just great starting spots.

Of course. I also love how pervasive AI is across different industries. Like I've heard ai, her, you know, discussed as a way to streamline healthcare consumer interactions.

I've heard it talked about for insurance companies to leverage while they're, you know, developing pricing policies. I think that's another reason why this trend is not something that's going away. And Conti is continuously evolving because it applies to all of the industries out there and can help streamline operations regardless of industry.

So maybe kind of pulling this conversation together and kind of narrowing us towards the end of our time here. Each of you are experts in this area. What are, for people watching the show, what are some tips you would have to get started with, uh, AI on the mainframe?

Mike, I can get started if you want. So, um, you know, I, I think you, it always comes down to the business case. I mean, mul multiple of us have said this.

I think we've all said it a little bit, right? It's, it's getting alignment with a business. AI is one of those areas where you have to do that.

Um, it's one of the most important elements. So picking the area that's going to have that business impact, focusing in on the data, um, that is going to make a difference, and that you can get value out of, whether that's improved customer interactions or improved efficiency and, and how you're processing things, or even just improving access to things on, on the mainframe. Um, pick those areas and then work on, now what do I wanna do?

Like what is the, the use case that I want, want to do? And a lot of times you're gonna do the training of a model, um, in the cloud, but always think about being able to run that natively on the mainframe directly using the, the, uh, IBM technology in those chips. I think that's a key element of, um, of the strategy, but start with looking at it through a data lens and, and usually you're gonna find where the opportunities are.

Yeah, that, I think that's good. Good, Andrew? Sure.

I'll, I'll just add a couple things to that because again, that was a tremendous, I think, uh, really response and starting point. If you're somebody that's coming at this from like a, a, let's say a technology perspective or you're, you're just interested in really understanding more about AI capabilities, um, obviously there's a, there's a lot of great resources out there in general, but there's a lot of great resources out there that can show you how you can actually leverage AI on the mainframe. Uh, one of them, my team just made available under the Open Mainframe project under Project Ambitus, and it's something we call solution templates, which are end-to-end walkthroughs that involve you creating simple models using synthetic data and deploying them right on Linux on Z environments or on ZCX, which is another Linux on Z environment that runs on ZOS.

And so there are a lot of assets out there like that that can help you get really jumpstarted very quickly, right? Just on, on mainframe technologies. Um, if you're coming at it from a use case perspective, I'll just add, or from a more of a business and project perspective, I'll just add, again, there are also a lot of resources out there.

Um, certainly, uh, for example, one of the team things that my team does is run a INZ discovery workshops, which, you know, can help you sort of pick out the right use case and kind of figure out just how to get jump started on some of these capabilities. But there's, there's, there truly is a lot out there and I would encourage folks to look at the, the solution templates we just released under the Open Mainframe project. 'cause they're great.

Just general information point, And I think Andrew and Michael really covered the key points here. I'll come up the come at the question, um, from a more generic lens as I often do. I, I think with AI like any other trend, you have to start by asking yourself the big questions.

You know, how does opportunity cost factor into the, your decision of whether or not to adopt AI in your organization, both in the short and in the long term? What challenges are you looking to solve with AI in your organization and how could you go about doing that? Um, if you implement AI in your environment, what data will you share?

Where will it run in your environment? What other data or data sets might, you know, come into contact with this? It all starts with answering those questions within your organization, and from there you work to develop a strategy and figure out how to maintain it in the long term too.

Well, hey, John, I've got one for you. Andrew mentioned that IBM has, uh, I guess donated or, or given over to the Open Mainframe project, uh, uh, some ip what other kind of AI mainframe kind of, uh, projects or work or initiatives are going on at the Open Mainframe project. So there's some very early stage, um, stuff and it's really all came to being of people reaching out and saying, Hey John, I'm doing this in AI and mainframe, um, and I'm like, cool, let's make some connections.

I, I'm seeing sort of two sort of thematics that are coming together. One, and I think, I think Michael, you might have mentioned it along the lines of, um, helping train sort of workforce, um, of helping pull together various educational materials, but also using ai, um, as a way to understand events that are happening on your mainframe system and to be able to provide guidance to, you know, that that same guidance to that 40-year-old or that 40 year mainframe veteran would have, um, and able to derive forward. So I'm seeing that as sort of one trend that we're starting to see some exploration.

And the other we're starting to see is, um, around language models, um, around, uh, various languages are specific to the mainframe. Um, COBOL is, I think a big one. Uh, we're starting to see of, you know, how can we provide guidance to organizations that are working on modernizing their, um, COBOL applications, you know, by looking at the code and identifying, Hey, here's some areas where of improvement, here are some areas where maybe you'd wanna rewrite it into Java, you know, things of that nature, which I know, um, Watson X, um, has done a lot of this on the IBM side and we're starting to see sign of a community coming together.

I, I think honestly, the one biggest area where we see an opportunity to be of value is on the data sets. Because if you don't have good data to train your AI models on, they're not gonna be very good. And, and we've seen this over and over, we've seen this in, you know, in our own realms, um, of a lot of the AI training even from, you know, chat, GBT and other tools have been using very public domain, uh, data.

And it's, it's fraught with issues, but then also there's the question of, well, you know, who came up with this data? Who's saying that this is the right data? There's all sort of governance models.

Um, so there's a lot of work and if, and if folks are interested, um, I can, can help connect. There's some projects that I think are gonna start forming here over the next couple months that are just going to get into that is how can we establish good quality data governed in an open and neutral way such that vendors can build solutions on top of it, um, along with some tooling under the hood to help do validations of these lms, um, to make sure that, you know, they're benchmarked properly. So there's that, that I'm seeing We're gonna see some exciting stuff and it's honestly gonna be things that are gonna accelerate so much of, um, the mainframe industry.

I like everything else. It seems to be touching. There's not a doubt in my mind that it's gonna rott some significant changes and let's all hope for the best.

Yep. Guys, we're, we're about out of time. I, first of all, I wanna thank each of you for coming on and spending a little time talking AI and mainframes with us.

And you know, some of it's real right now. Some of it again, is, is on the horizon, but we're looking forward to it. Um, I'll give you each a chance if you want to, anybody want to have some last words on this and then turn it over to John to close out?

Anyone? No. All right.

John, you wanna take this home? Yeah, let's take it home. I, great panel.

Uh, it's, it's really interesting the concept of AI and mainframe coming together, but I, I think it speaks to this being an area where investment is happening. You know, it's not, you know, how do we work with the past, but it's how do we invest in the future? And, you know, you're seeing here the use cases are here, the expertises here and the opportunities here.

And you know, if, if I look at every other industry I'm working with, Alan is wrestling with AI and how it applies. And I think this industry is in the midst of that too. And I think there's some great opportunities for collaboration so that we can accelerate faster, um, and do some really groundbreaking thing.

So it's exciting we're having this expertise. This is, this is really, really fun. Good stuff.

Alright folks, on behalf of the Open Mainframe Project of the Linux Foundation and Deck Strong and our friends at Rocket Software, who actually are sponsoring this quarter's, uh, open Mainframe show, thank you very much for tuning in. We hope you've enjoyed it. We'll be back next month with another great open mainframe, the video.

Until then, though, take care and be well. Bye-Bye.