HPE Discover 2024 | Red Hat and HPE: Fueling customer innovation with AI, virtualization, and hybrid cloud solutions

Clips
News
More from HPE Discover 2024

Ryan King

Senior Director, Hardware Partners

Red Hat

Hybrid cloud AI solutions take center stage in Red Hat and HPE’s strategic collaboration

Hybrid cloud AI solutions are becoming essential as business leaders navigate the rapidly changing landscape of technology and market demands.Red Hat Inc., in partnership with Hewlett Packard Enterprise Co., is driving innovation in virtualization and artificial intelligence. The companies’ efforts are helping organizations transition smoothly into the next era of technology.Red Hat’s Ryan King talks about hybrid cloud AI solutions.“GreenLake bringing pay-as-you-go into the context of the private cloud space and hybrid cloud is a great enabler for our platform,” said Ryan King (pictured), senior director of the global hardware partner ecosystem at Red Hat

play_circle_outline Red Hat's AI strategy and offerings, including OpenShift and RELL AI.

play_circle_outline Integration with HPE Green Lake for hybrid cloud adoption and consumption model benefits.

play_circle_outline Partnership with NVIDIA and Intel for AI capabilities and optimization.

play_circle_outline Introduction of Image Mode and InstructLab for containerized OS delivery and democratizing AI.

Info

Red Hat and HPE: Fueling customer innovation with AI, virtualization, and hybrid cloud solutions

Ryan King

Senior Director, Hardware Partners Red Hat

search

Rebecca Knight

>> Good morning everyone, and welcome back to The Cube's live coverage of HPE Discover 2024. I'm your host, Rebecca Knight, sitting alongside my co-host and co-analyst, Dave Volante. Dave, this has been a great show. We're about to welcome back an illustrious Cube alum.

>> Yes. Well, third year in a row for having Red Hat on. Red Hat is smoking hot. They're red, they're so hot.

Rebecca Knight

>> Red-hot. They're red-hot. Exactly. I know. We can rename them right now. I'd like to welcome back to the show, Ryan King, senior director Global Hardware Partner Ecosystems at Red Hat. Thanks so much for returning to the queue.

Ryan King

>> Yeah, thank you. Yeah, I think three years ago when I did this the first time, it was quite the experience. I was fresh into the role. My background, I've always been in this business and this industry, but about the 2017 era, I started to look into what was happening with generative AI and it really helped bootstrap the relationship we had with NVIDIA. And so that got me on the radar to say, hey, what should be our broader strategy for infrastructure and infrastructure partners? And that's how I stepped into the role. And one of my first forays was with HPE around Green Lake. So it's great to come back and talk about that journey, but also the additional focus on AI and vert, which seems to be the two topics that everybody is having deep strategic planning around.

>> Oh, absolutely. And Red Hat is so many places between OpenShift and RELL, the entire open source software, what you've done within Instruct Lab, I'd love to get your point of view on that, but hit the escape key for us and help us understand AI's just broadly point of view, or Red Hat's broadly, their point of view on AI. What's the strategy? How should we think about it?

Ryan King

>> Yeah, so of course everything's founded in open and open source with Red Hat, and it all builds up from what we do with RELL and above that with OpenShift. But of course the portfolio part of it, everybody knows OpenShift and now OpenShift AI as being two forays as platforms that we're offering into this space. Also, they may have heard it Red Hat Summit, something called Light Speed that brings generative AI into the experience of using our products. And so you're seeing that being placed to use AI to use our products and platforms within it. So it's AI within, but also AI as a platform. Our point of view is we've had a very successful journey. I think you've seen the analyst reports that have us in the very far right upper hand quadrant for container platforms. It's kind of like watching the DVD screensaver, you hit the corner, right? Something momentous. They actually drew the dot, it went outside the square. So we feel so strong in the container space and in the hybrid cloud space that we set the standard for how DevOps should be done in a cloud native process around DevOps. Now, what we're doing is we're taking that and we're making an MLOps journey. So what we did for DevOps with OpenShift, we're now doing for MLOps with OpenShift AI. And so to further that we've introduced RELL AI. So RELL AI is at its basics level, it's RELL, but it's in a new format called Image Mode, which we can talk a little bit about. It includes InstructLab, it includes the runtimes, and then it includes the Granite model. And so the model, there's both an English language model as well as a code model for it. And with that we're providing indemnity around that, IP indemnity. So for customers that are looking, how can I, in a dated sovereignty space around data is IP, I want to hold on to my data, but I want to take these foundational models and I want to fine tune them for my company's use case. And so what RELL AI is doing is making really democratizing that approach.

>> So if I use RELL AI, Granite, and get InstructLab with it, you indemnify against what?

Ryan King

>> IP indemnity. So it says like, hey-

>> So leakage.

Ryan King

>> Yeah, we can say, Hey, this is where this comes from, and we can guarantee that you're not going to be sued by somebody saying, "Hey, you put our IP into your model."

>> And the other interest thing about RELL is the edge, right? I mean, RELL can run anywhere.

Ryan King

>> Yes.

>> And so when you think about AI inferencing at the edge, RELL can play a key part there.

Ryan King

>> And OpenShift AI, to be frank. So Edge can come in a variety of flavors. So being able to serve up models at the edge is something a lot of our customers do. And obviously that draws down to smaller profile hardware. We also offer OpenShift as single node, OpenShift or small compact cluster, or there's even different variants there that they can run models within. We really do the model serving piece through OpenShift AI out to the edge. And so really the difference between RELL AI and OpenShift AI is really focus use case versus enterprise adoption.

>> And can you speak to the virtualization trends that are going on, how it might relate to your partnership with HPE generally and Green Lake specifically?

Ryan King

>> Yeah, of course. So I'll speak to the virtualization as I'll talk about the history of Red Hat around this. So inside of OpenShift, there's a capability called OpenShift virtualization. So that is not something you have to buy separately, it's something that comes with OpenShift. And what that is there's a project called Kubevert. And so that's based on KVM. So I think everybody in the industry knows KVM, like famously, like NVIDIA runs their entire like GPU cloud for gaming on KVM. So people have been doing virtualization in a Kubernetes context at scale for a very long time. And we offer that capability within OpenShift. And so customers have been saying, "Hey, can we run this as our virtualization platform?" We're like, "Fantastic." And what happened last August is we looked at VMs and you can qualify a VM, would it run well? I fit runs well in VMware, how well would it run in OpenShift vert? And the feature set that we added last August took that from a 20% of VMs are good candidates to 80% of VMs are good candidates. And so what's happened now is a lot of customers are saying, well, could we choose this as an alternative for our virtualization platform? So that's created a groundswell of interest around OpenShift in general, and it has all sorts of things that happen after, that that we can explore.

>> Well, so we had Rocco on earlier and we explained it to the audience, the whole changes in VMware's pricing strategies and essentially Broadcom doing exactly what it said it's going to do, but there's a lot of workloads that don't need the entire suite of VCF, et cetera. And so there's options out there and you guys are one.

Ryan King

>> Yeah, so we've been working with customers. They've been running OpenShift vert. So a lot of people were running OpenShift on VMware. So it's like a good portion of customers.

>> Right.

Ryan King

>> And more and more they're saying, hey, I'll just run this on bare metal, and that's where you run OpenShift vert. And so from an economic standpoint, it was really interesting to say, hey, I was paying for two things. I just paid for one now. Secondarily, when you run OpenShift vert, you get your RELL guests for free. So maybe in that virtualization context, they were paying for RELL as guests within that environment. Now those are free on top of OpenShift vert, and then maybe they were running Sentos, right? And they're saying, oh, what do I do? It's end of life. I got to pay Red Hat for this. We have great programs to convert people in that, but if you've moved OpenShift vert, you're going to get that for free too. So it's also a great way for people to look at what they do with their Sentos strategy or other Linux distribution because it just comes with OpenShift vert.

>> Now. What about HPE and Green Lake? How do you fit into that whole equation?

Ryan King

>> Well, I mean, as you know, third time, right? We've been with Green Lake from the very beginning. We've been pioneers in the concept of hybrid cloud. And so Green Lake bringing pay as you go into the context of the private cloud space and hybrid cloud is a great enabler for our platform. And so we started off, you launch a technology, you get your early customers, we're well into customer adoption, and we can go through all the different industries and say, we have customers now in every industry together, and we're watching them move to this consumption model within our platform. And so monthly they're saying, here's what we're using, that sort of thing. We get these consumption reports, it's very cloud-like. It's fantastic. And we're seeing two patterns, more customers, and we're also seeing that adoption ramp because when you have a consumption model, it's easy to consume and it's easy for people to try and become consumers. So organizational expansion is happening at the same time.

Rebecca Knight

>> So you're describing a lot of benefits that customers are already realizing as they modernize their architecture, but what else are you hearing even about the announcements from Red Hat Summit, which is not that long ago, in Colorado? Are you hearing things already from them, early reactions and responses?

Ryan King

>> Yeah, so I would say just a couple of things personally that I'm closest to. There's a broad set of announcements, Red Hat, but I draw in on the AI space and on OpenShift in particular. And then, we've started talk about tech preview of Image Mode. I think next year we'll be talking more about Image Mode, but I think it's probably the most subtle cool technology that's tucked into RELL AI. But in particular on the AI front, we have been on this march with NVIDIA. So we ran into each other at GTC, Dave, and that was on the back of us. NVIDIA was introducing NIM, their NVIDIA Inference Microservice capability, which is really about encapsulation and ease of consumption. And so we were already showing how you can use that with OpenShift AI right out of the gate. And then we added to that, and even last week, NVIDIA was announcing what we can do with them with KServe. So the ability to serve up NIMs within OpenShift AI. And so that capability is continuing to snowball. And so our partnership, our roadmaps between us and NVIDIA and capabilities are growing very strong. Separately with Intel, we announced... So we announced their new platform, their enterprise AI. So NVIDIA has AI enterprise, and Intel has enterprise AI. And so we're really so close with them because if you think about it really is a sandwich. You've got everything down to the kernel, and then you have the GPU and the driver, and you have layers like CUDA and you have PyTorch. And so at every layer we have to optimize with them, and then we have to manifest into real usability concepts, which we do with OpenShift AI for ML Ops. And they're helping to do with the consumability and availability of these different things through NIM or through the packaging that Intel's offering.

>> Can we double click on two things?

Ryan King

>> Yeah, please.

>> Image Mode and InstructLab.

Ryan King

>> Yeah.

>> Explain in more detail Image Mode and help us understand the roadmap and the adoption there. What's that like.

Ryan King

>> Yeah. So Red Hat Package Manager, RPM, I think some people are fairly familiar with that, certainly more technical part of the audience. That's like how you do packages within RELL. Image Mode is now you can do a containerized delivery of the OS. So very logical, but nobody done it. So you have essentially, okay, all our applications are containerized. Why are we doing packages in the OS? Why don't we containerize the whole thing? And so that's what Image Mode is, and it's simplest concept. It's got this thing called Bootsy, which allows you to boot from a container. And so right now, image mode... Sorry, RELL AI is like an immutable container that you just deploy to the server. And so when you think about over time, that becomes a concept of velocity and tooling. So all of the great tooling in DevOps that we know about delivering containers and repositories and all that stuff can now be leverageable for infrastructure. And so I think it's a very powerful concept, and I think they're going to see a lot of interesting concepts and approaches come out of Red Hat in our ecosystem as we start to push this beyond real AI and beyond tech preview with Image Mode.

>> And there's a security aspect to it as well, because it essentially simplifies the deployment because it's immutable, right?

Ryan King

>> Yeah.

>> And so that just checks one of the many boxes.

Ryan King

>> The applications have all these security concepts built around containers. So why not leverage that then have two different things. So anyway, it's early days, but when I first saw it, I was like, oh my gosh, I felt like I got the call from Doc Cloud when they said, "Hey, Ryan, we're going to change. We're going to become Docker. Go call Brian Stevens," or former CTO, "and let's have this conversation." I was like, oh, this must be cool. And so I learned about it. So I'm kind of having that moment again of like, oh, we're changing the OS, right? So I'm a big fan.

Rebecca Knight

>> Well-

Ryan King

>> We should talk InstructLab...

>> I just wanted to point out, so Stu and John Furrier were at, I believe it was the first where they were there for that announcement at the first Docker Con. Instructlab, very cool. The story I heard was basically the IBM labs had developed this tech. Matt Hicks evidently caught wind of it, or maybe somebody tapped on the shoulder and said, you got to look at this. Who knows how that actually went down. And then that began a collaboration, and you guys are bringing this to market, but explain InstructLab and we'll get into some of the magic.

Ryan King

>> Well, so behind the scenes, and so we always worked very closely with IBM research because we're always looking at the ecosystem. So myself and a couple other folks helped create this thing called Open Data Hub that became OpenShift AI. And so it was all ecosystem and collaboration, and we were kind of working with IBM in that. So we're always behind the scenes with them figuring out, where's this going? How can we better help our customers? How can we drive the community around it? So out of InstructLab, we saw this real great opportunity. First, the Grana model. And so open sourcing that and making that available within RELL AI and with OpenShift AI. But the InstructLab capability is... It's really in a democratization approach. And democratizing AI, we always hear about click ops, that sort of thing. What this is you can go in and you can enter your knowledge and data directly in as a non-coder practitioner and participate in a community that then selects the best data and knowledge from that, like simple YAML files into something that becomes... We do synthetic data generation, and that creates enough data so that it can be fine-tuned trained to re-weight the foundational model to now have your data in it, and then that's what becomes out of it. So InstructLab is just a great way... You can do it on your laptop and participate in the community. You can do a little kind of practice, kind of what they call inner loop, which is playing your laptop and try things out before it goes in the outer loop and it gets fully fine-tuned and everything. So it's really cool because it broadens the number of people that can put their knowledge and data into this generative AI era.

>> And I know you got to go, but just real quick, there's training. We all know about training, and then when you tune, you have to have a lot of human. With InstructLab, you feed it a little bit of human interaction, and then it's sort of autodidactic from there, which is quite amazing.

Ryan King

>> Yeah, it's very fun. I mean, just to have such... I mean, everything in generative AI is so amazing and enlightening to be able to contribute something that's novel and it broadens the number of people that can participate in it feels great. Personally from a company, it's all about community and openness.

>> Very cool.

Rebecca Knight

>> No, and I was there when it was announced, and to hear Matt Hicks talk about it, it's driving this learning culture too at Red Hat, but at other organizations too. So it really-

Ryan King

>> He put it out and it had a code name internally at first, and everybody's like, everybody needs to go in and just enter stuff they know and just go do it. Right? And so we all got on this private GitHub repo and we're just typing stuff in and trying stuff out, and then suddenly we're like, oh, wow, we're all doing this. Yeah, which is cool. No one's like one person, right? Can't hire that many of them. They really know how to do the really deep stuff.

Rebecca Knight

>> Makes it fun. Makes it, as you said, a community. Ryan, thank you so much for coming on The CUBE. It's great having you on.

Ryan King

>> Thank you so much. Yeah, great to see you again, John. Dave.

Rebecca Knight

>> I'm Rebecca Knight for Dave Volante. Stay tuned for more of The CUBE's live coverage of HPE Discover 2024. You're watching The CUBE, the leader in enterprise tech news and analysis.

Red Hat and HPE: Fueling customer innovation with AI, virtualization, and hybrid cloud solutions

search

Rebecca Knight

>> Good morning everyone, and welcome back to The Cube's live coverage of HPE Discover 2024. I'm your host, Rebecca Knight, sitting alongside my co-host and co-analyst, Dave Volante. Dave, this has been a great show. We're about to welcome back an illustrious Cube alum.

>> Yes. Well, third year in a row for having Red Hat on. Red Hat is smoking hot. They're red, they're so hot.

Rebecca Knight

>> Red-hot. They're red-hot. Exactly. I know. We can rename them right now. I'd like to welcome back to the show, Ryan King, senior director Global Hardware Partner Ecosystems at Red Hat. Thanks so much for returning to the queue.

Ryan King

>> Yeah, thank you. Yeah, I think three years ago when I did this the first time, it was quite the experience. I was fresh into the role. My background, I've always been in this business and this industry, but about the 2017 era, I started to look into what was happening with generative AI and it really helped bootstrap the relationship we had with NVIDIA. And so that got me on the radar to say, hey, what should be our broader strategy for infrastructure and infrastructure partners? And that's how I stepped into the role. And one of my first forays was with HPE around Green Lake. So it's great to come back and talk about that journey, but also the additional focus on AI and vert, which seems to be the two topics that everybody is having deep strategic planning around.

>> Oh, absolutely. And Red Hat is so many places between OpenShift and RELL, the entire open source software, what you've done within Instruct Lab, I'd love to get your point of view on that, but hit the escape key for us and help us understand AI's just broadly point of view, or Red Hat's broadly, their point of view on AI. What's the strategy? How should we think about it?

Ryan King

>> Yeah, so of course everything's founded in open and open source with Red Hat, and it all builds up from what we do with RELL and above that with OpenShift. But of course the portfolio part of it, everybody knows OpenShift and now OpenShift AI as being two forays as platforms that we're offering into this space. Also, they may have heard it Red Hat Summit, something called Light Speed that brings generative AI into the experience of using our products. And so you're seeing that being placed to use AI to use our products and platforms within it. So it's AI within, but also AI as a platform. Our point of view is we've had a very successful journey. I think you've seen the analyst reports that have us in the very far right upper hand quadrant for container platforms. It's kind of like watching the DVD screensaver, you hit the corner, right? Something momentous. They actually drew the dot, it went outside the square. So we feel so strong in the container space and in the hybrid cloud space that we set the standard for how DevOps should be done in a cloud native process around DevOps. Now, what we're doing is we're taking that and we're making an MLOps journey. So what we did for DevOps with OpenShift, we're now doing for MLOps with OpenShift AI. And so to further that we've introduced RELL AI. So RELL AI is at its basics level, it's RELL, but it's in a new format called Image Mode, which we can talk a little bit about. It includes InstructLab, it includes the runtimes, and then it includes the Granite model. And so the model, there's both an English language model as well as a code model for it. And with that we're providing indemnity around that, IP indemnity. So for customers that are looking, how can I, in a dated sovereignty space around data is IP, I want to hold on to my data, but I want to take these foundational models and I want to fine tune them for my company's use case. And so what RELL AI is doing is making really democratizing that approach.

>> So if I use RELL AI, Granite, and get InstructLab with it, you indemnify against what?

Ryan King

>> IP indemnity. So it says like, hey-

>> So leakage.

Ryan King

>> Yeah, we can say, Hey, this is where this comes from, and we can guarantee that you're not going to be sued by somebody saying, "Hey, you put our IP into your model."

>> And the other interest thing about RELL is the edge, right? I mean, RELL can run anywhere.

Ryan King

>> Yes.

>> And so when you think about AI inferencing at the edge, RELL can play a key part there.

Ryan King

>> And OpenShift AI, to be frank. So Edge can come in a variety of flavors. So being able to serve up models at the edge is something a lot of our customers do. And obviously that draws down to smaller profile hardware. We also offer OpenShift as single node, OpenShift or small compact cluster, or there's even different variants there that they can run models within. We really do the model serving piece through OpenShift AI out to the edge. And so really the difference between RELL AI and OpenShift AI is really focus use case versus enterprise adoption.

>> And can you speak to the virtualization trends that are going on, how it might relate to your partnership with HPE generally and Green Lake specifically?

Ryan King

>> Yeah, of course. So I'll speak to the virtualization as I'll talk about the history of Red Hat around this. So inside of OpenShift, there's a capability called OpenShift virtualization. So that is not something you have to buy separately, it's something that comes with OpenShift. And what that is there's a project called Kubevert. And so that's based on KVM. So I think everybody in the industry knows KVM, like famously, like NVIDIA runs their entire like GPU cloud for gaming on KVM. So people have been doing virtualization in a Kubernetes context at scale for a very long time. And we offer that capability within OpenShift. And so customers have been saying, "Hey, can we run this as our virtualization platform?" We're like, "Fantastic." And what happened last August is we looked at VMs and you can qualify a VM, would it run well? I fit runs well in VMware, how well would it run in OpenShift vert? And the feature set that we added last August took that from a 20% of VMs are good candidates to 80% of VMs are good candidates. And so what's happened now is a lot of customers are saying, well, could we choose this as an alternative for our virtualization platform? So that's created a groundswell of interest around OpenShift in general, and it has all sorts of things that happen after, that that we can explore.

>> Well, so we had Rocco on earlier and we explained it to the audience, the whole changes in VMware's pricing strategies and essentially Broadcom doing exactly what it said it's going to do, but there's a lot of workloads that don't need the entire suite of VCF, et cetera. And so there's options out there and you guys are one.

Ryan King

>> Yeah, so we've been working with customers. They've been running OpenShift vert. So a lot of people were running OpenShift on VMware. So it's like a good portion of customers.

>> Right.

Ryan King

>> And more and more they're saying, hey, I'll just run this on bare metal, and that's where you run OpenShift vert. And so from an economic standpoint, it was really interesting to say, hey, I was paying for two things. I just paid for one now. Secondarily, when you run OpenShift vert, you get your RELL guests for free. So maybe in that virtualization context, they were paying for RELL as guests within that environment. Now those are free on top of OpenShift vert, and then maybe they were running Sentos, right? And they're saying, oh, what do I do? It's end of life. I got to pay Red Hat for this. We have great programs to convert people in that, but if you've moved OpenShift vert, you're going to get that for free too. So it's also a great way for people to look at what they do with their Sentos strategy or other Linux distribution because it just comes with OpenShift vert.

>> Now. What about HPE and Green Lake? How do you fit into that whole equation?

Ryan King

>> Well, I mean, as you know, third time, right? We've been with Green Lake from the very beginning. We've been pioneers in the concept of hybrid cloud. And so Green Lake bringing pay as you go into the context of the private cloud space and hybrid cloud is a great enabler for our platform. And so we started off, you launch a technology, you get your early customers, we're well into customer adoption, and we can go through all the different industries and say, we have customers now in every industry together, and we're watching them move to this consumption model within our platform. And so monthly they're saying, here's what we're using, that sort of thing. We get these consumption reports, it's very cloud-like. It's fantastic. And we're seeing two patterns, more customers, and we're also seeing that adoption ramp because when you have a consumption model, it's easy to consume and it's easy for people to try and become consumers. So organizational expansion is happening at the same time.

Rebecca Knight

>> So you're describing a lot of benefits that customers are already realizing as they modernize their architecture, but what else are you hearing even about the announcements from Red Hat Summit, which is not that long ago, in Colorado? Are you hearing things already from them, early reactions and responses?

Ryan King

>> Yeah, so I would say just a couple of things personally that I'm closest to. There's a broad set of announcements, Red Hat, but I draw in on the AI space and on OpenShift in particular. And then, we've started talk about tech preview of Image Mode. I think next year we'll be talking more about Image Mode, but I think it's probably the most subtle cool technology that's tucked into RELL AI. But in particular on the AI front, we have been on this march with NVIDIA. So we ran into each other at GTC, Dave, and that was on the back of us. NVIDIA was introducing NIM, their NVIDIA Inference Microservice capability, which is really about encapsulation and ease of consumption. And so we were already showing how you can use that with OpenShift AI right out of the gate. And then we added to that, and even last week, NVIDIA was announcing what we can do with them with KServe. So the ability to serve up NIMs within OpenShift AI. And so that capability is continuing to snowball. And so our partnership, our roadmaps between us and NVIDIA and capabilities are growing very strong. Separately with Intel, we announced... So we announced their new platform, their enterprise AI. So NVIDIA has AI enterprise, and Intel has enterprise AI. And so we're really so close with them because if you think about it really is a sandwich. You've got everything down to the kernel, and then you have the GPU and the driver, and you have layers like CUDA and you have PyTorch. And so at every layer we have to optimize with them, and then we have to manifest into real usability concepts, which we do with OpenShift AI for ML Ops. And they're helping to do with the consumability and availability of these different things through NIM or through the packaging that Intel's offering.

>> Can we double click on two things?

Ryan King

>> Yeah, please.

>> Image Mode and InstructLab.

Ryan King

>> Yeah.

>> Explain in more detail Image Mode and help us understand the roadmap and the adoption there. What's that like.

Ryan King

>> Yeah. So Red Hat Package Manager, RPM, I think some people are fairly familiar with that, certainly more technical part of the audience. That's like how you do packages within RELL. Image Mode is now you can do a containerized delivery of the OS. So very logical, but nobody done it. So you have essentially, okay, all our applications are containerized. Why are we doing packages in the OS? Why don't we containerize the whole thing? And so that's what Image Mode is, and it's simplest concept. It's got this thing called Bootsy, which allows you to boot from a container. And so right now, image mode... Sorry, RELL AI is like an immutable container that you just deploy to the server. And so when you think about over time, that becomes a concept of velocity and tooling. So all of the great tooling in DevOps that we know about delivering containers and repositories and all that stuff can now be leverageable for infrastructure. And so I think it's a very powerful concept, and I think they're going to see a lot of interesting concepts and approaches come out of Red Hat in our ecosystem as we start to push this beyond real AI and beyond tech preview with Image Mode.

>> And there's a security aspect to it as well, because it essentially simplifies the deployment because it's immutable, right?

Ryan King

>> Yeah.

>> And so that just checks one of the many boxes.

Ryan King

>> The applications have all these security concepts built around containers. So why not leverage that then have two different things. So anyway, it's early days, but when I first saw it, I was like, oh my gosh, I felt like I got the call from Doc Cloud when they said, "Hey, Ryan, we're going to change. We're going to become Docker. Go call Brian Stevens," or former CTO, "and let's have this conversation." I was like, oh, this must be cool. And so I learned about it. So I'm kind of having that moment again of like, oh, we're changing the OS, right? So I'm a big fan.

Rebecca Knight

>> Well-

Ryan King

>> We should talk InstructLab...

>> I just wanted to point out, so Stu and John Furrier were at, I believe it was the first where they were there for that announcement at the first Docker Con. Instructlab, very cool. The story I heard was basically the IBM labs had developed this tech. Matt Hicks evidently caught wind of it, or maybe somebody tapped on the shoulder and said, you got to look at this. Who knows how that actually went down. And then that began a collaboration, and you guys are bringing this to market, but explain InstructLab and we'll get into some of the magic.

Ryan King

>> Well, so behind the scenes, and so we always worked very closely with IBM research because we're always looking at the ecosystem. So myself and a couple other folks helped create this thing called Open Data Hub that became OpenShift AI. And so it was all ecosystem and collaboration, and we were kind of working with IBM in that. So we're always behind the scenes with them figuring out, where's this going? How can we better help our customers? How can we drive the community around it? So out of InstructLab, we saw this real great opportunity. First, the Grana model. And so open sourcing that and making that available within RELL AI and with OpenShift AI. But the InstructLab capability is... It's really in a democratization approach. And democratizing AI, we always hear about click ops, that sort of thing. What this is you can go in and you can enter your knowledge and data directly in as a non-coder practitioner and participate in a community that then selects the best data and knowledge from that, like simple YAML files into something that becomes... We do synthetic data generation, and that creates enough data so that it can be fine-tuned trained to re-weight the foundational model to now have your data in it, and then that's what becomes out of it. So InstructLab is just a great way... You can do it on your laptop and participate in the community. You can do a little kind of practice, kind of what they call inner loop, which is playing your laptop and try things out before it goes in the outer loop and it gets fully fine-tuned and everything. So it's really cool because it broadens the number of people that can put their knowledge and data into this generative AI era.

>> And I know you got to go, but just real quick, there's training. We all know about training, and then when you tune, you have to have a lot of human. With InstructLab, you feed it a little bit of human interaction, and then it's sort of autodidactic from there, which is quite amazing.

Ryan King

>> Yeah, it's very fun. I mean, just to have such... I mean, everything in generative AI is so amazing and enlightening to be able to contribute something that's novel and it broadens the number of people that can participate in it feels great. Personally from a company, it's all about community and openness.

>> Very cool.

Rebecca Knight

>> No, and I was there when it was announced, and to hear Matt Hicks talk about it, it's driving this learning culture too at Red Hat, but at other organizations too. So it really-

Ryan King

>> He put it out and it had a code name internally at first, and everybody's like, everybody needs to go in and just enter stuff they know and just go do it. Right? And so we all got on this private GitHub repo and we're just typing stuff in and trying stuff out, and then suddenly we're like, oh, wow, we're all doing this. Yeah, which is cool. No one's like one person, right? Can't hire that many of them. They really know how to do the really deep stuff.

Rebecca Knight

>> Makes it fun. Makes it, as you said, a community. Ryan, thank you so much for coming on The CUBE. It's great having you on.

Ryan King

>> Thank you so much. Yeah, great to see you again, John. Dave.

Rebecca Knight

>> I'm Rebecca Knight for Dave Volante. Stay tuned for more of The CUBE's live coverage of HPE Discover 2024. You're watching The CUBE, the leader in enterprise tech news and analysis.