HiPerGator AI: Florida’s new AI supercomputer (and 100 new AI faculty)

supercomputer-ai-chip-nvidia

What could an AI-focused supercomputer and 100 new AI faculty do for the University of Florida? Particularly if it’s the 22nd most powerful supercomputer on the planet, HiPerGator AI?

In this episode of TechFirst with John Koetsier, I chat with UF provost Dr. Joseph Glover and NVIDIA’s Cheryl Martin about what they’re planning.

Short version: massive impact.

We’re talking AI-driven climate change research, medical research and infusing artificial intelligence throughout the entire curriculum of a massive university that graduates 10,000 students annually. And that could have a huge compounding impact.

Keep scrolling for the video, to subscribe to the podcast, and a full transcript …

Watch: the University of Florida’s new AI supercomputer, HiPerGator AI

(Subscribe to my YouTube channel so you’ll get notified when I go live with future guests, or see the videos later.)

Subscribe: what the University of Florida is doing with its new AI-focused supercomputer

 

Read: how the University of Florida is infusing AI into every part of its curriculum with HiPerGator AI

(This transcript has been lightly edited for length and clarity.)

John Koetsier: How will the most powerful AI supercomputer in higher education and one hundred new AI-focused faculty change education — and the future — at the University of Florida?

It’s pretty clear that AI is critical to the future of almost everything. If you want good cars, if you want safe power, if you want smartphones, if you want autonomous drones, maybe robots. If you want power-saving technology in our homes, or better speech-to-text when you chat with Siri. Or maybe if you want to model how climate change might impact Florida’s coastline and people — you need better, smarter, faster AI.

Well, the University of Florida and NVIDIA just completed HiPerGator AI.

It’s the 8th most powerful supercomputer in higher education globally, and the 22nd fastest supercomputer on the planet … and I want to know what they’re planning to do with it.

To chat with us, we have Dr. Joseph Glover, Provost and Senior VP of Academic Affairs at the University of Florida, as well as Cheryl Martin who leads Higher Education for NVIDIA. Welcome, everybody!

Joseph Glover: Morning.

John Koetsier: Morning, morning. Thank you so much for joining us. Let’s chat with you first, Joe. Tell us about this supercomputer.

Joseph Glover: You mean the specifications of it? 

John Koetsier: Absolutely! Geek out a little bit for us. What does this beast look like? What can it do? 

Joseph Glover: It is amazing to us. I think it has 1,120 GPUs in it. As you mentioned, it’s the eighth fastest in the United States and it’s the 22nd fastest in the entire world. It actually fills our entire data center. We were really lucky when Chris Malachowsky and NVIDIA approached us about giving us this supercomputer.

You have to have a place to put it, and we had built a data center that we thought would serve our needs for the next 25 to 30 years before we filled it. This supercomputer, HiPerGator AI, filled it in January. We flipped the switch and it’s been operating flawlessly for us. 

John Koetsier: Wow. Impressive. I mean, we thought the days of room-sized computers were behind us, right? But I guess not for supercomputers.

Joseph Glover: No, I think that’s true. I mean, most people of course on campus will never actually see HiPerGator AI, because it’s located off campus, but it does fill a room.

It takes a lot of air conditioning and a lot of power to service it. I’m told that when we were running the speed test it was drawing 1.1 megawatts. So it’s a pretty hefty machine when it’s up and running full speed. 

John Koetsier:  That is super interesting, and maybe we’ll chat with Cheryl about that in just a moment as well. I mean, you almost need your own power plant or you certainly need to get increased power supplies in there. Maybe the state dims a bit [laughing] while you turn it on. Perhaps there’s even some solar power running there.

But I want to ask, what can you do with this supercomputer? What do you expect it to be working on, Joe? 

Joseph Glover: Well, we expect it to be doing a couple of things. For one thing, we expect it to really be a new tool that’s going to be in the arsenal of the people doing research. As you know, the AI supercomputers are absolutely fantastic at crunching huge amounts of data. And we are tackling real-world problems that do involve enormous amounts of data.

In the field of medicine, we’re looking at better medical outcomes, we’re looking to bend the cost curve of medicine. In agriculture, we know that the United States and quite possibly the southeast is going to end up as the nation’s food basket, and so we are doing that. As you mentioned in your intro, climate change is a challenge for the state of Florida in terms of sea level rise and the changing environment. We’re trying to get a handle on that. All of these things involve huge amounts of data, and this is where the HiPerGator AI really excels.

So, in fact, we’ve already done one really large program — Cheryl is probably more technically able to explain it that I am — but it’s called GatorTron, in which we digested about a billion words of medical records from the past 10 years in our hospital system, to see if we could de-identify it to make it usable and to begin to find, to glean hidden truths in there. And so far, the doctors and the scientists working on this are very excited about the preliminary results.

It’s a little bit over my head, quite honestly, but I just wanted to give that as an example that we’re already engaged in really large projects. 

John Koetsier: Cheryl, maybe tell us a little bit about that. That sounds super interesting. I mean, I’m looking forward to an AI that I can give my health information to and everybody else’s, and in an anonymous way give me tips and insights about how I can be healthier. What was that project? 

Cheryl Martin: Yeah. So, this actually looked at the — had the data from the University of Florida’s health system. And what they’re able to do, if you look at what the healthcare industry has gone through over the last … I don’t know, what, 20 years … was to get all of those paper records onto computerized systems and have them digitized. And they went through all that effort and, okay, it’s there … there wasn’t a lot of value that came out of that other than, you know, maybe they’ve cleaned up rooms of paper and stuff.

But now there’s actually value inside of all that data.

And so, what they can do then is the nature it’s built on — something called Megatron from NVIDIA — and then they took this and they ran it through all that data. And so it can help them understand things like trends that are happening in different areas, but also help them with disease detection and hopefully lots of other things along that path by having this data.

And they’ve just started really. And when you look at this, there’s so many things you have to learn, like even dialect. So there’s data that comes out of that, but there’s different dialects from the Northern part of Florida to the Southern part of Florida. And so to really understand what’s in that data, it takes a lot of compute power to do that. But the benefits, the outcome, the vision, the future where we’re going, if you can see the vision of the future of where it can take us … is amazing.

John Koetsier: That is really interesting.

I love that point about dialects. I mean, we’ve all seen the videos on YouTube about the guy in Scotland or Ireland trying to get the voice-activated elevator to go [laughing] and it does not understand him at all. We all have some challenges with that.

Joe, I want to come back to you. Two things that you’re doing that are really interesting and, in my understanding, quite unique. One, you’re adding a hundred faculty in and around AI, teaching AI, learning with AI, all that stuff. And two, you are bringing AI across the curriculum. Talk to us about why you’re doing those things and how it’s working out.

Joseph Glover: Yeah. First, let me talk about adding a hundred faculty. Of course we already had several hundred faculty who were engaged with AI in one form or another, but we have really jumped into this with both feet.

We believe that this is going to be a transformational initiative for the University of Florida. We think that this is where higher education is going to inevitably go. And so we wanted to double down on our investment and really ensure that we strengthen the areas even further that were going to do AI.

Now, we don’t do just pure AI from a computer science sense. We do AI and applications. So we’re investing those extra 100 faculty across the entire University of Florida to do all sorts of things.

As I mentioned before, we’re doing AI in medicine, AI in drugs, AI in agriculture, AI in business. The College of Business just made AI a required introductory course for their entering freshmen. So we are in the process of adjusting our curriculum.

And right from the outset when Chris Malachowsky and NVIDIA approached us about this machine, they asked us the question, ‘What would you do with it?’ And one of our answers was we would teach AI across the curriculum. We are not limiting it to the computer science department or even the College of Engineering … it’s being taught in every college.

And I’m really pleased to say that the faculty have embraced this. They see this as the future. They see it as a wonderful tool and something that’s going to be a great advantage to the students to have in their skills portfolio. And equally importantly, the United States federal government has identified the creation of a 21st century AI-enabled workforce as one of the nation’s critical security problems — both from the point of view of literally national security, but also economic security. In order to build a 21st century AI-enabled workforce, you have to educate people at scale. And so, we believe that educating all of our students across the entire university is going to contribute significantly to growing this workforce.

We graduate about 10,000 students a year at the University of Florida. If even half of them come out with AI competence, that’s going to be a huge infusion into the economy of Florida, and the southeast United States, and the nation more generally.

And moreover, we think that this is a national model that could be imitated by universities around the country. So we’ve actually engaged with the state university system in Florida and talked to them about this, and are trying to convince them that this is the way to go. We’re also in conversation with the universities in the SEC to see if they would like to contribute on this scale as well. 

John Koetsier: I can’t tell you how impressive that is. It’s amazing, because I see it every day. I work with it every day. I work with technology. I report on technology all the time.

And AI is critical to almost every piece of technology that we have and almost every activity that we have. And for business and for products improving at scale and at speed, embedding intelligence in them, embedding the ability to learn from how they’re being used and to get better at adapting themselves to how people need them to work, is really impressive. And bringing that across the curriculum is really, really amazing.

Cheryl, let’s bring you back in here because this is not just one thing that’s going on. This is part of a larger effort to equip higher education with supercomputers. Talk with us a little bit about that for a moment. 

Cheryl Martin: Yeah. So we kind of look at this, there’s sort of three different parts to this, and Joe touched on all three of those.

And the first one really is the workforce enablement piece. So if you look at NVIDIA, it is engaged in AI across the spectrum of all the different industries. And every industry is in a different place, but we’re kind of still at the beginning of all this, but it’s moving so fast. If you think back even 10 years ago, even five years ago, how fast we’re moving. So when you look at the workforce readiness piece of this, we’ve got a long way to go. So we’re going to have graduating students going out there and we already have a lot of those industries that are just not finding the skills that they need today.

So there’s opportunities to re-skill the workforce, but to be able to bring in those skilled workers is incredibly important. So the work that Florida is doing to look at this cross-discipline is, you know, it’s in the retail industry, it’s in the healthcare industry, it’s in automotive, it’s an every single industry, right? So the workforce readiness piece is extremely important, and so NVIDIA is doing a lot of work to ensure that we help build the skills across the spectrum.

The second part of this is research.

And I think this is one of the ones that typically you might think of NVIDIA with, but if you look at research, the United States still typically leads in research, but other countries are starting to have more papers accepted, more and more papers accepted at the different conferences. And so it’s really important, the kind of research that we can do, the important research that could have never been done before. So everything, you know, huge amounts in the healthcare industry.

But again, everything that we do, and I think that that’s been one of the good benefits that we’ve seen at University of Florida because they have the infrastructure to take on that research, they have attracted a lot more funding to be able to start to do some of that research. So research is — and doing research that we could have never been done before. So, that’s the second part of this.

And then the third part, which Joe touched on a bit as well, which is the local economic impact.

So we all hear about the brain drain and the issue with so many of the really highly skilled people going to industry versus staying in research, or staying in the community that they’re in. So the work that Joe is and the University of Florida is doing to include the state university system. And then another one that we’re working on with them is the inclusive engineering. So as we look at the historically Black colleges and universities and the Hispanic speaking, you know, all of the minority serving institutions and being able to include them as part of this effort is really, really important as well. So it’s a pretty broad initiative and it’s extremely important for the United States in particular. But we’re seeing this play out in different scale, in different countries as well. 

John Koetsier:  Very cool. And maybe, Cheryl, geek out with us for just a moment here, tell us about HiPerGator AI in some detail. What does that computer include? What are the components that build it up and what is it capable of? 

Cheryl Martin: Well, Joe did a pretty good job there with his initial description.

And it’s 148 of the NVIDIA DGX systems, and it’s 1,120 of our A100 processors. And just to put that in perspective — this was true a couple of months ago and I think it still is — if you look at all of the, in North America, the systems that are available for researchers, it’s bigger than all of those combined … from a GPU perspective.

So this, the newest version of the GPUs and the type and GPUs, they’re obviously large CPU systems and older GPU systems as well, but this one is bigger than those. So it’s really taking things to a new level from that standpoint. 

John Koetsier: Very cool. Very cool. I want to thank both of you. Super interesting. Really neat to hear your perspective, Joe, on what you’re building, what you’re doing, what you’re providing. Thank you for taking some time. 

Joseph Glover:  Thank you! Been a pleasure. 

John Koetsier: Thank you, Cheryl. 

Cheryl Martin: Thank you, John. 

John Koetsier: Have a great day. 

5% of people make it down here

Made it all the way down here? Who are you?!? 🙂

The TechFirst with John Koetsier podcast is about tech that is changing the world, including wearable tech, and innovators who are shaping the future. Guests include former Apple CEO John Scully. The head of Facebook gaming. Amazon’s head of robotics. GitHub’s CTO. Twitter’s chief information security officer, and much more. Scientists inventing smart contact lenses. Startup entrepreneurs. Google executives. Former Microsoft CTO Nathan Myhrvold. And much, much more.

Subscribe on your podcast platform of choice:

 


Want weekly updates? Of course you do …