Empire AI is already making rapid progress in drug discovery

One of the most transformational projects in New York higher education is Empire AI, a first-in-the-nation consortium of public and private research universities to advance artificial intelligence research in the state. Headquartered at the University at Buffalo, Empire AI is predicted to have not just an academic impact on the state – including from research and the growth of new academic programs – but an economic impact with the research serving as a base for new companies. City & State talked with Venu Govindaraju, the University at Buffalo’s senior vice president for research, innovation and economic development, to discuss where Empire AI is going. Govindaraju has been working on artificial intelligence projects for over 30 years, earning him the moniker “the OG of AI.” This interview has been edited for length and clarity.

Where do things stand now with Empire AI?

Empire AI is, as we all know, a consortium of universities, right? So we already have a board, and all the consortium members are represented in the board. So that is one aspect. The other aspect is the real hardware, right? The computers that will, essentially, catapult us into the new age of research. So the way things stand is, I believe it was October of 2023 if I’m not mistaken, when the governor came and inaugurated the alpha machines. So the alpha machines were essentially coming from a gift from (the) Simons Foundation. And those, I think it was $10 million and then there is something more added on to it. Those computers are in UB’s Center for Computational Research in our downtown office in the economic development zone. Those are operational. They have been performing and all the consortium members have been running projects on those machines for the last year and a half and with some very, very exciting results to show for that. Then the next installment, buying computing hardware, what is called beta, I think we have about 280 or so of those GPUs, which are all just about to be operational in maybe a few weeks. These are very high density computing chips, so there were some arrangements that had to be made to accommodate the power requirements and also the cooling that needs to be done. These beta machines, I think, were an additional $30 million to 40 million of investment, and will give us about anywhere between 10 to 20 times speed up over the alpha.

Where do things stand with the permanent Empire AI data center?

The real data center is being built as the Empire AI computing center on the University of Buffalo campus. The drawings are all being approved as we speak, and groundbreaking should happen sometime in late summer. We are told that it should take 12 to 18 months. So my guess would be that that center should be ready maybe at the end of next year, sometime (after) for operation. That will be, again, another additional 10 to 20 times speed up, so you can see how you know we have this is really the latest and greatest computing infrastructure available.

What types of projects has Empire AI been used for so far?

At UB, we have had some very good, interesting projects in the health space, in the drug discovery space. I think the other consortium members have talked about helping with climate prediction, with the environment, with urban planning, so it’s the entire gamut of projects. My understanding is more than 120 projects have already been submitted to get time on these machines.

What was the pharmacology project that the University at Buffalo was working on with Empire AI?

This is really exciting and interesting for me, because from my understanding of how this works for drug discovery, scientists would usually design a particular kind of protein based on their understanding of its structure. Its protein gets folded, and somebody got a Nobel Prize a few years back trying to predict how this protein folds in various dimensions. The way scientists should work is they essentially design by simulation on a computer, and they come up with some kind of a predicted structure, then they go and actually fabricate that protein in a lab. Then they test that manufactured protein under various instruments and see how that behaves. Then they come back to the drawing board and tweak the design and go back and do the same cycle again. Before Empire AI, that cycle, each step, each iteration, would take anywhere from six to eight months, not because some of us don’t have the most sophisticated computing infrastructure, but because it is not what the latest Empire AI is offering. So now, with the alpha, one of our scientists, he was able to show that instead of taking six months for an iteration, he can just do it in six days. And the prediction is that once we have beta next month, that could be very well six hours.

What type of impact can Empire AI have long term on New York?

It’s not just getting some publications and scholarly activities going, but actually getting the translation part of getting startup companies launched and commercialization happening. So in New York’s case, AI can be used to bring down healthcare costs, because that is looming large in every state, especially New York, because we have such good benefits for everybody but the cost is going to skyrocket. It’s already skyrocketing. What can AI do? Drug discovery will be faster, so maybe it will get cheaper. Maybe the way we monitor and manage our hospitals, the beds, the patients, all those things. We have more than 100 projects of interest with AI, and more than half of them are in the healthcare space, because that’s most appealing to a lot of people, where AI and medicine can come together.

Can Empire AI serve as an economic catalyst?

That it is, attracting startups to relocate in this ecosystem. I think you must have covered (or) seen a story in The Buffalo News a year or so back where a startup company from Boston was relocating in Buffalo, just with the idea of being close to this ecosystem of Empire AI. Another effect it is having is we see more faculty members, university faculty members, choosing UB over perhaps universities which are similarly ranked, or even maybe higher ranked, but they’re preferring UB because of the access to Empire AI. We’re just getting started.

New York has seen a growth in technology investment between Empire AI, Micron and the federally designated technology hub spanning Buffalo, Rochester and Syracuse. What will the impact be for the state?

This is a game changer. I think the Empire AI, and the tech hub and the semiconductor industry with Micron and Syracuse, it’s going to take a little bit of time, but this is exactly what we need to catapult us to be the leader in the next decade in the technology space. Semiconductors are going to be so important. All these AI advances are coming out of these specialized chips, and we need to absolutely be on the cutting edge of that and also not be vulnerable to the supply chain issues (that) were exposed during COVID. Looking ahead, we have to be also well placed in quantum, and I see that our state has made huge investments. There have been investments in the Stony Brook area. There was an announcement by the governor that there will be four different quantum hubs, which will be funded and supported to make sure that New York is leading in that space. So AI, semiconductors, quantum, these are all the technologies where I think we are positioning ourselves very well, and ultimately, these will deliver the benefit to society, whether it’s in healthcare or whether it is in education, all those different sectors.

You’ve been called “the OG of AI,” and you’ve done some unique and interesting research in this field. What are your thoughts about that description and talk about some of your work in this field?

That’s a tongue in cheek comment. I heard it a year back, I think when Empire AI was being announced, and I think somebody quite important just called me that. I got into AI 30 years back at the University at Buffalo in the ’90s. We worked on postal automation, and we were the first to actually develop a system that could read handwritten addresses automatically. So just imagine, 30 years back, somebody drops an envelope in the mailbox by just handwriting an address. No human being lays eyes on that address. There is an image taken and a computer program … figures out where that mail should be delivered among all the almost 300 million possible destination places within the United States. It was a big success, because we did something in the real world (that) had not happened (until) then, because most of the artificial intelligence discipline work up to that point had not succeeded in the real world. If you go and look back, the big stories were: The computer can play “Jeopardy.” The computer can play chess and beat the grand masters. Excellent. But those were not helping the public, right? It was not a real-world solution, whereas postal automation was at scale. So, since I was there in the beginning, and that success has been touted as one of the first real-world AI success stories, I guess somebody just wanted to have some fun and call me the OG.