The University of Florida announced it will stand up a state-of-the-art AI supercomputer, powered by the latest Nvidia technology, to tackle major challenges such as rising seas, aging populations, data security, personalized medicine, urban transportation and food insecurity. Unveiled today, the $70 million project is undergirded by a $50 million gift — $25 million from UF alumnus and Nvidia cofounder Chris Malachowsky and a matching $25 million donation from Nvidia, comprising hardware, software, training and services. With UF investing $20 million, the combined resources will go to building the fastest AI supercomputer in higher education, available not only for UF but for educational institutions across the state of Florida, including HBCUs.
“I’ve reached a point in my life, both professionally and personally where I was ready to give back, and I focus those efforts on University of Florida,” said Malachowsky, who met his future wife at UF as an engineering student in the late 70s. “UF’s leadership has a bold vision for making artificial intelligence accessible across its campus. What really got Nvidia and me excited was partnering with UF to go broader still, and make AI available to K-12 students, state and community colleges, and businesses. This will help address underrepresented communities and sectors across the region where the technology will have a profound positive effect.”
Comprising 148 Nvidia DGX A100 systems, UF’s third-generation HiPerGator machine will be one of the top AI-focused supercomputers in the world. Like Nvidia’s in-house system Selene (ranked #7 on the Top500), the DGX SuperPOD installation is powered by Nvidia A100 Ampere GPUs, AMD Eypc Rome CPUs and 200 Gbps Mellanox networking technology. With a total 1,120 A100 GPUs, the system touts a peak performance of 700 “AI petaflops” using half-precision tensor core math and Nvidia’s sparsity feature. In HPC terms, theoretical performance comes out to 10.9 traditional double-precision petaflops, or 21.8 double-precision tensor petaflops, leveraging Nvidia’s new FP64 tensor core format.
The system is on track to be delivered by the end of the year and to enter production at the start of 2021. Two nodes of the technology were installed last month and are being used as an on-ramp for UF faculty and technical staff.
The new computational resources will support AI training, inference, and data analytics at UF and across the state of Florida. In a media briefing, UF Provost Joe Glover said a significant number of faculty are already engaging with artificial intelligence applications, spanning autonomous vehicle design and development, new drug discovery and finance. The university is committed to hiring an additional 100 faculty members to focus on AI technology and domain applications.
With its enriched AI ecosystem, UF will be expanding its focus to moonshot projects, interdisciplinary or cross disciplinary grand challenges. One of these is the iCoast initiative, which addresses the problem of rising sea level in the state of Florida, which in addition to submerging land causes salt water to infiltrate the freshwater aquifer. The iCoast initiative seeks to generate simultaneous data streams from the coastline involving water temperature, chemistry, salinity, and biological life in an effort to understand what’s happening and to formulate strategies to address it.
Further, UF, a land grant university of 55,000 students, plans to integrate AI across the curriculum. The university has instituted a bachelor’s degree in data science, it’s creating a certificate in AI, and it has installed bootcamps in AI. It’s also exploring other means to infuse AI across the disciplines and across the curriculum. “We think it’s important to do this at scale because we know that it’s important to train the next generation of an AI enabled workforce,” said Glover. UF expects to matriculate 30,000 AI-capable graduates by 2030.
Plans are in the works to develop public-private partnerships that benefit Florida and the nation. “We’re going to be reaching out to all of the industries in Florida, and throughout the United States,” said Glover. “The College of Engineering has deep connections with many engineering companies in the region and across the United States, our College of Medicine similarly. I think each college has pretty deep connections with an industry. As the flagship land grant university in the state, we are really responsible for the agricultural enterprise in the state of Florida. And precision agriculture is an area that AI will have and is having a profound impact in. So so we fully expect that this initiative will have a great impact on all of those industries, and that they will quickly be knocking on our door and asking how they can partner with us in bringing the benefits of this new technology to their particular industry.”
As part of its mission to broaden access and inclusion to AI, UF has established a partnership with an inclusive engineering consortium, comprised of electrical and computer engineering departments from 15 historically black colleges and universities, as well as two Hispanic serving institutions, with the goal of providing access to AI training, expertise and computing resources. During today’s announcement, Damon Woodard, director of AI partnerships and a faculty member in the Department of Electrical Engineering, reported that these efforts together with partnerships with industry stakeholders, such as the one with Nvidia, are critical to the development of an AI enabled workforce. “We believe that the result of these partnerships will be a diverse AI enabled workforce, as well as technological advances that address problems statewide, nationally and globally,” said Woodard.
A newly formed Nvidia Deep Learning Institute will collaborate with UF on developing curriculum and coursework for both students and the community, including programming aimed at engaging young adults in STEM fields. As well, the university will host an Nvidia AI Technology Center, which brings UF graduate fellows and Nvidia employees together to work on AI projects.
The partners see the charity-based funding mechanism being a replicable model for training the next generation of the workforce to be versed in AI. “I think it’s an easy fundraising pitch, [and] I really do hope that this becomes a blueprint for other land grants and other universities around the country, other states, to take advantage of their schools and their alumni and their state and local governments to work cooperatively to help do something that benefits everybody,” said Malachowsky.