The Center for Generative AI at The University of Texas at Austin, already among the most powerful artificial intelligence hubs in the academic world, is doubling its computing capacity to more than 1,000 advanced graphic processing units, or GPUs.
The additional AI computing power will speed discoveries in biosciences, health care, computer vision and natural language processing (NLP) that result in new vaccines, advanced medical imaging and video quality, personalized medicine and more accurate computer processing of human language. Such research often requires hundreds of GPUs working in parallel on massive datasets.
"This is a game-changer for open-source AI and research in the public domain, not only at UT but throughout academia," said Adam Klivans, director of the UT-led National Science Foundation Institute for Foundations of Machine Learning. "The scale of the cluster will allow us to create solutions to bigger real-world problems that make a difference in people's lives. It's exciting to accelerate discovery and to create more opportunities for our researchers to push the boundaries of what's possible."
The Texas Legislature recently appropriated $20 million to pay for a portion of the center's additional GPUs, which will include the most advanced chip technology.
Whereas much of UT's leading AI computing power is open to researchers from outside the University, the Center for Generative AI is dedicated to UT faculty and student researchers, giving them state-of-the-art computing power and frequency of access unrivaled by any university.
Open-source computing at UT and throughout academia is nonproprietary and can be fine-tuned to support research in the public interest, producing innovations that impact people and fields across all disciplines. The Center for Generative AI's computing cluster makes UT one of the few places where large models can be trained from the ground up, critical to ensuring a model's interpretability and accuracy for downstream applications. Interpretability allows researchers to understand which aspects of a model are most relevant to its findings, mitigating bias and guiding future experiments.
