April 6, 2018
PSC has a strong history enabling artificial intelligence (AI) research at scale. AI advancements at PSC include Never Ending Language Learning (NELL), led by CMU’s Tom Mitchell and William Cohen, on PSC’s Blacklight system and the 2017 history-making victory of Libratus, led by CMU’s Tuomas Sandholm, over the world’s top human poker pros on PSC’s Bridges system. Recognizing the vast strategic importance of AI to science and society, PSC has created a new Artificial Intelligence & Big Data (AI&BD) Group, directed by Paola Buitrago.
The AI&BD Group strengthens PSC’s AI environment for users, develops and disseminates best practices, enables and drives research, provides education and training and fosters collaboration. Buitrago and her team interface closely with AI hardware and software technology developers and other groups across PSC to provide valuable, forward-looking insights.
“The next step for AI is going to be enabled by scaling the algorithms and data access to an HPC level,” noted Buitrago. “People have been limited by access to expertise in the disciplines of HPC, AI and Big Data, but to attack the complex problems waiting to be solved, we will need that convergence. PSC is uniquely positioned to make that happen.”
“We’re seeing an explosion of interest in AI at PSC, spanning algorithm development, pure AI projects, integration of AI with traditional HPC and coursework,” said Nick Nystrom, interim director of PSC. “It’s transforming computational science, and Paola’s AI&BD group is ideally positioned to lead important contributions.”
Formation of the AI&BD Group supports PSC’s balanced approach to data-driven discovery, modeling and simulation, emphasizing access for all users, ease of use and support for communities and applications that have not traditionally used high-end computing resources.