Mathematics Colloquium
Mike Kirby
University of Utah
Title: Democratize Responsible AI: A ‘Many Voices’ Approach Spanning Mathematical and Computer Science And Beyond
Date: Friday, November 8
Place and Time: Love 101, 3:05-3:55 pm
Abstract. AI (Artificial Intelligence) is now an integral part of our daily lives, sometimes in ways that we do not even realize. AI has rapidly developed into the transformational technology of our time, with the potential to impact every aspect of our lives. From the perspective of those in the sciences and engineering, recent advances in artificial intelligence and machine learning (AI/ML) have introduced the possibility of using the tools of physics-informed machine learning (PIML), such as physics-informed neural networks (PINNs) and neural operators, as part of the modeling and simulation pipeline. We will begin by summarizing our recent work on domain decomposition strategies when using PINNs. We will then present our recent work on using Kolmogorov n-widths, a measure of the effectiveness of approximating functions, as a metric for the comparison of various multitask PIML architectures. We will then transition to discussing how AI is not solely a technological endeavor but one that requires us to muster all our talent towards solutions that serve our community. When used responsibly, AI can be a tremendous asset, positively impacting everything from routine daily tasks and services to urgent and important societal- and global-level challenges. We will present recent our team-science work applying AI/ML to problems ranging from oncology to political science. To leverage this inevitable transformation for social and societal good, we believe in the importance of having many voices at the table to represent diversity of outlook and expertise — hence our belief in the need to Democratize Responsible AI.