When: 11AM on December 9, 2022
Where: Joyce Cummings Center, Room 260
AI is increasing in use and importance not only in our daily lives but in scientific research too, as scientists use AI to solve data-intensive problems, reduce computational costs, or replace time-consuming tasks in their scientific workflows. In parallel, scientific publications have raise the question: “is there a reproducibility crisis?” They have documented cases, surveyed the community, started investigating technical solutions, proposed recommendations, and generally study how the classic principles of research of reproducible scientific processes and results apply in this new context. In this talk, we ask the question “how well do AI methods measure up?” We will discuss the challenges of reproducibility for AI applications, community definitions, and initial solution approaches. We will make the case that reproducibility and Findable, Accessible, Interoperable, and Re-usable (FAIR) principles are essential foundations for trustworthy AI, where users can understand and trust the results of the AI methods used.
Bio
Line C. Pouchard is an internationally recognized expert and senior researcher with over 2 decades of experience in scientific domains of interest to the Department of Energy and over 100 publications. She leads multi-disciplinary technical projects to create innovative approaches improving scientific data discovery, data management and curation. Her research focuses on provenance for workflows at scale, computational reproducibility, and text mining for big data. Prior to her position at Brookhaven National Laboratory, she was Assistant Professor at Purdue University and Staff Scientist at Oak Ridge National Laboratory. She has a PhD from the Graduate Center of the City University of New York, and an MS from the University of Tennessee, Knoxville.