How Robots Can Teach Us to Trust A.I.
Oct 29, 2021
2:30PM to 3:30PM
Date(s) - 29/10/2021
2:30 pm - 3:30 pm
There has been an explosion of interest in the field of artificial intelligence (A.I.) to guide materials science, this has resulted in the discovery of exciting new phase change materials, amorphous alloys, and catalysts. But our A.I.’s are powered by data and the scientific literature largely consists of one-off experiments without quantified uncertainties, sufficient metadata to ensure reproducibility, or access to the primary data used to draw conclusions. Here I will discuss the tenuousness of ground truth, the need for openly preserving expert disagreement within scientific data sets, and challenges associated with aggregating data from the open literature. This will drive home the difficulties in forming and capturing expert consensus, the impact of consensus variance on ML model evaluation, and the need to recreate important datasets that are born digital. To this end, I will conclude by discussing the emerging paradigm of autonomous research systems as an opportunity to (1) rapidly (in)validate new AI predictions, (2) generate as-needed data to supplement existing datasets, and (3) accelerate the discovery of new materials and phenomena.
Jason Hattrick-Simpers is a Professor at the Department of Materials Science and Engineering, University of Toronto, and a Research Scientist at CanmetMATERIALS. He graduated with a B.S. in Mathematics and a B.S. in Physics from Rowan University and a Ph.D. in Materials Science and Engineering from the University of Maryland. Prior to joining UofT Prof. Hattrick-Simpers was a staff scientist at the National Institute of Standards and Technology (NIST) in Gaithersburg, MD where he co-developed tools for discovering novel corrosion resistance of alloys, developed active learning approaches to guide thin film and additive manufacturing alloy studies, and developed tools and best practices to enable trust in AI within the materials science community.