AI for the Working Mathematician

Speaker: Lauren Williams, Harvard

Title: First Proof Batch 3: Mathematicians putting AI to the test

Abstract: First Proof is an initiative that aims to obtain a nuanced, objective assessment about the capabilities of large language models to prove specified mathematical statements. I will report on the results of First Proof’s third batch of problems. We will be scoring AI-generated solutions based on (1) mathematical correctness, (2) quality of exposition, and (3) quality of attributions.