Cameron McLeman

Artificial Intelligence in Pursuit of New Mathematics: The Case of Symbolic Regression

Abstract:

Much time has been spent on the phenomenal power of mathematics as employed in modern machine learning tools, put to use to further our understanding of every scientific discipline in existence. The field of mathematics itself, however, has thus far proven to be somewhat resilient to attack from our new AI overlords. Indeed, ChatGPT and other large language models can easily be prompted into producing mathematical falsities, and even more formally-oriented models frequently succeed only in sounding more logical and rigorous. This status changes as we upgrade from language models to reasoning models, and indeed the AI of the near future will very likely be more than a mere research assistant in our mathematical progress. In this talk we’ll survey some of the recent history along these lines and discuss one potential pathway for AI-driven mathematical innovation, that of symbolic regression for doing experimental mathematics.

Bio:

Dr. Cam McLeman is the director of the Division of Mathematics and Natural Sciences at the University of Michigan - Flint. His research spans number theory, algebraic graph theory, and the arithmetic of elliptic curves, with a more recent interest of applying artificial intelligence techniques to further research in all of these fields.