Interplay between Math and AI: Representation and Formal Mathematics
ABSTRACT
This talk explores how mathematics and artificial intelligence inform and shape each other along two lines of research. First, I discuss how mathematical structures, such as wavelet theory, can guide the design and analysis of neural networks, with connections to classical results like the universal approximation theorem, as well as recent perspectives on learning dynamics and spectral bias. Second, I turn to the use of AI in formal mathematics, highlighting recent progress in autoformalization and machine-assisted theorem proving in systems such as Lean. Together, these directions illustrate a broader vision: mathematics not only provides foundations for AI, but is increasingly becoming a domain where AI actively contributes to discovery and formal reasoning.
