From Tokens to Theorems: Building a Neuro-Symbolic AI Mathematician

AI is poised to automate mathematical discovery, potentially solving problems like the Riemann Hypothesis and transforming science. This future could involve millions of AI mathematicians working in parallel, compressing centuries of progress into short timeframes. The proposed approach is a neuro-symbolic one, combining large language models with symbolic solvers in a "generate-check-refine" cycle. Citing recent breakthroughs with models like GPT-5 and DeepMind's work, it suggests this revolution in scientific inquiry is closer than many expect.