
AI in Math and Theoretical Physics: Status and Prospects
Michael Douglas, a physicist at Harvard, argues that rapid AI progress means researchers should prepare for the possibility of artificial general intelligence within a few years, however implausible that sounds. Speaking at the IAS High Energy Theory Seminar, he draws parallels to earlier moments when machines surpassed humans, using chess as the clearest precedent for how a field adapts once computers outperform experts. The bulk of the talk covers concrete work on AI tools for mathematics: conceptual designs for mathematical copilots that assist human researchers and more ambitious systems aimed at autonomous mathematical discovery. Douglas assesses where these tools currently stand, what technical barriers remain, and why he expects some of them to become genuinely useful to working mathematicians and theoretical physicists within the next few years. The talk stays grounded in specific research directions rather than speculation about superintelligence, aimed at an audience of physicists and mathematicians weighing how AI might change their own research practice.