Terence Tao discusses the October 6 release of over 700 OpenAI research papers as a watershed moment for mathematics. While AI can make groundbreaking contributions, some mathematicians are unhappy with the disregard for community expectations of attribution and quality. Tao argues that the field should not be "hollowed out" simply because AI tackled major problems.
He notes that even if AI solved every problem in the Langlands program, it would not "solve number theory," comparing this to saying the periodic table solved chemistry. The recent OpenAI dump barely dented the general problems of functoriality and reciprocity. Hugo Duminil-Copin lamented that all major open problems he mentioned were claimed to be proved.
Tao draws a parallel to the 1972 classification of finite simple groups announced by Dan Gorenstein, which initially cooled interest in the subject despite remaining rich questions. He suggests mathematicians should follow the example of Gorenstein, Lyons, and Solomon in revising classifications, focusing on the responsibility not to pronounce subjects "done" but to explore new playgrounds. The post emphasizes that central problems are merely the natural boundaries of understanding, and unlocking them is the first step to exploration, not the end.
Source: Hacker News · Summarized by HeadlinesBriefing