At Stanford University, OpenAI CEO Sam Altman, facing hundreds of students worried about their career prospects, talked about the collective psychology of the mathematics community when confronted with AI. In his view, mathematicians went through a journey resembling "grief."
The first stage was disbelief, followed closely by pain. Altman recounted the typical reactions within the community: on day one, everyone felt "we're done for, the field of mathematics is finished too," even considering urging graduate students to switch fields; by day three, the emotion turned to anger, aimed at how "OpenAI's doing this is just way too much."
The shift came on day five. Mathematicians began to accept reality—in research and proofs, AI had become an unavoidable tool. Many instead came to believe that their path as mathematicians would change, but they "hope it will be better than before." Altman himself is optimistic, believing AI will not cause an employment catastrophe.
The End of the Romantic Era and the Controversy
Altman's optimism differs greatly from the feelings of many mathematicians. In September this year, OpenAI suddenly announced that its model had cracked the Navier–Stokes equations problem, unsolved since the 19th century, igniting fury. Researchers at New York University and Anthropic accused the company of using their chat logs as "ammunition"; others were annoyed that AI gives only answers without showing derivations, reducing what's left for humans to tedious verification.
Beyond the uproar, a new drama emerged. A group of Caltech undergraduates launched a "math marathon," encouraging the use of AI to tackle long-unsolved hard problems. More than 700 scholars signed a joint letter criticizing it for encouraging AI-generated garbage content; OpenAI then withdrew its sponsorship, and after the competition changed its rules, it is planned to begin in November.
Terence Tao stated bluntly that the "romantic era" of mathematical research is over, with top colleagues shifting from solving problems by hand to verifying AI's answers. This month OpenAI claimed to have solved more than 700 math problems, and Tao lamented: "Something is being lost." Altman, for his part, concluded: he had never seen a field go through this entire set of changes so rapidly.