Altman Says Mathematicians Have Moved From 'We're Doomed' to Embracing AI, but Academic Debate Rages On

OpenAI CEO Sam Altman recently spoke to several hundred students at Stanford University. He compared how mathematicians have reacted to artificial intelligence (AI) to the different stages of grief. He said mathematicians were shocked at first, then became angry, and finally started accepting AI as a tool that is now an important part of mathematical research.

According to Altman, on the first day, mathematicians reacted by saying, "We're doomed, and the field is doomed." Some even told graduate students to change what they were studying. By the third day, their fear had turned into anger. Some felt that "what OpenAI is doing is really too much." By the fifth day, more people were starting to accept that their work and future as mathematicians would change. They still hoped their future would be better than before.

But not everyone in mathematics shares Altman's positive view. In September, OpenAI suddenly announced that one of its AI models had solved a famous problem involving the Navier-Stokes equations. This problem has troubled some of the world's best mathematicians since the 1800s. After the announcement, several well-known mathematicians publicly said they were unhappy. Two researchers from New York University and Anthropic accused OpenAI of using information from their conversations with AI to help solve the problem. Other mathematicians complained that even if the AI really had found an answer, it had not shown the steps used to reach it. That left humans with the task of checking whether the answer was correct, which they felt was much less exciting than solving the problem themselves.

The disagreement continued. A group of undergraduate students at the California Institute of Technology worried that AI was changing mathematical research too quickly. They wanted to keep up with the technology, so they started a competition called the "Math Marathon." The competition encouraged people to use AI to work on mathematical problems that had remained unsolved for a long time. But more than 700 academics signed an open letter criticizing the competition. They said it encouraged people to produce large amounts of low-quality AI-generated work. After this criticism, OpenAI stopped sponsoring the event. The organizers later changed the competition rules, and the event is planned for November.

Recently, Terence Tao, a professor at the University of California, Los Angeles, said directly that the "romantic era" of mathematical research is over. Many leading mathematicians have started changing how they work. Instead of solving difficult problems on their own, they are increasingly checking answers created by AI. This month, OpenAI said it had solved more than 700 mathematical problems. In response, Tao said, "Something is being lost."

Altman also has a positive view of how AI will affect jobs. He said he does not think AI will cause a huge employment crisis. However, he admitted that mathematics is feeling the effects of AI much faster than other areas of study and other industries. He said, "I think this is the first time all of us have seen a field go through this whole cycle of change so quickly."

From the technology industry's point of view, the main question is no longer just who solves a mathematical problem. It is also about what role AI should play in basic scientific research. Should AI simply help researchers, or should it do the research itself? For OpenAI, solving mathematical problems is one of the clearest ways to show that its AI models can reason through difficult questions. Every time OpenAI announces that its AI has solved a problem, it strengthens the company's message about how powerful its technology is. But universities and researchers see another side of the issue. If an AI gives an answer without explaining how it reached that answer, it is difficult for other researchers to learn from it and build on it. There are also new disagreements about who should receive credit for the work, where the AI got its information, and who is responsible for checking whether its answers are correct.

These disagreements are now affecting other areas of scientific research as well. As leading researchers move from being people who solve problems to people who check AI-generated answers, the way research is judged may change. The way future researchers are trained and the types of research that receive funding may also change. For now, it is difficult to know whether mathematicians have truly reached the "fifth day" of acceptance that Altman described. Mathematicians still have questions about where OpenAI gets its data. The Math Marathon organizers have had to change their competition rules, and researchers such as Tao feel that something valuable is being lost. All of this shows that accepting AI as a reality does not necessarily mean agreeing with the new way it is changing mathematical research.