Altman Says OpenAI Will Keep Pursuing Major Math Problems
Altman said he feels conflicted about AI solving some of mathematics’ hardest problems. He argued that scientific knowledge should be shared, while expressing empathy for researchers and concern about how the work affects their communities. He said OpenAI is working out how to pursue research while helping specialists benefit from what comes next.
OpenAI has drawn criticism for using its models to tackle difficult mathematical questions. The company withdrew from a Caltech math hackathon after researchers objected, and 28 Fields Medal winners signed a letter warning that treating mathematical problems as AI benchmarks could harm the field and its community.
The company attracted attention by saying its models had found a counterexample involving the Navier-Stokes equations, which describe the movement of liquids and gases. The problem is among the Millennium Prize Problems, for which the Clay Mathematics Institute offered $1 million for a solution. Altman said the work involved about $10 million, possibly more.
Altman said OpenAI had spent hundreds of millions of dollars on mathematics overall and had more discoveries than the hundreds of results it had already released. He acknowledged that the effort briefly “got out of control,” describing employees using an internal Slack channel to direct models at difficult research problems.
OpenAI has also faced criticism over how it shares results, including complaints about limited context and verification. Some researchers have objected to the company withholding details of a claimed Yang-Mills theory result, which OpenAI says it is handling in line with advice from experts it consulted. Altman said he wants the company to be more sensitive to research communities, while maintaining that major scientific discoveries should not be kept from the public.
OpenAI’s Math Push Risks Leaving Researchers Behind
OpenAI’s push to apply AI to mathematics is disrupting the people who built the field, and Sam Altman’s sympathy has not changed the company’s course. He said the public deserves scientific knowledge, but also recognized that researchers are struggling with the effects of OpenAI’s work.
Mathematicians have objected to the company’s use of difficult problems as AI benchmarks. OpenAI withdrew from a Caltech math hackathon after objections, while 28 Fields Medal winners warned that the practice could damage mathematics and its community.
The criticism also concerns how OpenAI presents results. Researchers have complained about inadequate context, attribution and verification, leaving them to assess the work. Some have also challenged the company’s decision to withhold details of a claimed Yang-Mills theory breakthrough. OpenAI says that choice followed advice from a group of experts it consulted.
Altman said OpenAI has spent hundreds of millions of dollars on mathematics, including around $10 million, perhaps more, on its Navier-Stokes result. He said the company has more discoveries than the hundreds of results it has released, even as researchers face a flood of work to check and interpret.
Altman said OpenAI is trying to find a more considerate approach to research communities. But his insistence that discoveries should not be held back leaves the central tension unresolved: access to knowledge matters, and so does whether scientists can properly verify, understand and share credit for it.
Altman Says Scientific Discoveries Should Not Be Sidelined
Sam Altman’s message is that scientific progress should not be shelved because it unsettles established research communities. The OpenAI chief said he empathizes with mathematicians, but argued that people deserve fundamental knowledge about how the universe works.
That principle is being tested as OpenAI uses AI models to tackle major mathematical questions. The company says its models found a counterexample involving the Navier-Stokes equations, one of the Millennium Prize Problems. Altman said OpenAI spent around $10 million, possibly more, on that effort, and hundreds of millions of dollars on mathematics overall.
Researchers have criticized the company’s releases, including concerns about limited context and verification. OpenAI has also withheld details of a claimed Yang-Mills theory result, saying experts advised it on how to share the finding. That decision has itself drawn complaints from researchers who want access to the work.
Altman said OpenAI has more mathematical discoveries than the hundreds of results already made public. He acknowledged that the effort briefly got out of control, as employees directed models toward difficult research problems to test their abilities.
The company still needs to communicate its findings better, Altman said, and to consider the communities that advanced mathematics. But his position is clear: the answer to difficult questions about AI’s impact cannot be to keep major scientific discoveries from the public.
Reported by Business Insider, The Verge.
