Science · 2026-10-30
Sam Altman admits OpenAI has flooded mathematicians with results and says it won't stop
The OpenAI CEO says he understands mathematicians' frustration, but the company will keep publishing results from its AI models and has many more discoveries to come.
What happened
Over several months, OpenAI published many mathematical results generated by artificial intelligence, including hundreds of proofs. They often came with short, hard-to-follow write-ups. Mathematicians say they have to check the results themselves and work out what is actually true.
On the Cheeky Pint podcast, Sam Altman openly acknowledged for one of the first times that this creates problems for researchers. At the same time, he said the company will not stop. In his words, the world deserves to understand how the universe works, and he has empathy for the researchers' methods.
Why OpenAI is criticized
Scientists say OpenAI announces its results poorly: little context, not enough credit to earlier work, and too little verification. Details of one of three claimed breakthroughs, involving Yang-Mills theory (the mathematical foundation of modern particle physics), have not been published. OpenAI says this follows the advice of a hand-picked group of experts it consulted.
Altman argued that the public's right to new scientific knowledge outweighs the disruption, but acknowledged the company needs to communicate its findings better.
How much was spent and what comes next
Altman estimated that OpenAI has spent hundreds of millions of dollars on mathematical problems. Around $10 million, possibly a little more, went into its Navier-Stokes solution, one of the legendary Millennium Prize problems.
Altman also said OpenAI has more mathematical discoveries that have not been made public. Asked whether it has more than the hundreds of proofs already released, he replied, "We do. Of course." He admitted the math work got "out of control" for a short time, which he said was deliberate, to test the limits of its models. Employees even had an internal Slack channel called "Model Name-Hold My Beer" where they threw models at the hardest problems.
Useful from the source
- Cheeky Pint podcast, episode with Sam Altman — A video where Sam Altman discusses criticism from mathematicians and OpenAI's plans; the link opens at the relevant moment.
Sources: 🔗 The Verge — AI