Altman Sympathizes With Mathematicians but Says OpenAI's AI Math Releases Will Continue
OpenAI CEO Sam Altman says he has empathy for mathematicians overwhelmed by the company's AI-generated results, but he will not stop publishing. He also said OpenAI holds more discoveries that remain unreleased.
OpenAI chief executive Sam Altman, who leads the artificial intelligence company behind ChatGPT, said he feels empathy for mathematicians struggling to keep up with the stream of AI-generated results from his company. Speaking on the Cheeky Pint podcast in an episode released Thursday, he made clear the releases will go on and hinted that much more is still unpublished. The account comes from The Verge.
"I think, on net, the world deserves to understand how the universe works," Altman said, adding that he has empathy for the method. In his view, the public's right to new scientific knowledge outweighs the disruption caused by OpenAI's releases, though he acknowledged the company must communicate its findings better. He said humanity deserves fundamental knowledge and it should not be withheld.
Criticism from mathematicians
Mathematicians who spoke to The Verge have repeatedly criticized how OpenAI announces results, citing inadequate attribution, little context and insufficient verification. This month alone the company has released hundreds of mathematical results, often with brief writeups that researchers describe as poorly written and hard to follow.
At the same time, OpenAI faces the opposite complaint: that it is holding back research. After announcing three major breakthroughs in mathematics, physics and computer science, it has not made public the details of one result involving Yang-Mills theory, the mathematical foundation of modern particle physics. OpenAI says this follows the recommendation of a hand-picked group of experts, which runs counter to the advice of an earlier group of mathematicians it consulted.
Spending and unreleased results
Asked whether the company has more proofs than the hundreds already released, Altman answered: "We do. Of course." He conceded that the company's pursuit of mathematical problems "got out of control" for a short time, but said this was deliberate, to test the limits of its models. He described an internal Slack channel where employees threw the models at the hardest problems they could.
Altman estimated OpenAI has spent hundreds of millions of dollars on mathematical problems, including about $10 million on its Navier-Stokes solution, one of the Millennium Prize problems. Researchers worry such resources could price them out of their own fields.
