Altman: OpenAI Spent Hundreds of Millions on Math — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Sam Altman said during Thursday’s Cheeky Pint podcast episode that humanity deserves fundamental knowledge and the public’s right to new scientific knowledge outweighs disruption from OpenAI’s releases, while acknowledging that communication must improve.
- OpenAI released hundreds of mathematical results this month, often in brief writeups that mathematicians speaking to The Verge criticized as poorly written, hard to follow and deficient in context, attribution and verification.
- OpenAI announced three major breakthroughs across mathematics, physics and computer science but withheld public details for its claimed Yang-Mills theory result, citing recommendations from a hand-picked expert group — a choice contrary to advice from a previous mathematician panel.
- Sam Altman confirmed OpenAI has more mathematical discoveries than the hundreds of proofs it has already released, answering, “We do. Of course,” when asked whether a larger stockpile exists.
- OpenAI spent, in Altman’s estimate, hundreds of millions of dollars on mathematical problems, including around $10 million — “maybe a little more” — on its Navier-Stokes solution, one of the Millennium Prize problems.
- OpenAI deliberately tested its models on the hardest mathematical problems through an internal Slack channel called “Model Name-Hold My Beer”; Altman said this short “out of control” phase was necessary to probe the models’ limits.
Why it matters: OpenAI released hundreds of mathematical results this month and spent roughly $10 million on one Navier-Stokes solution, giving researchers more material to examine. The control tradeoff is sharp: outside scientists must verify terse writeups themselves, while details of the claimed Yang-Mills result remain unavailable to the public.
Ask SkimNews
