Skip to content
MaungaAI news, explained

Research

OpenAI says its AI refutes an old computer science hypothesis, and it has no plans to slow down

OpenAI claims its model has solved three problems in mathematics, computer science and physics, and it is now giving outside experts a voice. Its CEO, Sam Altman, understands why mathematicians feel overwhelmed, but he has no intention of stopping.

2 min read3 sources, one officialTranslated with AI from the Spanish original

Share: WhatsApp · Email

AI-generated illustration for Maunga. Not a photograph.

The essentials

  • On October 26, OpenAI published results from its model on the Kakeya conjecture, the exponential time hypothesis and quantum chromodynamics (OpenAI News, The Verge).
  • Claimed by one partyAccording to OpenAI, its model solves certain logic puzzles faster than was thought possible, which would refute a basic assumption of computer science (TechCrunch).
  • Outside experts are now guiding the computer science and physics results. Computer scientists have asked OpenAI to pay for a conference to debate them (The Verge, TechCrunch).
  • Claimed by one partySam Altman, OpenAI's CEO, said on a podcast released on October 29 that he understands the mathematicians, but that he won't stop (The Verge).

Why it matters

If the proof holds up, many papers that took the hypothesis for granted will have to be reviewed, according to Williams (TechCrunch). It would also lend a little more credence to the idea that very hard problems are easier to solve than thought. And Altman's words suggest the new approach won't stop the flood: review would keep chasing after the results.

The details

The exponential time hypothesis holds that, for a certain logic puzzle, even the fastest program takes vastly longer as the number of variables grows (TechCrunch). Levent Alpöge, a mathematician at Anthropic, hinted that OpenAI built on earlier work; hours later, Anthropic published its own refutation (TechCrunch). Ryan Williams, a professor at MIT, sees "two or three strange-looking parts" and believes the method can't be the best possible one (TechCrunch).

In early October, OpenAI released hundreds of manuscripts all at once, and Williams called it a cyberattack on mathematicians (The Verge, TechCrunch). This time, the math is on GitHub, the computer scientists have published an open letter and the physicists will decide how their results come out (TechCrunch). Even so, researchers interviewed by The Verge are tired of the company (The Verge). Altman argues that "the world deserves to understand how the universe works" (The Verge).

What we don't know

  • Whether the proofs will hold up under review by independent experts.
  • How far the model built on earlier work, as a mathematician at Anthropic hints.
  • How many more results OpenAI is holding back. Altman implies there could be many.

How this story has changed

  1. Sam Altman has said on a podcast that he understands why mathematicians feel overwhelmed, but that OpenAI won't stop and may have more results held back.
  2. We published the story.

Sources

This story is based on 5 sources. In the text, each fact names in brackets who reports it.

Was this story useful? Yes Not really

Spanish original