OpenAI says its AI proved a famous computer science rule is false
Hundreds of research papers relied on the rule. Outside experts, including two who first proposed it, will now check the AI's proof.
OpenAI says its AI has proved a famous computer science rule wrong. It shared the news in a post on October 26, along with two other new results in math and physics.
The rule is called the Exponential Time Hypothesis. Roughly, it says one classic kind of hard puzzle for computers cannot be solved much faster than by checking almost every possible answer. Experts widely believed it was true. Hundreds of research papers were built on it.
OpenAI says its system proved the rule is false. It also says a computer proof checker called Lean confirmed the result.
Now outside experts will check and explain the work. Two of them first proposed the rule, and they have written about it for 30 years. OpenAI also agreed to make all the raw data public and to pay for a conference about it.
The math result is about how waves behave as they focus. OpenAI says its models proved an idea first put forward in 1991, in every dimension.
The physics result covers part of a famous unsolved problem from the Clay Mathematics Institute. A group led by David Gross, who shared the 2004 Nobel Prize in Physics, will study it. That group chose to write the work up itself and hold back the AI's files for now.
OpenAI says it does not want academic credit for these discoveries. The experts are not paid and are free to share what they find.
Open it and check the claim yourself. That is the point.