OpenAI تقول إن نماذجها نقضت فرضية عمرها قرابة 30 سنة في علوم الحاسب، والخبراء للحين يتحققونOpenAI says its models disproved a nearly 30-year-old computer science assumption, and experts are still checking

OpenAI says its AI models have disproved the exponential time hypothesis (ETH), a fundamental computer science assumption proposed nearly 30 years ago, New Scientist reports. The result was announced in a blog post on 26 October alongside other claimed breakthroughs in maths and physics.

ETH tries to put a limit on how quickly computers can solve certain logic puzzles. The simplest approach is to try every combination of true and false, and ETH says that even with shortcuts, the work has to grow exponentially as the puzzle gets bigger. OpenAI says its models came up with an algorithm that beats that limit. Fedor Fomin at the University of Bergen in Norway called it a complete surprise and a shock, and said a textbook he wrote on the subject would become obsolete if the finding holds.

OpenAI says it verified the proof with Lean, software that checks mathematical proofs. Stefan Szeider at TU Wien expects the result to withstand scrutiny, but says researchers still need to check that the formal statement means what its authors say it means. OpenAI has commissioned an independent group to examine the findings, including Russell Impagliazzo and Ramamohan Paturi, who helped propose ETH.

Some academics are uneasy about the pace of announcements; Magnus Wahlström at Royal Holloway, University of London, calls it "mathematics done by press". And according to Leslie Goldberg at the University of Oxford, the result has no direct impact on the famous P versus NP problem, so it should not be seen as progress on it.