OpenAI says its AI has disproved ETH, and no, not Ethereum. ETH here is short for the Exponential Time Hypothesis, a decades-old computer science assumption. It says some hard computing problems will always take exponentially longer to solve as they grow. OpenAI says the hypothesis was widely believed to be true and has been used in hundreds of research papers. But its AI came up with an algorithm that solves these problems faster than the hypothesis allowed. The AI found the disproof on October 16, and Lean, a program that checks math proofs, verified the result a day later. OpenAI has published the proof and its Lean check on GitHub. Russell Impagliazzo and Ramamohan Paturi, two of the scientists who first proposed the hypothesis, will lead a group with MIT's Ryan Williams to verify the work. OpenAI says turning the result into faster real-world algorithms "will require significant further research." The company also announced two more AI discoveries. In math, OpenAI says its models resolved the Kakeya conjecture in every dimension. Mathematicians only proved the 3D version last year. In physics, OpenAI says its AI produced a proof covering part of one of math's $1 million Millennium Prize problems. A group led by Nobel Prize winner David Gross will study that work, which hasn't been released yet. Would you trust a proof written by AI before humans check it?