My CS professor spent a whole semester treating this one rule like a law of ph… · @loveAIwang · TwKit

@loveAIwang

My CS professor spent a whole semester treating this one rule like a law of physics 😭 Now an internal OpenAI model claims it broke it And there's a proof formalized in Lean Here's the part that gets me Russell Impagliazzo and Ramamohan Paturi helped come up with that rule back in 1999 Now they're helping lead the review, alongside MIT's Ryan Williams Imagine being asked to check an AI's homework that says one of your most famous ideas is wrong The rule is the Exponential Time Hypothesis Basically, problems like 3-SAT were believed to get exponentially harder as you add more variables, with no algorithm able to escape that kind of blowup Nobody proved it A whole lot of computer science was built around assuming it was true Hundreds of papers rely on it Before anyone loses it, this isn't P vs NP The new algorithm is still slow, and the review isn't finished If it holds up, a whole lot of what we thought we knew about computational hardness needs another look And if it doesn't, the fact that an AI got the people behind the hypothesis to examine its proof is still wild My old lecture notes are sweating

曝光 239 · 评论 2 · 点赞 8 · 书签 0