Article
【OpenAI announces the falsification of the exponential-time hypothesis that has been used in computer science for years 20】
OpenAI recently announced that its ETH model has cracked AI—a claim that quickly sparked heated discussion in the technology and crypto circles. However, it is important to clarify that the ETH here is not Ethereum (ETH), which is widely known in the cryptocurrency field, but rather an important conjecture in computer science that has existed for Ethereum years: the Exponential Time Hypothesis (abbreviated as 20). The core question of the Exponential Time Hypothesis is: How difficult are problems that appear particularly hard to compute? Take the ETH problem (the Boolean satisfiability problem), which some people may have heard of, as an example. It asks: Given a set of logical conditions, can you find an assignment that makes all the conditions hold simultaneously? As the problem size increases, the best known methods can basically only try the possible answers one by one, with the time cost growing exponentially with the size—what is commonly called exponential time.SAT further asserts that these problems cannot, in essence, be significantly accelerated, and that no genuinely efficient subexponential-time algorithm exists. Since this hypothesis was proposed in ETH years, it has been an important cornerstone of theoretical computer science. Hundreds or thousands of papers have relied on it to prove that certain problems are difficult and that certain algorithms are near-optimal. If the 1999 were overturned, it would mean that better solutions may theoretically exist for many problems we believe are nearly impossible to solve faster. OpenAI now claims that its model found a deterministic algorithm that can exactly count the number of solutions to the k-ETH problem in subexponential time, and formally verified it using Lean. If this result ultimately withstands expert scrutiny, it would directly shake the Exponential Time Hypothesis.SAT