OpenAI says its AI found a proof challenging the Exponential Time Hypothesis

In short: OpenAI says its AI system produced a formal proof that refutes the Exponential Time Hypothesis, a key idea about how hard some problems are, and researchers are now checking the claim.

What happened

OpenAI said its AI system created a mathematical proof that refutes the Exponential Time Hypothesis, often shortened to ETH. ETH is a long-standing guess in computer science that certain “satisfiability” problems (logic puzzles where you ask if there is any set of yes or no choices that makes all rules true) cannot be solved much faster than exponential time.

OpenAI said the result was formalized in Lean, which is a proof-checking system. You can think of Lean as a strict spellchecker for math proofs that checks every step, instead of relying on a human reader to fill in gaps.

Independent experts are still examining the proof and what it really implies. OpenAI also said that turning this kind of theory into real-world speedups for software will take substantial further research.

Why it matters

ETH has been used like a support beam in many areas of research, including graph theory, computational biology, databases, and computational geometry. If the proof holds up, some “if ETH is true, then X is impossible” results may need to be revisited, and it could point to new ideas for faster algorithms (step by step recipes for solving problems). Still, this is not the same as showing that all hard problems suddenly become easy, it does not prove P equals NP, and it does not mean your current apps will run faster tomorrow.

Source: NYTimes