A Closer Look At OpenAI’s AI Claim On The Exponential Time Hypothesis

🔍 Read the full analysis: A Closer Look At OpenAI’s AI Claim On The Exponential Time Hypothesis on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

OpenAI said on October 26 that an unreleased internal AI system had produced a proof of an algorithm that, if correct, would refute the Exponential Time Hypothesis (ETH), a long-standing conjecture about the difficulty of solving Boolean satisfiability problems. A Lean formalization of the proof is public, but the claim has not yet received independent expert sign-off, and the formalized result covers a weaker runtime bound than the preprint’s sharpest claim.

The preprint describes a deterministic algorithm that exactly counts satisfying assignments for any fixed-k SAT formula in subexponential time and polynomial memory. For 3-SAT, that would contradict ETH, which holds that solving such instances requires time exponential in the number of variables. The authors also say the result refutes the stronger Strong Exponential Time Hypothesis and three related hypotheses. These are claims in the preprint, not conclusions independently established by outside reviewers.

OpenAI says its system found the disproof on October 16, and that the Lean formalization followed a day later. The public code contains 3,977 files and about 710,000 lines, according to ThorstenMeyerAI.com’s count. Lean is a proof assistant that checks whether a formal proof follows from its definitions and rules. That check supports the formalized statement; it does not by itself establish that the paper’s strongest version has been checked or that the work has passed academic review.

The account says the preprint was revised on October 25. OpenAI has made the paper and code public at the request of an outside group that will explain and examine the computer-science result. The group is led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams. Impagliazzo and Paturi were among the original authors of the hypothesis. OpenAI says it will fund a conference on the result, and that its collaborators are unpaid, sign no nondisclosure agreements and speak only for themselves.

What an ETH Refutation Would Change

If independently confirmed, the result would force researchers to revisit a widely used assumption in theoretical computer science. ETH has helped researchers establish lower bounds by showing that many problems are unlikely to admit algorithms that are dramatically faster than known approaches. A valid refutation could affect those arguments and prompt new work on which hardness results still hold under different assumptions.

The immediate significance is primarily theoretical. The preprint’s described running time is roughly 2 to the power of n divided by the square root of log n: subexponential, but still much slower than any polynomial-time algorithm. OpenAI itself says practical improvements would require significant further research. The claim does not mean ordinary SAT instances, commercial workloads or other difficult computing tasks will become easy in the near term.

It also does not solve P versus NP. That question asks whether every problem whose proposed answer can be checked quickly can also be solved quickly. ETH is a stronger statement about the time needed for a particular family of problems. Refuting ETH would weaken that stronger assumption while leaving the P-versus-NP problem open.

The result also tests a process for evaluating AI-generated research: the system produces a mathematical claim, formal tools check a specified proof, and independent specialists scrutinize the work. Each stage has limits. A machine-checked proof can reduce the risk of mistakes in what was formalized, while human review is needed to assess the argument’s scope, its relation to the paper and its implications.

Popular right now

Breville Barista Pro Espresso Machine with Grinder, Brushed Stainless Steel

AmazonView Latest Price

Dyson Ball Animal 3 Extra Upright Vacuum

AmazonView Latest Price

Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

How ETH Became a Hardness Benchmark

SAT asks whether there is an assignment of true and false values that makes every logical rule in a formula true. In k-SAT, each rule contains at most k variables or their negations. With n variables, a direct search may examine 2ⁿ possible assignments, although researchers have developed algorithms that improve on brute force.

ETH grew out of work by Russell Impagliazzo, Ramamohan Paturi and Francis Zane in 2001. It proposes that 3-SAT cannot be solved in subexponential time. The preprint’s claim is not that SAT becomes easy, but that an algorithm can remove the exponential character of the running time for fixed k. The supplied source notes that researchers have long improved the constant in the exponent without eliminating that exponential dependence.

OpenAI presented the ETH paper alongside two other research announcements on October 26: a claimed proof of a local smoothing conjecture in mathematics and a physics result involving quantum chromodynamics and a proposed completion of the Standard Model. OpenAI says outside experts are guiding how each result is released. The different release choices underline that these are research claims at different stages of scrutiny, not a single independently validated package of results.

The announcement follows a large release of AI-generated manuscripts on October 6, according to the source. Repository history records that three papers were withdrawn after a sign error and 14 others were revised on October 7. That history makes the distinction between a manuscript’s claim, a formalized statement and independent review particularly relevant to assessing this new result.

“there’s by no means a consensus in the field that ETH is true”

— Scott Aaronson, in a 2017 survey, as quoted by ThorstenMeyerAI.com

The Proof’s Remaining Review

The central open question is whether independent reviewers will validate the result and its full scope. The formalization reportedly checks a weaker runtime bound than the sharpest bound stated in the preprint. The available account does not specify here the exact gap between those bounds, or whether the authors expect the stronger result to be formalized later.

Lean verifies the proof encoded in its formal system; it does not automatically establish that every claim in a research paper follows from that formal statement. Reviewers still need to examine the relationship between the code and the paper, the assumptions used, and the conclusions drawn. The source reports that the expert group will examine the work, but does not provide a completed review or a timetable for one.

The degree to which the result would alter practical algorithms is also unknown. Even if the theoretical claim holds, the stated running time remains superpolynomial, and no implementation or real-world performance gain is established in the source material. The status of the other two announcements is separate: their claims should not be treated as independently confirmed by the ETH formalization.

Independent Review and Formal Scope

The next milestone is the examination by the outside group led by Impagliazzo, Paturi and Williams. Its assessment should clarify whether the paper’s argument is correct, how much of the claimed runtime is covered by the Lean code, and whether the formalized result supports the stated consequences for ETH and related hypotheses.

Readers can compare the public preprint with its Lean formal statement and code while that review proceeds. Further revisions, a fuller formalization of the sharper bound, or a response from the reviewers could change how the result is described. Until then, the accurate account is that OpenAI has announced a potential refutation, a version of its proof has been machine-checked, and independent confirmation remains pending.

Key Questions

What does OpenAI say its AI proved?

OpenAI says its internal AI system found an algorithm for exactly counting solutions to fixed-k SAT formulas in subexponential time and polynomial memory. For 3-SAT, that would contradict the Exponential Time Hypothesis, if the proof and its stated implications are correct.

Has the proof been independently verified?

Not yet, according to the supplied account. A Lean formalization is public and an outside group is set to examine the result, but the group’s review has not been reported as complete.

Does this solve P versus NP?

No. ETH is a separate, stronger hypothesis about the time required to solve SAT problems. A refutation of ETH would not resolve the P-versus-NP question, which remains open.

Does the Lean check cover the strongest claim?

The source says the formalization checks a weaker runtime bound than the sharpest one in the preprint. The exact relationship and whether a stronger formalization will follow remain unclear.

Will this make computing faster soon?

No near-term practical improvement has been established. The claimed algorithm is still far slower than polynomial time, and OpenAI says practical gains would require significant further research.

Source: ThorstenMeyerAI.com