OpenAI Says Its AI Disproved A Hypothesis—Here’s What That Means

🔍 Read the full analysis: OpenAI Says Its AI Disproved A Hypothesis—Here’s What That Means on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

OpenAI said October 26 that an unreleased internal AI model produced a proof of an algorithm that would refute the Exponential Time Hypothesis (ETH), a major assumption in theoretical computer science about the difficulty of solving 3-SAT. The preprint and a Lean formalization are public, but the result has not yet been independently reviewed; if confirmed, it would revise a widely used basis for proving problems computationally hard.

The preprint describes a deterministic algorithm that exactly counts satisfying assignments for k-SAT formulas, for each fixed k, in subexponential time and polynomial memory. For 3-SAT, that would contradict ETH, which holds that no algorithm can solve the problem in time 2^(c·n) for any fixed positive c, where n is the number of variables. The authors also say the result refutes the stronger Strong Exponential Time Hypothesis and three related hypotheses.

OpenAI says its system found the proof on October 16, with a Lean formalization produced the following day. The preprint is dated October 17 and was revised October 25. Lean is software that checks whether a formal proof follows from its stated definitions and rules. The source account reports a formalization spanning 3,977 files and about 710,000 lines. That check is meaningful evidence about the formalized statement, but it is not the same as independent experts assessing the argument, its significance or how it relates to the paper’s claims.

There is also a scope difference: the formalization checks a slightly weaker runtime bound than the sharpest bound in the preprint, according to the source account. An external group led by computer scientists Russell Impagliazzo, Ramamohan Paturi and Ryan Williams is expected to examine the result. Impagliazzo and Paturi were among the original authors of ETH. OpenAI says the group will explain and verify the work, and that a conference on the result will receive OpenAI funding.

A New Test for Hardness Claims

ETH is used across theoretical computer science to show that many algorithms cannot run substantially faster than known approaches, unless the hypothesis fails. A proof against it could require researchers to revisit results whose hardness arguments rely on ETH, or to identify which conclusions still follow from weaker assumptions. The reach would depend on the proof and on the specific assumptions behind each result; the announcement alone does not invalidate every ETH-based finding.

The immediate importance is also methodological. OpenAI is presenting a chain in which an AI system generates a mathematical argument and a proof assistant checks a formal version, followed by review from human specialists. Formal checking can help catch logical gaps in what has been encoded, while expert review is still needed to assess whether the encoding matches the intended theorem and whether the result is sound and correctly understood. That distinction matters as AI systems are used to produce research claims.

The reported algorithm is not a practical shortcut for ordinary computing. The source account gives its running time as roughly 2^(n/√log n), which remains superpolynomial and grows rapidly as problem size increases. OpenAI itself says practical improvements would require significant further research. There is no basis in the announcement for saying that current software or hardware will suddenly solve large SAT instances faster.

Products Worth Considering

Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor

AmazonView Latest Price

LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor

AmazonView Latest Price

Apple iPad Air (4th Generation, 10.9-inch, Wi-Fi, 256GB) – Space Gray

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

What ETH Says About 3-SAT

In k-SAT, a formula is made from rules, each involving at most k variables or their negations. The task is to determine whether values can be assigned to the variables so that every rule is satisfied; counting versions ask how many such assignments exist. With n variables, trying every possible assignment requires considering as many as 2^n combinations.

Researchers have developed algorithms that improve on brute force by reducing the exponential base. ETH, proposed by Impagliazzo, Paturi and Francis Zane in 2001, makes a stronger claim: 3-SAT still requires exponential time, in the sense that it cannot be solved in 2^(o(n)) time. The new preprint’s claimed subexponential algorithm would contradict that claim.

This is not a solution to P versus NP. P versus NP asks whether every problem whose proposed solution can be checked efficiently can also be solved efficiently. ETH is a stronger kind of hardness assumption; rejecting it does not establish that P equals NP. The Clay Mathematics Institute continues to list P versus NP as unsolved.

The development comes amid a broader OpenAI research release effort. The source account says OpenAI published more than 700 AI-generated manuscripts on October 6, and that three were withdrawn and 14 revised the next day after a sign error was found. It also reports that about 42% of the repository’s top-line results were formalized in Lean. Those details do not establish whether this proof is correct, but they show why formalization and independent scrutiny are central to evaluating the current claim.

Review and Formal Scope Remain

The central open question is whether independent experts will verify the argument. Lean checking is not independent peer review: it checks the formal proof supplied to it, while reviewers must also examine the paper’s claims, the formalization’s scope and any connection between the two. The source account says the formal result is weaker than the preprint’s sharpest runtime bound, leaving the status of that strongest version particularly important to clarify.

The review has not yet concluded, and the source material does not provide a timetable or a final assessment from the external group. It is also unclear which ETH-dependent results would need revision if the claim is upheld. Their implications would have to be examined individually rather than inferred from the headline result.

Independent Review Will Settle the Claim

The next milestone is the examination by the group led by Impagliazzo, Paturi and Williams. Its members are expected to explain the result and assess the proof; a conference funded by OpenAI is also planned, according to the source account. Public materials include the preprint and Lean code, allowing researchers to inspect the work.

Readers should look for a clear account of whether the paper’s strongest runtime claim is supported, how it compares with the Lean-checked statement, and whether outside reviewers accept the reasoning. Until those questions are answered, the accurate description is that OpenAI has announced a machine-checked proof that claims to refute ETH, not that the research community has confirmed the hypothesis is false.

Key Questions

What does OpenAI say its AI proved?

OpenAI says an internal AI model produced an algorithm for exactly counting satisfying assignments to fixed-k SAT formulas in subexponential time. For 3-SAT, that would contradict the Exponential Time Hypothesis.

Has the proof been independently confirmed?

No independent review is reported as complete. Lean has checked a formalized statement, but an outside group is expected to examine the result. The formalization reportedly covers a slightly weaker runtime bound than the preprint’s sharpest claim.

Does this solve P versus NP?

No. ETH is a hardness hypothesis stronger than the basic claim that P is not equal to NP. A refutation of ETH would leave P versus NP unresolved.

Will this make SAT problems faster to solve now?

There is no evidence of an immediate practical speed-up. The claimed running time remains superpolynomial, and OpenAI says practical improvements would require significant further research.

Why does the result matter if it is theoretical?

ETH is used to support hardness results across theoretical computer science. If the proof is upheld, researchers may need to reassess conclusions that rely on ETH, while also examining which results can be established under weaker assumptions.

Source: ThorstenMeyerAI.com