Making Sense Of OpenAI’s AI Claim About A Fundamental Complexity Hypothesis

🔍 Read the full analysis: Making Sense Of OpenAI’s AI Claim About A Fundamental Complexity Hypothesis 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 on October 26 that an unreleased internal AI model had produced a proof of an algorithm that would refute the Exponential Time Hypothesis (ETH), a foundational conjecture about the difficulty of solving certain logic problems. A Lean formalization has been machine-checked, but independent expert review has not been completed, and the formalized statement is weaker than the fastest bound claimed in the preprint.

The 67-page preprint claims a deterministic algorithm that exactly counts satisfying assignments for any fixed k-SAT problem in subexponential time, using polynomial memory. For 3-SAT, such an algorithm would contradict ETH, which says that solving 3-SAT requires time exponential in the number of variables, up to a positive constant in the exponent. The preprint also says the result refutes the stronger Strong Exponential Time Hypothesis and three related hypotheses.

OpenAI says its model found the computer-science result on October 16, and that the Lean formalization followed the next day. The preprint is dated October 17 and was revised October 25. The source account reports that the public formalization contains 3,977 files and about 710,000 lines, while the formal statement itself is short. Those figures describe the released project; they do not establish that outside researchers have verified the proof’s significance or all of its claims.

A key qualification is that Lean checks a weaker runtime bound than the preprint’s sharpest claimed result. Lean is proof-assistant software that can verify whether a formal proof follows from its stated definitions and steps. That check is evidence about the formalized theorem, not an independent assessment of the paper’s strongest claim, the research assumptions, or the result’s implications. OpenAI said an external group led by researchers including Russell Impagliazzo and Ramamohan Paturi will examine the work.

A Challenge to SAT Hardness

If the result holds as stated, it would overturn ETH, a widely used assumption in theoretical computer science for proving that many computational problems are hard. Researchers often use ETH to show that an algorithm for one problem cannot run too quickly unless it also yields an unexpectedly fast method for solving 3-SAT. A refutation would require rechecking results that depend on that assumption, though it would not automatically invalidate every such result: some may rest on other assumptions or have independent proofs.

The claimed algorithm is subexponential, not polynomial-time. The source gives a runtime roughly 2 to the power of n divided by the square root of log n, where n is the number of variables. That still grows much faster than a polynomial as problem size increases. OpenAI’s announcement, as relayed in the source, says practical improvements would require significant further research; it does not establish that current software or commercial workloads will become faster.

The development also draws attention to a possible research workflow: an AI system produces a mathematical argument, formal software checks a coded version, and specialists review the result. That is a consequential demonstration if validated, but it is not yet proof that AI-generated research can reliably settle major open questions. Formal checking and independent mathematical review answer different questions, and both matter here.

Popular right now

KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream

AmazonView Latest Price

De’Longhi Classic Matte Black Espresso Machine with Milk Frother

AmazonView Latest Price

Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

What ETH Says About 3-SAT

In k-SAT, each rule is a logical clause involving at most k variables, and the task is to determine whether assignments of true or false can satisfy all clauses. A brute-force search over n variables examines as many as 2ⁿ assignments. Researchers have developed faster algorithms, but the longstanding question is whether the exponential dependence can be removed or reduced to subexponential time.

The source traces ETH to a 2001 paper by Russell Impagliazzo, Ramamohan Paturi and Francis Zane. ETH says that 3-SAT cannot be solved in subexponential time in the relevant sense. It is stronger than the claim that P is not equal to NP: refuting ETH would not show that every problem with efficiently checkable answers can be solved efficiently. The Clay Mathematics Institute continues to list P versus NP as unsolved.

OpenAI’s October 26 post announced three research claims, according to the source: the ETH result in computer science, a proof concerning local smoothing in harmonic analysis, and a physics result involving quantum chromodynamics and a proposed completion of the Standard Model. The release arrangements differed by field. For the ETH work, OpenAI made the preprint and Lean code public at the external group’s request; the source says materials for the physics work were held back. These are OpenAI’s reported claims and release choices, not independent confirmation of the scientific results.

The release follows a larger publication of AI-generated manuscripts on October 6. According to the source, OpenAI’s repository history records the withdrawal of three manuscripts over a sign error and revisions to 14 others on October 7. That history helps explain why formalization and outside scrutiny are relevant, but it does not determine whether this separate ETH proof is correct.

Review Still Has to Test the Proof

The central open question is whether independent specialists agree that the proof is correct and that it establishes the claims made in the preprint. The source says a group led by Impagliazzo and Paturi, two of ETH’s three original authors, will examine it. No completed review or final verdict is reported in the material provided.

There is also a specific gap between the paper and the formal check: the sharpest runtime bound is not covered by the Lean formalization. The formalized, weaker result may still be important, but readers cannot treat the machine check as confirmation of every version of the theorem presented in the preprint. The scope notes and precise relationship between the formal statement and the paper’s strongest bound will need close scrutiny.

The source does not provide the full proof details, the external reviewers’ findings, or a timeline for their assessment. It also reports that views on ETH itself vary; a result challenging the hypothesis is not the same as a consensus that every application of ETH was misguided. Until the review is public, the accurate description is an announced claim supported by a machine-checked formalization, not a settled refutation.

Independent Review and Formal Scope

The next milestone is the external group’s examination of the preprint and the Lean code. Readers will need to see whether reviewers endorse the claimed algorithm, whether the formalized result captures the paper’s main contribution, and whether the stronger runtime bound can also be verified. OpenAI has said it will fund a conference on the result, which could provide a venue for researchers to question and assess the work.

Until those steps produce public findings, the practical consequences remain limited. The claim does not make P versus NP solved, and the stated algorithm does not offer a polynomial-time solution. The development to watch is not an immediate speed-up in applications but whether specialists can independently validate the mathematical argument and clarify exactly which complexity assumptions it overturns.

Key Questions

What is OpenAI claiming its model proved?

OpenAI says its internal model produced a proof of a deterministic, subexponential-time algorithm for exactly counting solutions to fixed-k k-SAT problems. For 3-SAT, that would contradict ETH, if the proof and its stated result hold.

Has the result been independently confirmed?

Not according to the source material. A Lean formalization has been machine-checked, and an external group is expected to review the work. The source does not report a completed independent review.

Does this solve P versus NP?

No. ETH is a stronger hypothesis about the difficulty of 3-SAT. Even a valid refutation of ETH would not by itself prove that P equals NP, and the P-versus-NP problem remains open.

Will this make software or AI systems faster now?

No immediate practical improvement is established. The claimed algorithm remains superpolynomial, and OpenAI said practical improvements would require significant further research.

What is the main limitation of the Lean check?

The source says Lean checks a weaker runtime bound than the sharpest one claimed in the preprint. The formalization therefore should not be described as checking every part of the strongest claim.

Source: ThorstenMeyerAI.com