OpenAI’s AI Research And The Limits Of The Exponential Time Hypothesis Claim

🔍 Read the full analysis: OpenAI’s AI Research And The Limits Of The Exponential Time Hypothesis Claim on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the little things that make your day 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 produced a proof refuting the Exponential Time Hypothesis (ETH), a central conjecture about the difficulty of solving certain logic problems. A Lean formalization has been machine-checked, but the paper’s strongest runtime claim is not covered by that formalization, and independent experts have not yet signed off.

The 67-page preprint claims a deterministic algorithm that exactly counts solutions to any fixed-width k-SAT formula in subexponential time and polynomial memory. For 3-SAT, that would contradict ETH, which says such problems require exponential time. The authors say the result also refutes the stronger Strong Exponential Time Hypothesis and three related hypotheses.

OpenAI says its system found the disproof on October 16, with a Lean formalization produced the next day. The company released the preprint and formalization after an outside group asked for them to be made public. Thorsten Meyer’s review of the materials reports that the Lean repository contains 3,977 files and about 710,000 lines of code; the formal statement itself is much shorter.

The formalized result does not match the strongest bound in the paper. According to Meyer’s account, Lean checks a slightly weaker runtime bound, while the sharpest bound remains supported by the preprint rather than the formal proof. That difference makes the distinction between a machine-checked statement and the paper’s full claim important: Lean verifies the formalized theorem, not automatically every claim in the manuscript.

Why ETH Shapes Complexity Research

ETH is used in theoretical computer science to support claims that many algorithms cannot run substantially faster than known approaches. If the preprint’s central result survives review, researchers would need to revisit results whose hardness arguments rely on ETH or related assumptions. The extent of any revisions would depend on which specific conclusions use those assumptions.

The claimed algorithm is not a practical shortcut for ordinary computing. Its runtime is described as roughly 2 to the power of n divided by the square root of log n: subexponential, but still far slower than polynomial time. OpenAI says practical improvements would require significant further research, and the result does not show that real-world SAT instances can now be solved efficiently.

The development also tests a proposed route for evaluating AI-generated research: publish the work, encode its proof in a formal system, and have domain experts inspect the result. Formal checking can catch errors in a proof as represented in the system, but it does not replace expert assessment of the theorem’s meaning, assumptions, or relation to the paper’s claims.

Products Worth Considering

Philips Hue Bridge Pro Smart Lighting Hub

AmazonView Latest Price

Ring Battery Doorbell Plus, Satin Nickel

AmazonView Latest Price

Shark Rocket Pet Plus Corded Stick Vacuum, Bordeaux, HV322

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

From SAT Algorithms to ETH

In k-SAT, each rule is a clause involving at most k variables, and the task is to determine whether truth values can satisfy all clauses. With n variables, checking every assignment takes 2ⁿ combinations. Researchers have developed faster-than-brute-force algorithms, but ETH predicts that for 3-SAT the runtime must still grow exponentially with n.

Russell Impagliazzo, Ramamohan Paturi and Francis Zane formulated ETH in 2001. The hypothesis is stronger than the statement P ≠ NP: refuting ETH would not establish that every efficiently checkable problem is efficiently solvable. P versus NP remains open, and the Clay Mathematics Institute lists it as an unsolved problem.

OpenAI presented the ETH paper alongside two other research claims: a proof concerning conjectures in harmonic analysis, and a physics result involving quantum chromodynamics and a proposed completion of the Standard Model. The company says outside experts are guiding the handling of each project. The ETH materials were released; the physics materials are being held back at the request of the group overseeing them.

Review and Bound Still Open

The outside review has not concluded. A group led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams is expected to examine the result; Impagliazzo and Paturi were among ETH’s original authors. The available account does not provide a completed assessment from that group or a timetable for its findings.

It is also unresolved whether the paper’s sharpest runtime bound can be formalized and checked, or whether independent reviewers will accept the argument as written. The Lean check is meaningful evidence about the statement encoded in its files, but it should not be described as confirmation of every claim in the preprint.

OpenAI’s announcement and the source account describe the result as an ETH refutation, while expert scrutiny is still pending. The scope of any consequences for existing hardness results will depend on the final theorem, its assumptions, and how researchers assess those applications.

Independent Review Comes Next

The next milestone is the external group’s examination of the preprint and its formalization. OpenAI says the group will explain and verify the computer science work, and that the company has agreed to fund a conference on the result. No review date or conference date is specified in the source material.

Researchers will also be watching for clarifications or revisions to the paper and Lean code, particularly on the gap between the strongest bound claimed in the preprint and the weaker bound covered by the formalization. Until the review is complete, the claim remains a major announced result rather than a conclusion independently endorsed by the field.

Key Questions

What does OpenAI say its AI proved?

OpenAI says an internal model produced a proof of a deterministic, subexponential-time algorithm for exactly counting solutions to fixed-width k-SAT formulas. Applied to 3-SAT, the claim would refute ETH.

Has the proof been verified?

A Lean formalization has been machine-checked, but it covers a slightly weaker runtime bound than the sharpest claim in the preprint, according to the source account. An outside expert group has not yet completed its review.

Does this solve P versus NP?

No. Refuting ETH would challenge a stronger hypothesis about exponential-time algorithms, but it would not settle P versus NP, which remains open.

Will computers solve SAT problems faster now?

Not on the basis of this announcement alone. The claimed algorithm remains far slower than polynomial time, and OpenAI says practical improvements would require significant further research.

What happens next?

An external group led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams is expected to examine the paper and formalization. The source material gives no timetable for its findings.

Source: ThorstenMeyerAI.com