OpenAI’s AI Results: A Challenge To The Exponential Time Hypothesis?

🔍 Read the full analysis: OpenAI’s AI Results: A Challenge To The Exponential Time 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 produced a proof refuting the Exponential Time Hypothesis (ETH), a central conjecture about the limits of solving hard logic problems. The company has released a preprint and a Lean formalization, but the result has not yet been independently reviewed, and the formalized statement covers a weaker runtime bound than the paper’s sharpest claim.

The preprint claims a deterministic algorithm can exactly count satisfying assignments for any fixed-width k-SAT formula in subexponential time and polynomial memory. In 3-SAT, where each rule can involve up to three variables, that would contradict ETH, which predicts that no algorithm can solve every instance in time below an exponential function of the number of variables. The paper says the result also refutes the stronger Strong Exponential Time Hypothesis and three related hypotheses.

OpenAI says its system found the disproof on October 16, and that the Lean formalization followed the next day. The preprint is dated October 17 and was revised October 25. Lean is a proof assistant that checks whether a formal proof follows from its stated definitions and rules. The source article reports that the public formalization contains 3,977 files and about 710,000 lines; the short formal statement can be read separately. Those counts are the source author’s tally, not a figure independently confirmed here.

The distinction between the paper and the formalization matters: according to the source material, Lean checks a slightly weaker runtime bound, not the sharpest bound claimed in the preprint. OpenAI has made the preprint and code public at the request of an external group that will explain and examine the result. The source reports that the review group is led by Russell Impagliazzo, Ramamohan Paturi and Ryan Williams; Impagliazzo and Paturi helped formulate ETH.

Why an ETH Refutation Would Matter

If the paper’s central claim holds up, it would change a widely used assumption in theoretical computer science. ETH underpins conditional lower-bound arguments: researchers use it to show that certain algorithms are unlikely to run substantially faster than known approaches. A refutation could force those results to be revisited, although it would not automatically provide practical algorithms for everyday problems.

The claimed algorithm remains superpolynomial—its running time is still far slower than any polynomial-time method. The source material describes its time roughly as 2 to the power of n divided by the square root of log n, and says any practical improvement would require substantial further research. The development is therefore as much about proof production and verification as immediate computing performance: an AI-generated argument is being paired with a machine-checkable formalization and an expert review process.

For readers outside theoretical computer science, the result does not solve P vs. NP. That question asks whether every problem whose proposed solutions can be checked efficiently can also be solved efficiently. ETH is a stronger, separate conjecture about the time needed for particular hard problems. Its possible refutation would be a major change to complexity theory, not evidence that all hard problems can now be solved quickly.

Products Worth Considering

Amazon Kindle Paperwhite (12th Generation, 2024), 16 GB, 7-inch Display, Jade

AmazonView Latest Price

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

AmazonView Latest Price

Sony WH-1000XM4C Wireless Noise-Cancelling Headphones, Black

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

The Assumption Behind the Claim

In k-SAT, a formula consists of logical rules involving at most k variables, and the task is to find an assignment of true or false values that satisfies all rules. With n variables, trying every assignment takes 2^n possibilities. Researchers have developed faster methods by reducing the exponential base, but ETH says that for 3-SAT the exponential dependence itself cannot be eliminated: the running time must remain at least 2 raised to a positive constant times n, for some fixed constant.

ETH was introduced in work by Impagliazzo, Paturi and Francis Zane in 2001. Its role in the field is substantial, but the degree of agreement about its truth should not be overstated. OpenAI describes it as widely believed and says it has supported hardness results in hundreds of papers. The source material also cites a 2017 survey by Scott Aaronson saying there was no consensus that ETH is true. That qualification does not settle the mathematical claim; it describes the limits of claims about researchers’ views.

The ETH announcement was one of three results OpenAI described in its October 26 post about AI-generated research. The company also announced mathematics and physics work. The source material says OpenAI presented the manuscripts as raw research data and worked with outside experts on how to release each one. It also records earlier repository corrections: three manuscripts were withdrawn after a sign error and 14 were revised. That history makes independent checking relevant, but it does not by itself establish whether the ETH proof is right or wrong.

What Review Must Establish

The central open question is whether specialists can verify the paper’s reasoning and its claimed consequences. Independent review is pending, and the Lean check, as described in the source material, does not formalize the paper’s sharpest runtime bound. It remains to be explained whether the formalized weaker statement is enough to establish the paper’s headline refutation of ETH, and how the remaining gap will be addressed.

The source material gives no completed assessment from the review group and no final verdict from the mathematical community. It also does not establish whether the proposed algorithm can be implemented in a way that improves real-world performance. Until review is complete, the result should be described as OpenAI’s claim, supported by a public preprint and a machine-checked formalization, rather than an independently confirmed refutation.

Independent Review and Formal Checks

The next milestone is the external group’s examination of the preprint and formalization. The source material says the group, led by Impagliazzo, Paturi and Williams, will explain and verify the computer-science result. Their assessment should clarify whether the proof establishes the stated ETH consequence, what the formal Lean result covers, and whether further formal work is needed to match the paper’s strongest bound.

OpenAI has also agreed to fund a conference on the result, according to the source material. No date or schedule for that conference is provided. Until the reviewers publish their findings, readers should treat the work as an important but unsettled research claim; any consequences for algorithms or applications would depend on the proof surviving scrutiny and on further research.

Key Questions

What is the Exponential Time Hypothesis?

ETH is the conjecture that 3-SAT cannot be solved in subexponential time as the number of variables grows. It is used to support conditional claims about the difficulty of other problems.

Does this result solve P vs. NP?

No. P vs. NP remains open. The claimed result concerns ETH, a stronger and distinct hypothesis about the running time of particular problems.

Has the proof been independently confirmed?

Not yet, according to the source material. OpenAI has released a preprint and Lean formalization, and an external group is expected to examine the result. A machine check is not the same as independent expert review of the claim and its scope.

What does Lean verify in this case?

Lean checks a formalized proof against its definitions and rules. The source material says the public formalization establishes a slightly weaker runtime bound than the sharpest bound in the preprint, leaving the relationship between those claims to be clarified.

Will this make current computing faster?

There is no evidence of an immediate practical speed-up. The claimed algorithm is still superpolynomial, and OpenAI says practical improvements would need significant further research.

Source: ThorstenMeyerAI.com