AI News

OpenAI model proves that Landau-Siegel zeros do not exist; the academic community discusses its distance from P≠NP

Phoenix TechnologyUpdated Nov 9 11:46

AI summary

An internal OpenAI model produced a proof that "Landau-Siegel zeros do not exist"; the original proof was only 8 pages, and mathematicians subsequently compressed it to 5 pages. On October 6, OpenAI released 722 manuscripts on GitHub (withdrawing 3 the next day), classified into 372 result families; the internal model attempted about 4,000 open problems, averaging about 3 hours of ChatGPT Pro compute per problem. Stanford number theorist Jared Duker Lichtman said that the existence of such a proof strongly suggests P≠NP, but Lance Fortnow pointed out that the AI's solve rate is less than 10%, and that not a single result brings people closer to resolving P vs NP; only about 42% of the manuscripts come with Lean formal proofs.

Why read this

AI score 72

The original text reviews multiple manuscripts from OpenAI models on number theory and complexity theory, and lists the academic community's doubts about the verification progress and the gaps in the core approach.

Topics

Tags

Body · AI translation

OpenAI Cracks a Century-Old Number Theory Mystery, Revealing a Major Clue to P≠NP! The Millennium Prize Problem Inches Closer to the New Intelligence Era November 09, 2026 09:57:11 From Beijing Updated November 09, 2026 10:16:52

A number theory mystery that humanity has been stuck on for decades—did AI solve it in just 8 pages?

Over the past two days, the entire mathematics community and the computer science world have been in enormous shock.

On October 18, Stanford University number theorist Jared Duker Lichtman's words on X instantly set the whole internet ablaze—

"The existence of such a proof strongly suggests P≠NP."

The trigger for this statement was the astonishing answer sheet just delivered by an internal OpenAI model—a proof that "no Landau-Siegel zero exists."

This key problem in analytic number theory has been stuck for decades, and the academic community universally regards it as extremely difficult.

As a result, not only did OpenAI solve it, but the original proof was actually only 8 pages long!

What's even more exciting came afterward. Top mathematicians quickly followed up:

Bill Banks of the University of Missouri first wrote a detailed 19-page breakdown;

Then, Bryce Kerr used biquadratic extensions

and Masser's classical results to compress the proof to 6 pages;

Subsequently, University of Oxford professor Ben Green cut it down again, directly giving a 5-page ultimate version!

From decades of hopelessness with no solution to just 5 pages. Abhishek Saha, a professor at Queen Mary University of London, posed a soul-searching question:

If this problem can be proved in just 5 pages, why did mathematicians miss it for decades?

Finding it is like climbing to heaven, verifying it is like drinking water

Why would a 5-page number theory proof make a Stanford professor directly think of the holy grail of computer science—P vs NP?

The reason is simple.

A 5-page proof can be read and verified by any expert in a few hours; but for these 5 pages, humanity's brightest minds struggled and groped for decades.

"Finding is extremely hard, verifying is extremely easy." These eight words are precisely the underlying intuition of P≠NP!

In the comments section, Anthropic researcher Joshua Batson pointed out incisively—

Behind the joke is an extremely serious question: when AI delivers more than 700 mathematical manuscripts within a few weeks, is the millennium prize problem P vs NP also coming soon?

Three High Walls Lock Humanity Down for Fifty Years

Explain P vs NP in one sentence: Are all problems whose answers can be quickly verified also problems whose answers can be quickly computed?

P is the set of problems that "can be solved quickly," and NP is the set of problems that "can be verified quickly."

Sudoku, class scheduling, and codebreaking all belong to NP: give you the answer, and you can check whether it is right or wrong in a second; but to find it from scratch, you can only do something close to brute force.

If P=NP, modern cryptography would instantly collapse, global finance would be left exposed, but all complex problems would gain an ultimate shortcut;

If P≠NP, it means the universe contains problems that are essentially impossible to shortcut.

This problem was posed in 1971, with a $1 million prize. The vast majority of scientists believe P≠NP, but for more than fifty years no one has been able to prove it.

Where is the difficulty? Because humanity has not only failed to find the key, but has instead proved that existing lock-picking tools are doomed to be ineffective.

These are the famous three "barriers":

1. The relativization barrier (1975): classical methods relying on simulation and diagonalization simply cannot solve it;

2. The natural proofs barrier (1994): most proof approaches for circuit lower bounds, if they could prove P≠NP, would conveniently break pseudorandom number generators, leading to self-contradiction;

3. The algebrization barrier (2008): upgraded methods that also bring in algebraic techniques have likewise been proved to lack sufficient firepower.

In June this year, complexity theorist Lance Fortnow pessimistically stated: at present, humanity does not even have one feasible route of attack.

He suggested that challengers first prove a much simpler problem such as "SAT has no quadratic-time algorithm," then added a finishing blow: "You won't succeed there either."

On October 6, OpenAI released 722 manuscripts at once on GitHub (withdrawing 3 the next day), classified into 372 result families.

The internal model attempted about 4,000 open problems, averaging only about 3 hours of ChatGPT Pro compute per problem.

Among them, theoretical computer science accounted for 40 families. Although it did not directly kill P vs NP, several papers landed precisely on the main road leading to it.

In June this year, complexity theorist Lance Fortnow wrote: apart from simulation, we have no way to handle general polynomial-time algorithms; routes such as circuit complexity, proof complexity, and algebraic geometry are all stuck far below the full power of polynomial time. His conclusion was that at present there is not even one feasible route of attack. Coincidentally, OpenAI's six results this time planted a flag on each of these routes.

Family No. 112 (depth-3 circuit lower bounds): the most favored route for proving P vs NP is to prove that some NP problem requires "superpolynomial-size circuits." Humanity has long been stuck on explicit lower bounds for depth-3 circuits at

Family No. 102 (Unique Games Conjecture, UGC): proved this core conjecture proposed by Subhash Khot in 2002 (including partial Lean verification). It sits closely adjacent to P vs NP and greatly strengthens the inapproximability limits of many NP-hard optimization problems.

Basic algorithms and derandomization: achieved full derandomization of randomized logarithmic space (L=RL=BPL); compressed the matrix multiplication exponent to

; integer multiplication and DFT fell below

time.

Around Family No. 102, the "Unique Games Conjecture," there was also a heart-wrenching episode.

The wife of the famous scholar Scott Aaronson, Dana Moshkovitz, devoted her entire career to this conjecture. On the night the results were released, their 9-year-old son ran over to mock his mother:

"I heard there's a robot that solved the problem you've been researching your whole life!"

Quanta Magazine also disclosed that MIT's Dor Minzer team, after hearing rumors about OpenAI on September 11, rushed to release their own proof of a related variant before OpenAI's release so as not to be scooped by a machine.

Lance Fortnow published an article the next day pointing out that AI's solve rate on given problems was less than 10%, and that none of these results brought us any closer to resolving P vs NP.

Although Scott Aaronson called October 6 "one of the most important days in the history of mathematics," he also pointed out that the list contained not only no P≠NP, but not even "smaller" hard problems like P=BPP—clearly not because no one had tried.

He revealed on a podcast that OpenAI and Anthropic are competing to crack P vs NP using their strongest internal models, but neither has succeeded yet.

Looking at it calmly, there are still four major obstacles before a breakthrough:

1. Verification is far from complete: only about 42% come with Lean formalized proofs; the rest rely entirely on humans. Aaronson admitted that almost no human has truly understood any of the long proofs.

2. Terrible readability: Moshkovitz commented that the AI's Unique Games proof was written so "badly that you can't understand it at all without the help of AI."

3. Lean is not omnipotent: machines can only guarantee that the reasoning is correct; they cannot guarantee that the formalized definitions are completely consistent with the original problem, nor can they judge whether the result is important.

4. The core route remains blank: what AI broke through were depth-3 circuits and algebraic circuits, while the "general Boolean circuit lower bounds" that could imply P≠NP remain unwritten.

First, the "hardness" of hard problems has been repriced. The 5-page proof of the Siegel zero shows that what some dead ends lack is the one step no one thought of. How many conjectures are still just one layer of window paper away? No one is sure.

Third, scale and speed are not on a human scale. 4,000 problems, averaging 3 hours each. When major labs aim their next-generation strongest compute squarely at P vs NP, quantitative accumulation is highly likely to trigger qualitative change.

As for the final outcome, an interesting divergence in bets has even emerged in academia.

Aaronson still bets heavily on P≠NP; but some scholars believe that if AI really can break through, the answer is more likely to be P=NP—because getting AI to search out a clever algorithm is much easier to imagine than using existing mathematics to prove "impossible."

However, the most intriguing humor of the whole affair lies in this—

The more powerful AI becomes, the more it itself looks like living evidence of P≠NP.

It expends massive compute to "find" proofs, while humans spend a few hours to "verify" the results.

If one day AI really burns through its compute and proves P≠NP, that will be the most perfect self-proof of this millennium problem.

Source

Phoenix Technology