On the same day, the two giants announced that the Century Guess ETH was faked.

  It's the first time in the history of the world.

The world of theoretical computing is also being bombed by AI!

Last night,AI has completely crossed the Rubicon River of scientific research and has officially begun to "overturn" the knowledge base of humanity.

OpenAI is a breath of fresh air in the three fields of theoretical computing, pure mathematics and theoretical physics:

This is probably the most explosive tech news of the year, not even one.

Claude is even more straightforward that if these results are true, it will be the most important day in the history of mathematics and theoretical computing.

Among them, the mathematical breakthrough is highly related to earlier work on the Hanging Valley hypothesis by Felts Prize winner King Neon, whereas the physical breakthrough may be the most important of the three, or solves the dilemma left by Nobel laureate Xi Jinping.

This is a breakthrough in the field of computers: the time-index assumption is not valid.

According to Claude, this is a Turing Award-level disruptive achievement, "If correctly stated, it will change computer science".

Sebastian Bubeck, now a researcher at OpenAI, former vice president of artificial intelligence at Microsoft and a distinguished scientist, praised the "greatest surprise I've seen in 20 years".

And, on the same day, two AI's announced this breakthrough almost simultaneously!

OpenAI and Anthropic, the two biggest AI giants, announced almost simultaneously that their internal models had overturned the "index time hypothesis" (ETH) that had ruled theoretical computer science for 27 years.

What's even more alarming is that in a paper by Anthropic, which explicitly labels the core algorithm, the internal model Claude was working on.Zero human interventionIn the case of independent derivative discoveries.

Claude is the only author to sign the paper, which could be included in the History of the Humanities.

In simple terms, in 1999,For example, the index time assumption (ETH)It was proposed by Russell Impagliazzo and Ramamohan Paturi of the University of California, San Diego.

You can imagine it as a password lock with n switches, which is the 3SAT problem.

For half a century, countless brilliant minds have tried their best to unlock it, but no one has been able to break one thing:The time for breaking the lock has always swelled to "index" as the number of switches increases.

For example, suppose a lock needs to fill in n numbers, each number between 1 and 9. There are 9 options for each location, so there are a total of 9 n possibilities, i.e. 9 times n.

We can solve this problem by adding up all the possible 9n numbers. But this method is extremely ineffective: for a single number of n = 100 digits, that is, using all the computers in the world, it is necessary to wait for the sun to cool down.

For a long time, people have tried to find algorithms that run time with n growing slower than 9n, but with no success.

In 1999, scholars wrote this collective intuition as a formal hypothesis:3SAT does not have a sub-index time algorithm.

You can improve your speed a little bit, but you can't always overturn this "index wall". It's like "the speed of light cannot be exceeded" in physics.

This assumption has been made for more than two decades, and hundreds of papers have been written on the premise of ETH to explore the complexity of various computational problems.

From graph theory, combinatorial optimization, to database and computational biology, many studies are based on this hypothesis.

Lyman speculates that if overturned, it's mathematics; ETH is overturned, it's complexity and parametric algorithms.

ETH's share of the world's population has dropped.P vs NP"or Lyman's guess of the millennial level".

But it is stronger and more specific than P≠NP: if ETH is formed, it can produce P≠NP; but ETH is overturned and cannot produce P=NP.

The findings of conditional studies that rely heavily on ETH need to be reassessed.

A new wave of "victims" in AI has emerged.

There is also a younger brother in the group, who wrote a paper on the premise that ETH was founded.

Now that the premise may be clear, the conclusions of the paper should be reviewed.

Imagine that you spent several years researching a problem, writing a paper, proving it, and even an academic in a field studying this hypothesis for more than two decades.

As a result, one day an AI model will tell you that the premise you've always believed might be wrong.

A dramatic scene soon followed.

Former OpenAI president Greg Brockman has just posted that his former opponent Anthropic is directly behind a paper on the same topic.It's hard on the face, it's a crash!

It's not a coincidence that Claude's internal model is changing.16th of OctoberIt's been independently overthrown by ETH!

Moreover, Anthropic left a "source statement" on the paper page:

The core algorithm was found by an internal Claude study model on October 16, consuming 14 million tokens without any human input.

This means:AI has completely crossed the Rubicon River for scientific researchIt is no longer a human tool, but a true "first author".

What is the concept of 14 million tokens? A three-dimensional set of tokens is about 900,000 words. 14 million tokens, the equivalent of Claude's crazy breath, is a draft of dozens of three-dimensional tokens.

A number of leading scholars have estimated that this is the equivalent of an AI model completed in a matter of days without eating or drinking, without sleeping or resting.One to two years of post-doctoral research in human topics！

In those days, Claude himself explored the path, devised his own formula, verified his own logic, and even cross-referenced cold-blooded mathematical tools that could not be tamed, finally putting a proof in front of humans.

This time, it's not "AI helps mathematicians look at literature" or "AI does an Olympic puzzle".

It's AI in the hardest, purest, theoretical discipline of humanity, where a single-gun horse breaks the sky.

Is AI saying that it's overturned? Is AI saying that it's overturned?

The most magical scene happened. Both companies, in an attempt to prove themselves, did not just throw out a PDF, but made it public at the same time.Lean formalized proof of code。

Lean was a "selfless electronic reviewer" who, even with a logical leap, made direct mistakes. Since the code could run, the logic underlying almost sealed off the possibility of error. Delivering a "machine verification" curve was the undeniable essence of AI.

However, this does not fully confirm the correctness of the proof.

Russell Impagliazzo and Ramamohan Paturi, the two founders who proposed the ETH speculation that year, were reportedly asked to leave the mountain in an emergency.

Twenty-seven years ago, two scholars wrote a speculation; twenty-seven years later, they were awakened to examine how an AI could overturn it.

This is probably one of the most complex moods in academic history.

Currently, the top teams at universities such as MIT are conducting step-by-step testing of both codes.

At the same time, another front is advancing rapidly.

In the face of AI, this 60-page "extraterrestrials" show that the human brain can read extremely hard.

Shiva Kintali, a senior AI architect at INVEDA, has proven the authenticity of the more than 60 pages of the original OpenAI ETH.Hardcore Life is compressed into an abridged version of 22 pages!

He cut off the AI-heavy steps, didn't pursue the optimum pair factor, only sought a clean profit margin to falsify basic statements, and supplemented arithmetic cases.

He has only one goal:"Let your fellow human beings really understand".

Now, an unprecedented academic miracle has emerged:

AI is responsible for the attack: knocking on a door that hasn't been opened in 27 years in a matter of days;

Humans are responsible for digesting: to drive AI's extensive evolution, refining it into knowledge that humans can understand and keep pushing forward.

Now, it's time for humans to simplify the task of AI.

As early as March 2023, OpenAI released a groundbreaking report predicting AI's impact on the job market.He is a mathematician who has been ranked as one of the "biggest AI-impacted professions". 。

At the time, most people snapped at the idea that mathematics required the most rigorous human logic, and how could an AI with a full-mouthed train do theoretical mathematics?

However, three years later today, with ETH being falsified by AI overnight, Kakeya speculates that the connection has been broken by the model...

The bullet, fired three years ago, has finally caught the eye of human theoretical research.

Does this mean that scientists are collectively unemployed?

As Weijie Su, a North Korean alumnus and scholar, saw three years ago:AI will not simply "replace" any field roughly, but will reshape it. Future research will greatly reward those who can make deep connections with real-world challenges and other disciplines.

As with the invention of the camera, painters no longer had to write "who painted better", and so Impressionism and modern art were born.

When AI takes over the hard work of "reckoning with evidence", humans can finally devote their precious energy to more fundamental issues:Looking for the next question worth exploring and building real-world connections across disciplines.

Twenty-seven years ago, when the two scholars wrote the ETH conjecture, it was absolutely unthinkable that it would not end with humans.

I can't even imagine that it would be two AIs on the same day.

"The world is not being changed by AI, it has already been changed".