Is Physics About to Be Conquered by AI? Five open quantum problems that have remained unsolved for more than 20 years have been broken through.

That's terrifying, it's only been a few days?

How come AI just messed up math, and now it's messing up physics right after......

Just now, 5 tough quantum information problems were announced to be solved with AI assistance.

The company behind this is a mysterious firm called Physical Superintelligence (PSI for short).

They stated that with the help of their self-developed AI scientific research system, they have made achievements on 5 open problems in quantum information theory. All papers, Lean formal proofs and verification codes have been posted on Github for academic review.

Who on earth is this PSI, which was established less than two months ago?

What exactly has PSI solved?

The quantum information science that PSI focuses on this time mainly studies how to use the special laws of quantum mechanics to store, process and transmit information, so as to break through the computing power limit of traditional computers and explore more secure communication methods.

The five tough problems they challenged with AI this time are all from the 27-year-old Open Quantum Problems (OQP) catalog.

This open problem catalog was established by German physicist Reinhard Werner in 1999, specifically collecting important long-standing research problems stuck in the quantum information academic circle.

The 5 problems this time cover five core directions: entanglement distillation, quantum field theory, high-dimensional quantum communication, quantum channel capacity and quantum causality, including:

OQP 2｜Undistillability of Entanglement in Negative Partial Transpose States

OQP 12｜Vacuum Entanglement and Bell's Inequalities

OQP 27｜Optimal Measurements for the CGLMP Bell Inequality

OQP 38｜The PPT Square Conjecture and Entanglement Breaking

OQP 43｜Purifiability of Quantum Causal Processes

The specific proof process will not be expanded here. It is enough to know that 3 of the 5 problems that PSI has made progress on this time (OQP 2, OQP 12, OQP 27) were all included in the early 21st century, which speaks volumes about their difficulty.

"The World's Most Advanced Physics Research Laboratory"

Clicking into PSI's website, it can be described as completely empty. This breakthrough in solving quantum information problems is also their first achievement.

After all, this company was officially unveiled on September 1 this year.

But it never shows up without a blockbuster, they made a grand ambition at the beginning of their debut, claiming that they want to build "the world's most advanced physics research laboratory".

PSI's founding team is quite interesting, and Alex Wissner-Gross, founder and chief scientist, is the most remarkable one.

This person once completed three undergraduate majors at MIT at the same time: Physics, Electrical Engineering and Mathematics, and graduated with the first place in the School of Engineering......

What does it mean to complete these three majors at the same time? To put it this way, MIT canceled the option of "studying three majors at the same time" after him, on the grounds that "too few people can do this, so there is little point in offering this option".

Therefore, Wissner-Gross even has a title of "the last undergraduate in MIT history to graduate with three majors".

Perhaps after testing the waters, this top talent decided that he still loves physics more, so in 2007, he obtained his PhD in Physics at Harvard.

It has to be said that his understanding of "AI + Physics" is far ahead of the era.

As early as 2013, when deep learning was still in its infancy, he stood on the TEDx stage, using the concept of "causal entropic force" in physics to explain intelligent behavior.

He put forward a very bold idea.

The so-called intelligence can be understood as a tendency to continuously increase the possibility of future actions.

This may not be easy to understand, let me give you an example.

For example, when you squeeze into the subway, why do you try your best to keep your balance and avoid falling as much as possible?

From his theory, this is because once you fall, the number of actions you can choose will decrease. Standing still means you can still move forward, backward, left and right in the future.

Then, as long as the rule of "maximizing future possibilities" is followed, the system may spontaneously show intelligent behaviors similar to tool use and cooperation.

Although it sounds reasonable, this idea has not been proved so far.

Such a genius as Wissner-Gross could no longer hold back this year, and proposed to "Solve Everything".

In two articles he published in February this year, he put forward the current PSI concept.

There are quite a few inspirations that echo the present: "The Singularity needs a physics division."

"The era of proof has arrived."

Apart from this top talent, Alex Klokus, co-founder and president of PSI, is also very interesting. He is the founder of Futurism, a well-known American technology media.

The financing background can show their strength at a glance. When they debuted in September, PSI announced that it had completed a $58 million seed round of financing.

The lead investor is Breakthrough Energy Ventures founded by Bill Gates.

This fund gathers top figures from the technology and business circles including Amazon founder Jeff Bezos, Alibaba founder Jack Ma, and Virgin Group founder Richard Branson, and specifically invests in cutting-edge technologies that have the potential to drive energy transition.

The follow-on investment lineup also includes SV Angel, a long-established Silicon Valley venture capital firm. Its founder Ron Conway is a famous angel investor who bet on technology companies such as Google, PayPal and Twitter in the early years.

Not only that, Balaji Srinivasan, former partner of a16z, also participated in the investment, and it also attracted individual investors from technology companies such as OpenAI, NVIDIA, Hugging Face and Oracle.

Realize "Mass Production" of Scientific Discoveries

After talking so much, to sum up PSI's vision in one sentence, it is actually to make scientific discoveries "From Genius to Scale".

For hundreds of years, the leaps of human basic science have almost relied on individual geniuses like Newton and Einstein who had sudden flashes of inspiration.

But PSI believes that this purely manual workshop-style scientific research has hit the ceiling, otherwise why are we in the 21st century still standing on the physics foundation laid in the 20th century?

Scientific discoveries should no longer be limited by human time.

PSI calls the basic research unit Autonomous Campaign, hoping to make research run continuously from problem definition, hypothesis formulation, verification, experiment to engineering deployment.

PSI also believes that in the AI era of scientific research, although ideas are no longer scarce, the truth is still scarce.

Large language models can quickly generate hundreds of conjectures, but these are far from becoming truths.

To solve this problem, PSI's self-developed AI scientific research system Emmy (named in tribute to Emmy Noether, the great 20th century mathematician and founder of theoretical physics) separates the two steps of "generation" and "verification".

Its workflow is roughly as follows.

First, when facing a huge open problem in physics, Emmy calls cutting-edge large models to break it down into countless tiny, verifiable mathematical sub-hypotheses.

Second, the hypotheses are directly put into the Lean formal prover, symbolic computing engine and high-precision verification code for strict screening. In this way, wrong hypotheses are eliminated on the spot, and paths that pass local verification are retained and continue to derive new ones.

As long as the final Lean code is compiled successfully, it means that this physical theorem is 100% mathematically and logically impeccable.

The verification results in the process can also become reward signals for training the next generation of AI physicists, promoting the system to continuously improve, so as to realize the scaling of scientific discoveries.

One More Thing

It sounds great, but here comes the question: how can a company that studies physical theories make profits?

I went through the projects they are working on, and they are very pragmatic.

They plan to start making money from the most practical problems first.

Where? AI data centers.

The current supercomputing cluster is essentially an extremely complex macroscopic physical system, with highly coupled engineering optimization problems between power supply, cooling, network and computing resources.

This is exactly their area of expertise. The Emmy system can start from the underlying physical laws, take power supply, heat dissipation, communication and computing power as an overall equation to solve, and find higher energy efficiency ratio at the physical limit.

More interestingly, beyond business, PSI also took on a private job from an individual investor (probably a sci-fi fan).

The Fermi Explorer interstellar mission.

PSI announced: Start building spacecraft right now!

According to the official announcement, they plan to launch the deep space interstellar probe "Fermi Explorer" before 2029.

Using the optimal lightweight engineering scheme optimized by the Emmy system within the existing physical limits, this spacecraft will fly out of the solar system, and take 80,000 years to fly to Alpha Centauri, which is 4.2 light-years away from us, the home of Trisolarans and Avatars.

Although probably no human will live to see that day, wish them success.

Reference links:

[1] https://psi.inc/

[2] https://github.com/physical-superintelligence/oqp-batch-2026

This article is from WeChat official account "QbitAI", author: Focus on cutting-edge technology, authorized for release by 36Kr.