SIAgentNews.com
← All news

Claude solves two long-standing theoretical physics problems

Anthropic's AI model completed complex calculations regarding black hole formation and quantum gravity limits that had stumped physicists for decades.

· First reported by anthropic.com

61parallel attempts Claude failed before finding the correct method

The short version

  • Claude completed a 20-year-old calculation showing that black hole formation is an abrupt phase transition.
  • The AI proved the exact limit for a technique used to rule out inconsistent theories of quantum gravity.
  • The results were verified using Lean, a computerized proof checker, to ensure mathematical accuracy.
  • The AI achieved these results by running a small research group of independent copies of itself.

Why it matters

While these findings are theoretical and don't have immediate practical use, they help bridge the gap between general relativity and quantum mechanics. They provide evidence that certain properties of gravity are built into the underlying quantum laws of the universe.

Bridging two worlds of physics

Modern physics currently relies on two separate frameworks that do not work well together. General relativity explains the behavior of massive objects like stars and the curvature of spacetime, while quantum mechanics governs the behavior of subatomic particles like atoms.

The ultimate goal for many physicists is to find a single theory of quantum gravity that unites these two systems. Black holes serve as a primary testing ground for this goal because they contain immense mass within a tiny space, forcing both relativity and quantum mechanics to matter simultaneously.

To study this, researchers often use holographic duality. This principle suggests that a universe with gravity can be described as a quantum theory without gravity, similar to how a 3D hologram is encoded on a 2D surface. This allows scientists to translate difficult gravity problems into potentially simpler quantum physics problems.

The black hole transition

One problem Claude solved involves what happens when a theoretical universe is heated up. At low temperatures, the universe contains a gas of particles, but above a certain temperature, a black hole forms instead. In 1983, physicists showed this switch is an abrupt phase transition, much like water boiling into steam.

However, that original calculation only worked when the universe was large enough for general relativity to be accurate. When the universe shrinks to a tiny scale, spacetime becomes blurry and loses its definite shape, making the gravity side of the math impossible to solve. Scientists knew a transition existed on the quantum side but could not determine if it was abrupt or smooth.

Claude completed a calculation set up in 2003 that settled this question. By running a small group of independent agents to check each piece of math, the AI showed the transition remains abrupt. This suggests that black hole formation is a fundamental feature of the underlying quantum theory, not just a side effect of relativity.

Ruling out wrong theories

The second problem involves a simplified universe with only two dimensions of space. Because physicists cannot yet build a perfect theory of quantum gravity, they often use a method called bootstrapping to rule out incorrect theories by showing they are internally inconsistent.

Since 2009, researchers have used this method to set a maximum weight limit for the lightest black hole in such a universe. While numerical calculations suggested where the best possible limit lay, they could not provide a formal proof. The best proven ceiling remained significantly lower than what the numbers suggested.

Claude noticed a pattern in these numerical results that humans had missed. It borrowed a method from a different area of mathematics to calculate the precise ceiling for the simplest version of this technique. The AI proved that this ceiling could not be lowered, pointing researchers toward the correct direction for future theories.

How the AI performed the work

The discovery happened during a side project where an Anthropic researcher asked Claude to identify open problems in theoretical physics. Claude identified 22 candidates, and the researcher set up separate agents to work on them independently. These agents were able to read each other's progress and collaborate on related tasks for about six weeks.

One agent identified the pattern in the bootstrap literature, while another agent working on sphere-packing in high-dimensional space provided the necessary mathematical method. Claude then applied this method to the bootstrap problem, which it had previously failed to solve in 61 parallel attempts. Other agents independently arrived at the same answer as well.

To ensure the results were accurate, the findings were formally verified in Lean, which is a computerized proof checker. After the verification, Claude wrote the research paper with minimal guidance. A human physicist then reviewed the paper to confirm every main result was correct.

The path forward for physics

Neither of these results describes our actual universe directly, and they do not have immediate practical applications for technology. Instead, they represent significant progress in the long-running effort to understand the fundamental architecture of reality. They provide a clearer picture of how gravity and quantum mechanics might eventually be unified.

The success of these experiments highlights the potential for AI to handle the extreme complexity of theoretical physics. By breaking down massive calculations into modular pieces and checking them against known results, AI can navigate mathematical landscapes that are too dense for human calculation alone.

As these models become more capable, they may help physicists identify which theories of quantum gravity are most viable. By ruling out inconsistencies and completing unfinished calculations, AI acts as a tool to narrow down the search for the ultimate laws of physics.

The significance of the results

The first result regarding the Hawking-Page transition is significant because it shows that the transition is consistent across different descriptions of the universe. It confirms that the sudden appearance of a black hole is a robust feature that persists even when spacetime becomes blurry and loses its traditional shape.

The second result regarding the bootstrap method provides a definitive mathematical ceiling. While it does not identify the final correct theory of quantum gravity, it provides a rigorous proof that narrows the field of possibilities. It confirms that certain limits are mathematically fixed, preventing researchers from pursuing impossible paths.

Together, these problems show that Claude can handle high-level reasoning and complex mathematical verification. The fact that the AI reached the same conclusion as a human physicist who was using Claude independently serves as a useful cross-check on the reliability of these automated solutions.

Technical context of the duality

To understand the first problem, it helps to visualize the anti-de Sitter spacetime as a giant bowl. In this model, gravity pulls everything back toward the center, allowing the bowl to stay at a steady temperature when heated. This environment allows physicists to see two different versions of the universe competing for dominance.

One version is a gas of particles, and the other is a black hole. At the Hawking-Page temperature, the universe switches from one to the other. The holographic principle suggests that the physics inside this bowl is equivalent to a quantum theory existing only on the rim of the bowl.

Claude studied the quantum theory on the rim, known as N=4 super Yang-Mills theory. This theory is related to how particles interact in the real world but includes extra symmetries to make it easier to study. By solving the transition problem, Claude proved that the quantum theory on the rim behaves exactly like the gravity theory inside the bowl.

Verification and collaboration

The process used to solve these problems demonstrates a new way of using AI for scientific discovery. Rather than just asking for an answer, researchers set up a collaborative environment where multiple agents could share progress and cross-reference data. This mimics the way human research teams operate to ensure accuracy and find hidden patterns.

One of the agents was able to spot a pattern in numerical results that had been overlooked by the physics community. It then combined this with a method from a different mathematical field to produce a proof. This ability to synthesize information from unrelated areas of study is a key strength of the model.

The final step of using Lean to verify the proofs is crucial for scientific integrity. Because Lean is a computerized proof checker, it ensures that the logic holds up without human error. This combination of AI-driven discovery and automated verification provides a high level of confidence in the results produced by Claude.

Implications for the future

These results suggest that AI can be a powerful partner in exploring the most difficult questions in theoretical physics. By completing calculations that were abandoned decades ago, Claude is helping to clear the path for human scientists to focus on new questions. It serves as a proof of concept for using AI to tackle problems that require both deep reasoning and massive computational power.

As the field of quantum gravity continues to evolve, the use of AI to rule out inconsistent theories will likely become more common. The bootstrap method, in particular, could become a standard way to prune the tree of possible physical laws. This accelerates the pace of discovery by focusing human attention on the most promising candidates.

While we are still far from a complete theory of everything, these small steps are vital. They provide the mathematical scaffolding needed to understand how the universe works at its most fundamental level. Claude's ability to navigate these complexities marks a new chapter in the intersection of artificial intelligence and fundamental science.

Summary of the findings

In summary, Claude has successfully navigated two distinct hurdles in the quest for quantum gravity. It confirmed that the transition from a gas to a black hole is an abrupt event that holds true even in the most extreme, blurry regions of spacetime. It also established a rigorous mathematical ceiling for the weight of black holes in simplified universes.

These achievements were made possible by a multi-agent system that could collaborate, share information, and verify its own work. The AI was able to identify patterns in existing data and apply methods from unrelated fields to solve problems that had remained open for years.

The results were formally verified using a computerized proof checker and reviewed by human experts. This ensures that the findings are mathematically sound and can be used by the physics community to further the search for a unified theory of the universe.

The role of holographic duality

Holographic duality is the core concept that allowed these problems to be solved. It provides a bridge between two different ways of looking at the universe: one involving gravity and one involving quantum mechanics. By using this duality, physicists can translate a problem that is impossible to solve in one

What to watch

  • Whether other AI models can achieve the same level of verification in Lean.
  • How physicists will use these new ceilings to narrow down candidate theories of quantum gravity.

Try it

Look up the 'Hawking-Page transition' to see how physicists describe the point where a gas of particles turns into a black hole.

Sources and links

First reported by anthropic.com. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news