Anthropic's Claude Solves Nine-Loop Physics Challenge

Overview

Anthropic has successfully computed a nine-loop scattering amplitude in N=4 super Yang-Mills theory, a feat that researchers previously deemed computationally prohibitive. The achievement demonstrates that large language models can tackle frontier problems in theoretical particle physics using standard academic resources, bypassing the traditional hardware barriers.

This milestone answers a public challenge issued by science writer and former physicist Matt von Hippel, who sought to determine if artificial intelligence could solve complex problems limited only by computing power rather than a lack of conceptual innovation.

The Challenge

Matt von Hippel, a frequent contributor at 4gravitons.com, initiated the challenge to test the practical limits of current AI models in his former field. He specifically asked AI companies to solve one of the big outstanding problems in the field of scattering amplitudes: computing results for N=8 supergravity at seven loops or N=4 super Yang-Mills at nine loops.

If AI companies want to impress people like me, they need to tackle my old field. Show that an AI can take the kinds of computer resources an academic has access to, and solve one of the scattering amplitudes field’s big outstanding problems.

Hippel’s goal was to see if an AI could overcome a computational ceiling that human researchers had encountered not because the problem was theoretically unsolvable, but because it required more time and processing power than was reasonably available. While LLMs have recently shown impressive strides in pure mathematics, Hippel wanted to test a domain where the difficulty was purely computational rather than creative.

Understanding the Physics

The problem resides in amplitudeology, a branch of theoretical particle physics focused on calculating scattering amplitudes. These formulas allow physicists to predict the likelihood of subatomic particles reacting in specific ways based on their momenta and energies. Accurate predictions are essential for interpreting data from experiments like the Large Hadron Collider and for searching for evidence of new physics, such as dark matter.

Computing these formulas is extremely difficult. Consequently, physicists typically rely on approximations that stop at a specific number of loops, which measure the complexity of particle interactions. Most calculations in the field are limited to two or three loops, with only a handful reaching five.

To develop new techniques, researchers often test them on "toy models" like N=4 super Yang-Mills. While this theory is not a realistic description of the real world—it predicts four supersymmetric partners for every particle rather than one—it is mathematically elegant and serves as a rigorous stress-test for new computational methods.

Claude’s Achievement

Anthropic’s team utilized Claude to perform the nine-loop calculation for N=4 super Yang-Mills. This level of precision is orders of magnitude beyond what was previously considered feasible using conventional computing budgets. The calculation required handling an astronomical number of terms and symmetries that would typically overwhelm standard supercomputing clusters assigned to academic projects.

The success suggests that AI systems can now act as effective research assistants for heavy computational lifting in theoretical physics. By automating the tedious and expansive algebraic manipulations involved in high-loop calculations, Claude freed researchers to focus on interpretation and further theoretical development.

This result challenges the skepticism of those who argue that LLMs have hit a ceiling in their ability to perform deep, specialized work. It also counters the fears of those who believe AI is merely a parrot without genuine problem-solving capabilities, demonstrating instead that these models can navigate highly constrained, technical domains with precision.

Implications for Science

The ability to compute higher-loop amplitudes has significant implications for the future of particle physics. More precise theoretical predictions allow scientists to compare experimental data with greater accuracy. If discrepancies arise between observation and high-precision calculation, they may point toward new physics beyond the Standard Model.

Furthermore, this achievement validates the use of AI as a tool for exploring mathematical frontiers. It opens the door for other complex problems in physics and mathematics that were previously abandoned due to sheer computational cost. Researchers may now tackle problems that were once thought to be out of reach, potentially accelerating discoveries in quantum field theory and string theory.

While the nine-loop result was achieved in a toy model, the techniques and workflows developed can potentially be adapted for more realistic theories. As AI models continue to improve, their role in scientific discovery is likely to expand from simple information retrieval to active participation in complex research workflows.