Nobel Laureate Utilises AI to Solve Decade-Old Physics Problem

Artificial intelligence (AI) technology has helped Nobel Prize in Physics laureate Giorgio Parisi solve a mathematical problem in physics that had left scientists at a dead end for more than a decade. The findings were published in the Journal of Statistical Mechanics: Theory and Experiment in early July 2026. Together with physicist Francesco Zamponi from Sapienza University of Rome, Parisi utilised Claude, a generative AI model made by Anthropic, to find a mathematical proof that had previously eluded conventional methods. This success has drawn attention because it demonstrates that AI can play a role in scientific research requiring high-level reasoning. The story began in 2014 when Parisi and his team were researching the jamming phenomenon, a condition where a collection of particles transitions from moving freely to becoming interlocked due to extremely high density. This phenomenon is commonly found in granular materials, glass, and complex particle systems. In that research, they discovered two mathematical parameters, a and b, which always satisfy the equation a + b = 1. Unfortunately, no mathematical explanation could prove why this relationship was always true. Other research conducted by physicist Matthieu Wyart using a different approach also produced a similar relationship. Although this strengthened the suspicion of an as-yet-ununderstood physical principle, various attempts to find an analytical proof continued to fail for over 10 years. After years of deadlock, Parisi tried a different approach by using Claude as a scientific discussion partner. The AI was asked to study previous research before searching for a mathematical proof of the unsolved equation. According to a Live Science report, in only about 40 prompts, Claude managed to propose a new approach. Although it still contained some technical errors, the main idea was deemed correct. Armed with this idea, Parisi and Zamponi then verified and refined all the calculations to produce a complete proof, which was finally published in the scientific journal. Parisi stressed that AI did not replace the role of scientists in this research. Claude merely provided a new perspective that helped open a way out of a problem that had remained unsolved for years. The entire process of verification, calculation refinement, and result validation was still carried out by the researchers. This success serves as an example of how generative AI can function as a research aid to solve highly complex mathematical and physics problems. A number of experts assess that collaboration between humans and AI has the potential to change the way research is conducted in the future, from formulating hypotheses to exploring various possible solutions, while scientific decisions remain in the hands of researchers.