Parisi and his collaborators used Claude to solve a long-standing mathematical problem related to the physics of “jamming,” according to a study published in the journal Jul. 1. Journal of Statistical Mechanics: Theory and Experiment.
Jamming refers to the point at which a collection of flowing particles suddenly becomes stiff while remaining disordered. According to , the concept was originally used to describe materials such as bubbles and particulate matter, but has since found applications in fields such as neuroscience and artificial intelligence. physics.org.
In a 2014 study, Parisi, Sapienza University of Rome physicist Francesco Zamponi, and their collaborators found that two mathematical quantities that describe how particles stick together at the point of disturbance always follow the same relationship. Repeated calculations confirmed this pattern, but the researchers could not explain why, and the problem remained unsolved for more than a decade.
 |
|
A person struggling with a complex formula on a blackboard. Pexels illustrations |
When generative AI models emerged, Parisi chose unsolved problems as a test of mathematical reasoning. Zamponi said Claude was chosen because he “appeared to have somewhat advanced mathematical reasoning abilities.”
The researchers first asked Claude to reproduce the calculations from the 2014 study. Within about 40 prompts, the AI generated core ideas for evidence, which researchers validated and refined into publishable results. This proof solves the puzzle and confirms that two independent theories describe the same underlying physical laws.
“A pure mathematician working full time on that kind of research…[s] If we solved the equation, we might have found a solution,” Zamponi said. live science. “But this is of particular interest to us, because it highlights how Claude gave us immediate access to a vast repository of mathematical training and formal skills that were just outside the usual realm.”
Zamponi said the experience changed his perspective on the role of AI in theoretical physics. While it made him rethink reasoning, intuition and creativity, he said he plans to continue leveraging AI to speed up routine tasks and provide new perspectives on difficult research problems.
