+- MKLab (https://mklab.gr)
+-- Forum: [INDEX] (https://mklab.gr/forumdisplay.php?fid=1)
+--- Forum: MATHEMATICS (https://mklab.gr/forumdisplay.php?fid=3)
+---- Forum: ARTICLES (https://mklab.gr/forumdisplay.php?fid=13)
+----- Forum: AI AND TECHNOLOGY (https://mklab.gr/forumdisplay.php?fid=158)
+----- Thread: AI model helps physics Nobel laureate (/showthread.php?tid=1725)
AI model helps physics Nobel laureate - mklabgr - 08-22-2026
AI model helps physics Nobel laureate out of a decade-old mathematical jam
Source:Physics World
Author: Isabelle Dumé
Published: 7 August 2026
Topic: Artificial intelligence, mathematical physics, complex systems / jamming theory
Physicists Giorgio Parisi, winner of the 2021 Nobel Prize in Physics, and Francesco Zamponi used Anthropic’s Claude to help resolve a mathematical question that had resisted them for roughly a decade. The problem comes from jamming theory, which studies how a collection of particles can suddenly become rigid while remaining disordered—for example, spheres packed so densely that none can move. In their 2014 work, Parisi, Zamponi and collaborators had numerically observed a striking relation between two critical exponents, $a$ and $b$, describing the distributions of contact forces and gaps between particles near the jamming transition:
$
a+b=1.
$
Although numerical calculations consistently supported the relation, the researchers had been unable to derive a satisfactory analytical explanation.
Rather than simply asking Claude to produce a proof, the researchers first had it reproduce their earlier numerical calculations. After it succeeded, they asked why $a+b=1$ should hold. Claude proposed a conceptual route that, while initially containing small errors, revealed the key idea. Through several rounds of human checking and refinement, this led to a proof connecting the researchers’ abstract, infinite-dimensional theory of jamming with a more physically intuitive framework developed by Matthieu Wyart and colleagues at EPFL. Zamponi emphasizes that the surprising contribution of the AI was not an enormously complicated calculation but recognizing a simple connection that the human researchers had overlooked after years of searching for a deeper hidden symmetry.
The episode illustrates an emerging role for LLMs as research collaborators or “telescopes for the mind”: systems that can reproduce calculations, explore alternative arguments and rapidly connect ideas from different areas of the literature. Zamponi believes such tools could lower barriers between scientific specialities and dramatically accelerate research, although he also warns that they can make it easier to produce large quantities of weak or pseudo-scientific work. For this reason, the researchers published their full interaction with Claude alongside the scientific work, arguing that transparency should become standard when AI contributes substantially to a discovery. Zamponi is already using the same approach on problems involving random sequential adsorption of hard hyperspheres, with connections to high-dimensional geometry and error-correcting codes.
Key takeaways
- AI did not simply output a finished proof: Claude supplied the crucial conceptual insight, while the physicists corrected, refined and verified the mathematics.
- The result explains the previously mysterious numerical relation $a+b=1$ in jamming theory.
- The most interesting lesson is that AI may sometimes advance science not through brute-force calculation, but by spotting simple connections experts have overlooked.
- The case strengthens the argument that future scientific work may involve a partnership in which humans formulate problems and verify results while AI searches rapidly through possible conceptual routes.
- It also raises an important norm for AI-assisted mathematics: disclose the AI interaction so that the origin and verification of new ideas remain traceable.
ARTICLE