AI, theoretical physics and LPTMS: to prompt or to be prompted?
Leonardo Mazza
The spectacular success of generative artificial intelligence has the potential to profoundly reshape the way science is performed. Mathematicians and translation scientists have experienced that already; theoretical physicists might be next. Is it possible to be prepared for this change? Some people argue that AI should be completely banned from science; others say that we should simply let it go, without any form of control. I will present a thesis according to which theoretical physics could be better prepared to the upcoming revolution by (i) capping the number of published papers and (ii) letting our policy-makers acknowledge that. If output volume stops being the metric, the incentive to flood the literature shifts toward slower, deeper work. A corollary of the discussion is the problem of AI and scientific training: is it possible to define the “little bit of ai” that is helpful and not harmful? I would like to know what the theoretical physicists working at LPTMS think about that, and how to improve the reasoning.
Ref. https://proofsandprompts.com/2026/10/05/ai-and-the-mismeasure-of-science/