Artificial intelligence can search vast spaces of possible experimental setups and discover new configurations that outperform designs developed by human intuition alone.
Designing a physics experiment often means choosing the right configuration from an enormous number of possible combinations of lasers, detectors, optical elements, and other components. Within an international research team we argue that artificial intelligence can systematically explore this vast design space and identify experimental concepts that humans may never have considered. Rather than relying on large language models, the approach treats experimental design as an optimization problem, using physical simulations to evaluate and improve candidate setups. Particularly in emerging fields such as quantum electron microscopy, where human intuition is still developing, AI could become a powerful tool for discovering entirely new measurement concepts.
TU Wien and derStandard , opens an external URL in a new windoware featuring our work, which was recently published in Nature.
J. Klimesch, S. Arlt, C. Ruiz-Gonzalez, C. Rodríguez, X. Gu, P. Haslinger, P. Vischia, C. Haack, Y. Drori, R. Adhikari, M. Arndt, M. Kagan, L. Heinrich, M. Krenn, Designing physics experiments with artificial intelligence, Nature (2026). DOI: 10.1038/s41586-026-10898-6
https://www.nature.com/articles/s41586-026-10898-6, opens an external URL in a new window