4 November 2026

Photo: UHH/Wegner

The meeting brought together members of the Cluster of Excellence Quantum Universe and the CRC 1624 and had a mixture of graduate students, postdoctoral students, and senior faculty.

AI tools are rapidly becoming part of everyday research at the Cluster of Excellence Quantum Universe and the Collaborative Research Center 1624. From writing code to constructing mathematical proofs, large language models (LLMs) are finding their way into the workflows of researchers across all career stages - and the community is actively shaping how this happens.

"The exceptional growth in capabilities of large language models in areas physicists and mathematicians care about, for example writing computer programs or constructing proofs, means these tools are of great interest to our researchers," says Principal Investigator Gregor Kasieczka. What is particularly notable is how organically the adoption has developed: "This potential integration of such AI tools is very broad and happens from the bottom up, with PhD students and young postdocs finding new and creative uses of these tools."

This enthusiasm is matched by dedicated research efforts. A research area within Quantum Universe is specifically focused on developing new AI and machine learning approaches for fundamental physics. Decades of experience with large datasets and complex algorithms have made particle physics and astronomy natural early adopters of these techniques, and that expertise now feeds back into the development of AI itself. Researchers at the University of Hamburg and DESY are building foundation models, systems that learn from large and diverse datasets and then perform well on entirely new problems with very little additional training, as well as generative models to simulate particle interactions in detectors at a fraction of the computational cost of traditional methods.

To keep pace with a rapidly changing landscape and share knowledge across groups, Kasieczka recently organized a meeting bringing together members of Quantum Universe and the CRC 1624. The turnout reflected genuine demand: graduate students, postdoctoral students, and senior faculty attended, and the conversations revealed an unexpected common ground. Experimental physicists and mathematicians are discovering shared interests through shared technology, finding new ways to collaborate that cut across traditional disciplinary boundaries. Further meetings are already planned.

Kasieczka is frank about both the promise and the work still ahead. "Technically, these tools are already at the level where they can carry out complex programming or data analysis tasks. However, we still need substantial work in figuring out how to make these models reliable and validate their output, so we do not give up the required precision in our fields." The ambition, he stresses, is not simply to automate existing workflows: "We need to use these tools to produce better results and more ambitious research, not just be happy that they could automate our old work. And in the end, human physicists and mathematicians are responsible for the result."