Large-scale research facilities produce incredible amounts of data. There is no doubt that computational science can provide sustainable solutions to extract information and knowledge from Big Data. In late March 2023, more than 20 scientists from Europe and the US met at Schloss Dagstuhl – Leibniz-Zentrum für Informatik in Wadern, Germany, for an in-depth seminar about that topic. The aim was to develop a strategic vision on how to move jointly towards facilities that employ useful artificial intelligence (AI)-based solutions. The week-long event was organized by Peer-Timo Bremer, Brian Spears (both Lawrence Livermore Laboratory, Livermore, USA), Michael Bussmann (CASUS at Helmholtz-Zentrum Dresden-Rossendorf, Görlitz, Germany) and Tom Gibbs (NVIDIA Corp., Santa Clara, USA.

“Large-scale experimental and computing facilities could greatly benefit from a more automated and integrated approach”, says CASUS Founding Manager Dr. Michael Bussmann. The high energy density physics expert has been working e.g. at the European X-Ray Free-Electron Laser Facility (European XFEL) near Hamburg. “My various experiences at different research facilities convinced me that innovative methods from mathematics, system theory, data and computer science are needed to advance data-intensive research. Ultimately, this resulted in the foundation of the Center for Advanced Systems Understanding where we develop such innovative methods”, Bussmann adds.

While AI is one of many tools used at CASUS, it was center stage at the Schloss Dagstuhl seminar. Titled “AI-Augmented Facilities: Bridging Experiment and Simulation with machine learning (ML)”, the seminar brought together experimental and computational scientists, experts on edge and HPC computing, and machine learning researchers. The participants’ experiences came from a variety of application areas ranging from physics to life sciences to manufacturing. But they all had a common goal in mind: integrating predictive computation and simulations on one side and world-class experimentation on the other side with the help of AI. “A common language, or at least a coordinated process could speed up the exchange of insights from from one side to the other. In the end, both sides – simulations and experiments – will benefit heavily from this integration”, says Bussmann.

In the course of the week, the participants drafted a perspectives paper that they aim to publish in a peer-reviewed journal later this year. Besides Bussmann, also the designated director of CASUS, Prof. Thomas D. Kühne, was among the seminar participants. “I enjoyed the highly productive atmosphere of the event. This topic will gain importance in the years to come and it will certainly continue to be on the research agenda of CASUS”, Kühne reckons.