This time, Alina Bazarova, Thomas Lippert, Director of Jülich Supercomputing Centre (JSC) Centre, and CESOC Director Martin Schultz spoke with Helga Hermanns about why exascale computer JUPITER is particularly suited for AI, what sets it apart from previous supercomputers at Forschungszentrum Jülich, the high level of interest from researchers, and its future role in the global exascale landscape.
In short: JUPITER is essential for training very large AI models. Beyond the well-known large language models, there are specialised large models for weather and climate simulations. Training models that capture the complex interactions of the climate system and generate high-resolution predictions for local impact assessments requires enormous computing power. International collaboration is key, and we’re excited to work with strong partners from academia and operational centers like European Centre for Medium-Range Weather Forecasts (ECMWF) and Deutscher Wetterdienst (DWD).
Martin Schultz leads the Earth System Data Exploration (ESDE) research group at JSC and has long been involved in AI applications for climate and meteorology, from predicting extreme events like heavy rainfall and thunderstorms to developing foundational AI models for climate and atmospheric research.
Alina Bazarova, in the Applied Machine Learning & AI group led by Stefan Kesselheim, applies AI language models to protein analysis for biotechnology and medicine. On JUPITER, she’s now extending this work to discover improved enzymes and develop new drugs and diagnostics, including for cancer research.
Put your headphones on and enjoy: WDR5 Wirtschaftsmagazin.