Amadeus Bramsiepe, KIT; portrait: private

Artificial intelligence (AI) has become part of everyday life. The models are supposed to constantly improve, but there is a lack of awareness regarding their sustainability and efficiency. Charlotte Debus' aim is to increase the robustness and scalability of AI models, but also to improve their energy efficiency and carbon footprint. "Until now, breakthroughs in deep learning have always been accompanied by increased resource consumption," says Debus. She advocates for the introduction of a transparent benchmark to determine the energy consumption of an AI for the entire training period.

Sandra Goettisheim, KIT

Artificial intelligence (AI) can play a key role in the energy transition. To ensure its reliable application, its methods must be transparent and comprehensible. This is the goal pursued by the Helmholtz Investigator Group “Data-Driven Analysis of Complex Systems” (DRACOS), led by Tenure-Track Professor Benjamin Schäfer, which was evaluated and rated by KIT’s Executive Board as “extraordinarily successful.” Schäfer’s goal is to make AI models used to analyze large volumes of data from energy systems more transparent. Until now, it has often been unclear which factors an AI uses to predict household energy consumption.

Markus Götz, KIT

Wildfires, floods and droughts: a new artificial intelligence (AI) from KIT promises to be a game-changer in providing more precise, faster, and energy-efficient predictions of such events. In the "WOW - a World model of Our World" project, researchers will develop an AI world model that combines multiple specialized AI sub-models through shared “latent spaces”. "Modern AI methods can not only cost-effectively imitate physics-based simulations, but can even learn correlations directly from observational data," says project coordinator Peer Nowack. The Carl Zeiss Foundation is funding the project with six million euros.