AI Reveals the Secrets of Planetary Systems

(University of Bern, September 12, 2025)

Researchers from the University of Bern and partners have created an AI that quickly predicts the layouts of planetary systems, greatly helping to uncover new planets. This AI, developed by Prof. Yann Alibert, Dr. Sara Marques, and Dr. Jeanne Davoult, speeds up the process a million-fold compared to old methods, saving observation time and boosting the odds of finding Earth-like planets. It uses the Transformer architecture, similar to that in advanced language models, to anticipate planet sequences from initial properties, enhancing target selection for planet observation and advancing our grasp of planet formation. This method, fueled by tens of thousands of simulations, not only streamlines the search for habitable worlds but also redefines our approach to studying the cosmos. By merging AI technology with astronomical expertise, the University of Bern's team opens new pathways in the quest to decode the universe's secrets, offering a smarter way to pinpoint potential Earth-like planets and understand planetary system formation.