A research team led by Associate Professor Atsushi Ishikawa and doctoral student Taishiro Wakamiya of Institute of Science Tokyo (Science Tokyo) developed a method enabling AI to propose promising catalyst materials with both high activity and durability for fuel cells, which generate electricity from hydrogen and oxygen but rely on expensive and scarce platinum (Pt) catalysts. The team created a workflow where AI proposes candidates and learns from evaluation results, allowing efficient searching through vast possibilities while keeping computational costs manageable. Without being told what atomic arrangement leads to high-performance catalysts, the AI began identifying patterns on its own through repeated learning cycles, discovering features associated with high-performance catalysts and proposing structures resembling those previously identified as promising.