Discovered Materials, a San Francisco-based Y Combinator startup, has launched a $1 million prize challenge for AI models to solve five major problems in materials science. The company, which uses artificial intelligence to find new substances for advanced chips, recruited an academic advisory group from leading global universities to define the five open problems. These challenges include explaining cuprate superconductivity, predicting structures of silicon and silicon dioxide interfaces, mapping grain structure to thermal properties, predicting light-induced behavior changes, and improving density functional theory.
Before launching the public prize, Discovered Materials tested frontier lab models on these problems by giving AI agents internet access and autonomous work time for days, but the models made little headway. The startup hopes the $1 million prize pool will establish clear goalposts for artificial intelligence progress in a complex domain where problems lack the straightforward yes-or-no definitions found in mathematics.
This initiative highlights a shift in scientific inquiry as artificial intelligence demonstrates the ability to uncover unexpected insights and techniques. By establishing these verification challenges first, the startup aims to deepen humanity's understanding of physics and shape the future manufacturing of electronics and energy storage materials.