Beyond hill climbing: the path to superhuman scientific discovery

<p>At RAAIS 2026, Google DeepMind&#39;s Roberta Raileanu lays out a recipe for superhuman scientific discovery: AI systems that make groundbreaking discoveries across domains faster than people can. She walks through three ingredients - reinforcement learning to discover solutions where progress can be measured, open-ended divergent search to find new problems rather than climb known ones, and meta-learning to speed up discovery on problems no one has posed yet. The through-line: we can search for anything we can measure, but we still cannot measure what makes a discovery good. The bottleneck isn&#39;t the search. It&#39;s the signal.</p><p><strong>Chapters:</strong></p><p>00:00 - Introduction</p><p>00:51 - Defining superhuman scientific discovery</p><p>01:40 - The state of play: real progress, real plateau</p><p>06:58 - Ingredient one: discovery as reinforcement learning (Move 37, MLGym)</p><p>12:51 - Ingredient two: open-ended search and why greatness cannot be planned</p><p>18:30 - Rainbow Teaming: quality-diversity in practice</p><p>21:07 - Ingredient three: meta-learning the process of discovery (DiscoBench)</p><p>25:06 - The recipe, and the missing signal</p><p><br></p>