From Hype to Hypotheses: Why AI-for-Science Is Hard—and What It Takes to Make It Work

Will AI actually accelerate discovery—or are we still stuck at the “hype” stage? AI-for-Science matters because scientific environments impose real-world constraints and “TRADE”-offs across testing, resources, assurance, data, and expressivity that can prevent promising models from working reliably in practice. There is also a scientific readiness gap: without tools to assess and close it through performance testing, integration, and continuous monitoring, autonomous AI-enabled discovery remains out of reach. Attendees will learn how to think in terms of TRADE constraints and how to evaluate and stress-test AI systems for robustness, safety, data quality, and operational fit. They’ll also leave with a clearer view of what new research and end-to-end “AI technology stack” optimization are needed to make autonomous scientific discovery feasible. Attend to move from hype to hypothesis—and learn what it practically takes to partner AI with science to achieve real scientific objectives.