Big Ideas Lab AI at the Lab
May 20, 2025
Brian Spears is the Director of the AI Innovation Incubator at Lawrence Livermore National Laboratory, and Kelli Humbird leads the computational physics group there. They discuss how AI, particularly through cognitive simulation, is revolutionizing scientific research by optimizing experiments and accelerating drug discovery. The duo reveals impressive applications, from enhancing fusion ignition experiments to predicting outcomes in national security. They highlight the importance of rigorous validation and envision a future where AI models assist scientists in generating innovative hypotheses.
AI Snips
Chapters
Transcript
Episode notes
AI-Driven Feedback Loop Accelerates Science
- Cognitive simulation pairs physics simulations, experimental data, and AI to create faster hypothesis–test cycles.
- Brian Spears says this loop lets researchers narrow thousands of possibilities into a few promising experiments.
CogSim Guided Fusion Ignition
- CogSim helped identify experiments that raised the predicted chance of fusion ignition above 50%.
- Brian Spears recounts that the tool told them they were more likely than not to ignite, a historic moment.
Simulation+Data Produces Practical Predictions
- Cognitive simulation generalizes beyond fusion to materials, national security, and drug discovery.
- Kelli Humbird explains simulations give a theoretical map while experiments correct its location and scale.