Keynote: Noam Brown, Research Scientist at OpenAI in Conversation with Joe Spisak
summary
Leads reasoning work at OpenAI and traces a path from superhuman poker AIs to Cicero diplomacy agent that placed second in its league yet was preferred by human players, to current inference-time scaling and multi-step reasoning methods that already speed literature reviews and small proofs. Claims models are starting to contribute large chunks of research papers within one to two years, highlights sample inefficiency of RL as the main bottleneck, argues multi-agent reasoning buys parallelism and diversity, and treats open-sourced environments as low risk.
AI-written summary · about
chapters (10)
- 0:00Career background
- 2:27Beyond games to AGI
- 4:21Scientific contribution
- 5:58Verification in AI
- 9:13Open source and safety
- 12:46Reinforcement learning
- 15:37The research through-line
- 18:51Audience Q&A: RL paradigms
- 22:11Multi-agent systems
- 24:48Advice for researchers
related talks
19mRedefining Intelligence: How Reasoning Is Re-Shaping AI in 2025
30mKeynote: Fireside Chat
33mKeynote: Fine-tuning our way towards openness in AI
16mKeynote: Open Source for AI and AI for Open Source
05mKeynote: How Open Protocols Unleash AI’s Next Wave of Creativity and Impact
25mFrontier AI Models: A Review of Past Nine Months
08h 52mWF2026: Autoresearch & Keynotes ft. Anthropic, Google DeepMind, Amazon AGI, Sonar, Arena, Recursive
19mTraining Agentic Reasoners