This podcast episode features Noam Brown, a prominent AI researcher, who recounts his transformative journey in AI game-playing, particularly through the lens of poker, showcasing how the integration of planning and search capabilities vastly outperformed traditional, model-based AI methods. Brown discusses the historical underestimation of planning in AI, its critical role in games like Go, Hanabi, and Diplomacy, and the implications of trade-offs between training and inference for scaling AI capabilities. He emphasizes the future of AI hinges on a broader understanding of general methods that enhance computational efficacy, calling for renewed focus on planning as a pivotal avenue for innovation in the field.
