How to build Skynet = Adaptive Multi AI Agents

In this podcast, we delve into a fascinating new study called AdaSociety, conducted by researchers from Peking University, New York University, Tsinghua University, and the University of Science and Technology of China. The study investigates adaptive multi-agent AI systems within a unique environment that features evolving physical and social structures. Here, AI agents learn to navigate complex social dynamics through cooperation and competition, forming coalitions, hierarchies, and communities. While traditional reinforcement learning methods often struggle with such intricacies, the study finds that curriculum learning is a game-changer, enabling agents to gradually adjust to increasing social complexity. The podcast wraps up by discussing the broader implications of this research for creating collaborative and adaptive AI systems, while also cautioning against the overhyped notion of "superhuman intelligence" and advocating for a focus on distributed, cooperative AI instead.