Scaling Rational Cooperative Intelligence for Pluralistic AI Futures

Attend this 58-minute conference talk exploring how to develop cooperative artificial intelligence systems for pluralistic AI futures where multiple AI agents, systems, and services operate across diverse applications and stakeholders. Learn about rational cooperative AI, a paradigm for building coherent AI agents that are inherently cooperative through rational inference over intended goals and constraints, avoiding the unreliability of current generative AI while promising significant efficiency improvements. Discover recent research on automated negotiation and market intermediaries as institutional technologies that could enable efficient solutions to social dilemmas, while examining why these systems require deeper understanding of human values and rationality to preserve what matters most to humanity. Explore the computational foundations of human cooperation and normativity, and understand how reverse engineering these processes could enable development of human-level AI cooperators and cooperative infrastructure for an increasingly automated world. Gain insights into scaling cooperative intelligence through rational AI engineering that spans AI alignment, probabilistic programming, and computational cognitive science, with practical applications for creating AI systems that promote mutual flourishing rather than conflict or anti-human outcomes.