From Smith's Invisible Hand to Distributed Optimization and Control

Explore a 58-minute lecture where David Wolpert from the Santa Fe Institute delves into the fascinating connection between Adam Smith's economic concept of the invisible hand and modern distributed optimization systems. Learn how economic theory's invisible hand principle can be adapted for distributed control systems by implementing reinforcement algorithms for individual agents, mimicking human behavior in economies. Discover the innovative approach of designing agent reward functions to achieve Nash equilibrium that optimizes entire distributed control systems. Examine how these distributed control techniques extend to distributed optimization, encompassing methods like cross-entropy, genetic algorithms, and evolution of distributions. Understand the formal relationship between distributed optimization and machine learning, leading to enhanced optimization algorithms that achieve state-of-the-art performance. Presented at IPAM's Naturalistic Approaches to Artificial Intelligence Workshop at UCLA, this talk bridges economic theory, distributed systems, and artificial intelligence in a comprehensive exploration of optimization and control strategies.