Abstract
In both cognitive science and computer science, intelligence has traditionally been viewed solipsistically, as a property of unitary agents devoid of social context. Yet converging evidence in behavior, neuroscience and evolution shows that natural intelligence emerged at multiple scales in networks of interacting agents. For those interested in reverse-engineering intelligence, these findings suggest constraints on the space of workable algorithms. Recent breakthroughs in artificial intelligence (AI) run parallel to these findings, with specific population structures enabling agents to master complex strategic games like Capture-The-Flag and StarCraft II. We posit that moving beyond a solipsistic view of agency will benefit both cognitive science and machine learning: understanding intelligence demands a multi-agent context.
Authors
Edgar A. Duéñez-Guzmán, Suzanne Sadedin, Jane X. Wang, Kevin R. McKee, Joel Z. Leibo
Venue
Nature Machine Intelligence