Now is the time to decipher what makes the brain both flexible and dependable—and to apply those lessons to AI—before an unaligned agentic system wreaks havoc.

By combining large language models with modular cognitive control architecture, Robert Yang and his collaborators have built agents that are capable of grounded reasoning at a linguistic level. Striking collective behaviors have emerged.

Many of the recent developments underlying the explosive success of artificial intelligence have diverged from using neuroscience as a source of inspiration—and the trend is likely to continue.