Optimization alone won’t get us to humane AI. Current AI is based on the principle of maximizing some proxy for value — whether that’s clicks, dollars, or correct predictions. But no one knows how to turn “be good for humanity” into an optimization problem. CHAI is a research lab devoted to re-founding AI on provably beneficial principles. We develop theory, run experiments, educate the public, and advise on policy.

We are part of UC Berkeley’s world-class computer science department, but the problems we work on are deeply interdisciplinary. CHAI is a hub for a network of scholars and organizations spanning social science, law, philosophy, public health, economics, governance, and more.