What is the Center for Human-Compatible AI's research agenda?

The Center for Human-Compatible AI (CHAI) is an academic research group based at UC Berkeley, focused on the development of "provably beneficial [AI] systems".

Its research focuses on Cooperative Inverse Reinforcement Learning and Assistance Games, a new paradigm for AI where they try to optimize for doing the kinds of things humans want rather than for a pre-specified utility function. have been working on inverse reinforcement learning (where the AI infers human values from observing human behavior) and corrigibility, as well as attempts to disaggregate neural networks into “meaningful” subcomponents (see Filan, et al.’s “Clusterability in neural networks” and Hod et al.'s “Detecting modularity in deep neural networks”). (CHAI) have been working on inverse reinforcement learning (where the AI infers human values from observing human behavior) and corrigibility, as well as attempts to disaggregate neural networks into “meaningful” subcomponents (see Filan, et al.’s “Clusterability in neural networks” and Hod et al.'s “Detecting modularity in deep neural networks”).