When artificial intelligence was in its early stages of development, its creators were not envisioning technology that interacts with humans in the same way humans interact with each other. But as AI has advanced and been stirring our imaginations with its possibilities, the idea of human-compatible AI is gaining traction, says Subbarao Kambhampati, a Fulton Schools professor of computer science and a past president of the international Association for the Advancement of Artificial Intelligence. Reaching that goal will require overcoming big challenges, he says. AI must develop social intelligence and be capable of adopting the kinds of mental states that guide human interactions, and be able to understand human emotions and values. To see a future in which AI agents work successfully with us, Kambhampati says, there must be close collaborations between AI experts and those in other fields, particularly behavioral psychology, sociology and the humanities.