Benji Rosman: Acting Rationally, The Challenges of Building Intelligent Agents

About Benji Rosman

Our most recent meetup, in collaboration with Conscium, focused on one of the central challenges in AI: how to build intelligent agents that act rationally in the world. We were honoured to be joined by Professor Benjamin Rosman, who was recently named in The Times 100 Most Influential People in AI.

Benji is a professor at the University of the Witwatersrand in South Africa, where he heads the Robotics, Autonomous Intelligence, and Learning (RAIL) Laboratory. Benji is also the founding director of the Machine Intelligence and Neural Discovery (MIND) Institute, which explores the science of intelligence across humans, animals, and machines.

In his talk, “Acting Rationally: The Challenge of Building Intelligent Agents”, Benji walked us through his research on reinforcement learning, knowledge generalization, and decision-making. He emphasized that building intelligent agents isn’t just about performing a single task well, but about developing systems that are flexible, instructable, and reliable. From robots navigating office environments to agents learning abstract skills, his work highlights both the technical challenges and the broader implications for how intelligence may evolve in artificial systems.

This session was part of PRISM’s ongoing series on conscious AI, and Benji’s insights offered a valuable perspective: focusing on intelligence as a foundation while opening pathways to questions about consciousness.

We’re grateful to Benji for joining us and to everyone who contributed thoughtful questions. Stay tuned for details of our next meetup, where we’ll continue exploring the science and philosophy of AI.