CogNovo research fellow Thomas R. Colin has presented an overview of some of his work at the Artificial Intelligence seminar at the University of Alberta (Canada). He seeks to reproduce insightful problem-solving in robots, using hierarchical reinforcement learning techniques to cause diversified exploration for difficult problems.
Thomas stayed for a two-months research visit at the Reinforcement Learning and Artificial Intelligence lab at the University of Alberta, improving his theoretical knowledge of Reinforcement Learning and exchanging ideas on artificial curiosity and creativity. The lab is directed by Richard Sutton and pursues the long-standing goals of AI (understanding the mind, reproducing human abilities) based on reinforcement learning ideas.