Thesis talk titled "Reinforcement Learning Beyond Rewards: Decision Making in the Language of Visitations".
Missing first 30 seconds. Transcript: "Hi everyone, thank you for coming to my thesis defense. Let’s get started. In the course of past few years we have seen a huge increase in AI adoption and a number of these AI system are decision making agents. Reinforcement Learning was developed as the framework to learn sequential decision-making from tasks specified through reward functions but this quickly becomes inconvenient as most people interacting with these AI agents lack technical expertise. Instead, they treat these AI agents as their collaborators and interact and give feedback to these systems as they would with other humans. As researchers, we need to think about these interactions and leverage them to make the systems better. In my thesi,s I take foundational steps towards developing a learning framework that embodies these principles."