DeepMind’s AlphaGo and its descendants learned to play the board game of Go better than any human player in the history of the game. Later versions of AlphaGo learned only through self-play, with no input from human experts beyond the rules of the game. This was made possible by an elegant, principled, and intuitive approach to machine learning — called reinforcement learning — that directly addresses the core problem of artificial intelligence: how to interact with the world in order to achieve desired outcomes. Reinforcement learning agents learn from their own experience, through trial and error, by observing the consequences of their actions and making appropriate adjustments to their behaviour. Application areas include healthcare (e.g. personalised treatments), drug discovery, robotics, finance, traffic control, power systems, autonomous vehicles, natural language processing, and nuclear fusion. This talk will provide an introduction to reinforcement learning, describe some of its applications, and discuss its future potential in the pursuit of artificial intelligence.

Bio

Özgür Şimşek is a Professor of Artificial Intelligence and Deputy Head of Computer Science at the University of Bath, where she leads the Artificial Intelligence and Machine Learning Research Group. She conducts research in artificial intelligence, focusing on machine learning, drawing from a strong multi-disciplinary background that encompasses computer science, engineering, operational research, and psychology. She is particularly interested in the theory and applications of reinforcement learning. Prof Şimşek received her PhD in Computer Science in 2008 from the University of Massachusetts Amherst. She subsequently joined the Centre for Adaptive Behaviour and Cognition at the Max Planck Institute for Human Development in Berlin, Germany, first as a Postdoctoral Research Fellow, then as a Research Scientist. She joined the University of Bath in 2017. She is the Machine Learning Theme Lead of the UKRI Centre for Doctoral Training in Accountable, Responsible and Transparent Artificial Intelligence (ART-AI). Prof Şimşek is a co-author of Classification in the Wild: The Science and Art of Transparent Decision Making (2020, MIT Press).