30.10.23
30.10.23
Embedded Files
Professor Frans Oliehoek and Dr. Robert Loftin
Professor Frans Oliehoek and Dr. Robert Loftin
Talked about
Talked about
Ad Hoc Teamwork is Theoretically Impossible, But We Should Do It Anyway
The talk was recorded and can be viewed here
Ad Hoc Teamwork is Theoretically Impossible, But We Should Do It Anyway
The talk was recorded and can be viewed here
Abstract
Most real-world applications of AI require agents that can cooperate with partners they have never encountered before, and whose strategies cannot be known in advance (i.e., humans). Any effort to create such artificial agents must confront the question of: what does is mean for an AI to be "good" at ad hoc teamwork? In this talk we survey attempts to provide a theoretical answer to this question, drawing connections (and highlighting the gaps) between theory and practical solutions. We then present our recent work [Loftin and Oliehoek, 2022] in which we show that, even after ruling out trivial failure modes, no single agent can learn to cooperate with all possible adaptive partners. This result implies that any definition of ad hoc teamwork that can be satisfied by some algorithm must necessarily impose restrictive assumptions on the target class of partner agents. We conclude by asking whether there exists a "universal" class encompassing all reasonable strategies, or whether we must necessarily incorporate background information about potential partners if we are to ensure successful cooperation.
About the Speakers
About the Speakers
Dr. Frans A. Oliehoek is an Associate Professor at Delft University of Technology, where he is a leader of the sequential decision making group, a scientific director of the Mercury machine learning lab, and director and co-founder of the ELLIS Unit Delft. He received his Ph.D. in Computer Science (2010) from the University of Amsterdam (UvA), and held positions at various universities including MIT, Maastricht University and the University of Liverpool. Frans' research interests revolve around intelligent systems that learn about their environment via interaction, building on techniques from machine learning, AI and game theory. He has served as PC/SPC/AC at top-tier venues in AI and machine learning, and currently serves as associate editor for JAIR and AIJ. He is a Senior Member of AAAI, and was awarded a number of personal research grants, including a prestigious ERC Starting Grant.
Frans Oliehoek
Frans Oliehoek
Delft University of Technology
Dr. Robert Loftin is a Lecturer in Machine Learning at the University of Sheffield. He received his PhD in Computer Science from North Carolina State University in 2019. He then completed a post-doc with Microsoft Research Cambridge (UK), exploring the use of Reinforcement Learning in commercial games (Xbox games). He also completed a post-doc at TU Delft with Dr. Frans Oliehoek, applying game theory to human-AI cooperation.
Robert Loftin
Robert Loftin
University of Sheffield
Page updated
Google Sites
Report abuse