Superhuman AI for Multiplayer Poker

3/5/20 | 4:15pm | E51-325
Reception to follow.





Tuomas Sandholm

Carnegie Mellon

Abstract: In recent years there have been great strides in AI, with games often serving as challenge problems, benchmarks, and milestones for progress. Since the 1950s, poker has served as such a challenge problem in AI, game theory, and OR. Past successes in such benchmarks, including poker, have been limited to two-player games. However, poker in particular is traditionally played with more than two players. Multiplayer games present fundamental additional issues beyond those in two-player games, and multiplayer poker is a recognized AI milestone. In this paper we present Pluribus, an AI that we show is stronger than top human professionals in six-player no-limit Texas hold’em poker, the most popular form of poker played by humans. To our knowledge, this is the first superhuman AI for a multiplayer game. It is based on our new techniques such as depth-limited lookahead for imperfect-information games and equilibrium-finding algorithms that are significantly more scalable than prior approaches. This is joint work with my PhD student Noam Brown.

Bio:Tuomas is Angel Jordan Professor of Computer Science at CMU and Co-Director of CMU AI. He holds appointments in the Computer Science Department, Machine Learning Department, Ph.D. Program in Algorithms, Combinatorics, and Optimization (ACO), and CMU/UPitt Joint Ph.D. Program in Computational Biology. He is the Founder and Director of the Electronic Marketplaces Laboratory. In parallel with his academic career, he was Founder, Chairman, and CTO/Chief Scientist of CombineNet, Inc. from 1997 until its acquisition in 2010. During this period the company commercialized over 800 of the world's largest-scale combinatorial multi-attribute sourcing auctions with $60 billion in volume and over $6 billion in generated savings. His CMU algorithms run the UNOS national kidney exchange, which includes 173 transplant centers, that is, 73% of the transplant centers in the US. Since the founding of the exchange in 2010, his algorithms make the life-and-death kidney exchange decisions autonomously for those centers together each week. He is Founder and CEO of Optimized Markets, which is bringing a new optimization-powered expressive market paradigm to advertising campaign sales, scheduling, and pricing—in linear and nonlinear TV, Internet display, video and audio streaming, mobile, game, radio, and cross-media advertising. He is Founder and CEO of Strategic Machine, which is fielding his game-solving techniques to business, recreational gaming, and finance applications. He is Founder and CEO of Strategy Robot, which is fielding his game-solving techniques to defense and intelligence applications. Among his honors are the IJCAI Minsky Medal, Computers and Thought Award, inaugural ACM Autonomous Agents Research Award, Allen Newell Award for Research Excellence, Edelman Laureateship, Sloan Fellowship, Carnegie Science Center Award for Excellence, and NSF Career Award. He is Fellow of INFORMS, ACM, and AAAI. He holds an honorary doctorate from the University of Zurich.

Event Time: 

2020 - 16:15