This website uses cookies.

By clicking the "Accept" button or continuing to browse our site, you agree to first-party and session-only cookies being stored on your device to enhance site navigation and analyze site performance and traffic. For more information on our use of cookies, please see our Privacy Policy.

« Back to Results

Invited Session on AI, Machine Learning, and Economics

Paper Session

Friday, Jan. 5, 2018 10:15 AM - 12:15 PM

Marriott Philadelphia Downtown, Independence Ballroom II
Hosted By: Econometric Society
  • Chair: Leeat Yariv, California Institute of Technology

Super-human AI for Strategic Reasoning: Beating Top Professionals in Heads-up No-limit Texas Hold’em

Tuomas Sandholm
,
Carnegie Mellon University

Abstract

Poker has been a challenge problem in AI and game theory for decades. As a game of imperfect information it involves obstacles not present in games like chess and Go, and requires totally different techniques. No program had been able to beat top humans in large poker games. Until now. We describe Libratus, our AI that, in a 120,000-hand competition against four top human specialist professionals, became the first---and so far only---program to defeat top humans in heads-up no-limit Texas hold'em, a game that has 10^161 decision points and is the leading benchmark for testing algorithms for imperfect-information games. Libratus is powered by new algorithms in each of its three main modules: 1) an algorithm for computing a blueprint Nash equilibrium approximation for the overall strategy, 2) a safe algorithm that fleshes out the details of the strategy as play progresses, and 3) an algorithm for fixing Libratus’s own strategy to play even closer to equilibrium based on what holes the opponents have been able to identify and exploit. The algorithms are domain-independent and have potential applications to many real-world imperfect-information games such as negotiation, business strategy, cybersecurity, physical security, military applications, strategic pricing, product portfolio planning, certain areas of finance, auctions, political campaigns, and steering biological adaptation and evolution, for example, for medical treatment planning.

Machine Learning Methods for Economists

Guido Imbens
,
Stanford University

Abstract

Machine Learning Methods for Economists
JEL Classifications
  • C1 - Econometric and Statistical Methods and Methodology: General
  • C7 - Game Theory and Bargaining Theory