The growth of Artificial Intelligence (AI) has been staggering in recent years with expectations that the trajectory won’t be slowing down anytime soon. One benchmark of this progress is how AI performs against humans in competitive game play. To date, AI has proven its ability to defeat top human players in checkers, chess, and more recently Go. However, No Limit Texas hold’em, the most popular form of poker, presents an imperfect information game that has presented challenges for AI to conquer.
In a newly published piece in Science Magazine, Noam Brown and colleagues developed an AI system named Libratus, which is capable of developing winning strategies for imperfect knowledge games. As they state, “hidden information makes a game far more complex for a number of reasons. Rather than simply search for an optimal sequence of actions, an AI for imperfect-information games must determine how to balance actions appropriately, so that the opponent never finds out too much about the private information the AI has.” The example used is the common practice of bluffing, which is necessary in competitive poker strategy, but bluffing all the time would be a losing strategy.
In a 120,000 hand competition, Libratus defeated four top human specialist professionals in heads-up no-limit Texas hold’em. As the authors describe their “game-theoretic approach features application-independent techniques: an algorithm for computing a blueprint for the overall strategy, an algorithm that fleshes out the details of the strategy for subgames that are reached during play, and a self-improver algorithm that fixes potential weaknesses that opponents have identified in the blueprint strategy.”
Importantly the real world is filled with “games” that involve strategic interactions with imperfect information and thus the development of Libratus may become an important stepping stone in the widespread application and further adoption of AI.
The paper, Superhuman AI for heads-up no-limit poker: Libratus beats top professionals, was co-authored by Tuomas Sandholm.