Moral AI in the News: Poker-Playing AI

Game-playing computers have made headlines for besting human players at Jeopardy, chess, and go in the past. However, all of the aforementioned are games of perfect information: the computer had access to every piece of information about the game every time its opponent made a move. In January 2017, the AI Libratus defeated professional poker players in a tournament of heads-up no-limit Texas hold’em. Libratus trained by playing against itself in thousands of hands of the game, testing strategies’ effectiveness from the bottom up. The supercomputer’s repeated wins are an important step forward because poker is an imperfect information game: the computer has to guess information, and it has to bluff against its opponents. Furthermore, the AI could learn from experience, adapting its strategies to even greater levels of impenetrability at the end of every poker session. The ability for AI to perform well in situations where it does not have access to all possible relevant information has important implications for its involvement in economics, another imperfect information game of a sort – albeit a far more extensive and complex one.

8/31/16 Article describes a Libratus predecessor belonging to the same creators:

1/23/17 Article gives an overview of AI playing games in the context of Libratus (and quotes Vincent Conitzer!):

1/24/17 Article describes the Libratus vs human poker tournament and discusses corporate uses for and involvement in AI advances:

1/24/17 Article discusses Libratus in the context of AI playing games:

1/25/17 Article discusses how Libratus makes bets differently from humans and how it learns from and corrects its poker weaknesses:

1/30/17 Article describes how Libratus played against and beat high ranking professional poker players:

1/31/17 Article gives overview of Libratus’s strategy:

1/31/17 Article gives an outline of the poker tournament between Libratus and human players:

1/31/17 Article gives details about the games between Libratus and human players as well as the AI’s strategies, and human concerns about what the AI’s win means for poker:

1/31/17 Article describes how Libratus played against human players, and the human players’ opinions on playing against a machine:

2/1/17 Article gives details about how Libratus functions:

2/1/17 Article describes what Libratus is and how it learned to play poker:

2/4/17 Thought article that describes the possibility of deceptive AI, as well as some of the ways AI in daily life could be beneficial or harmful:

2/4/17 Article addresses the potential applications of poker-playing computers outside of poker:

2/17/17 Article describes Libratus and game-playing AI, and makes some speculations about the future of AI in relation to the trolley problem:

3/3/17 Article describes some the algorithms Libratus uses:

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at

Up ↑

%d bloggers like this: