Humans vs Robots: AI Bluffs Poker Players

Poker AI, Libratus, defeats four of the world’s best poker players.

Professional poker players versed AI Liberatus. Source: The Guardian

Artifical Intelligence (AI) continue to advance in the gaming industry, with the most recent example being Libratus, a Poker AI, defeating four of the world’s best poker players. Held at the Rivers Casino in Pittsburgh, Libratus beat its rivals with an astounding $1.7 million in chips at closing.

The AI, manufactured by Carnegie Mellon University (CMU), successfully defeated competitors in a game of no-limit Texas Hold’em. Throughout the 20-day tournament, 15 of these days saw the AI claim victory.

Taking into account the complexity of this game that requires prediction and betting strategies throughout numerous hands, this emphasises how machine learning is progressing. Last year, AI defeated one of the greatest players in the complex board game ‘Go’ 10 years before experts expected.

However, AI’s victory in poker differs significantly from board games, as both players have full knowledge of the board throughout the game which makes it easier to predict your opponents moves.

In poker, the machine calculates with incomplete information, without knowing what cards his opponent holds; it has to calculate and bluff in a much more sophisticated way.

Adrian Weller, from the Centre for the Future of Intelligence at Cambridge University, commented on the uniqueness of AI’s victory with a lack of perfect knowledge.

“No-limit Texas Hold’em is a game of incomplete information where the AI must infer a human player’s intentions and then act in ways that incorporate both the direct odds of winning and bluffing behaviour to try to fool the other player,” he said.

What is even more intriguing is that if the players spotted a weakness in the AI’s strategy, it could quickly identify and address any issues at the end of the day with the help of three supercomputers.

The fact that AI can analyse situations and reach conclusions has unlimited potential. Co-founder of Libratus and Professor at CMU, Tuomas Sandholm, noted the potential applications of this.

Professor Sandholm, CMU. Source: CMU.

“It can be used in any situation where information is incomplete including business negotiation, military strategy, cyber security and medical treatment.”

Sandholm’s partner, Frank Pfenning, has similar ambitions for the future of Libratus with more practical uses, including cost-cutting.

“The computer can’t win if it can’t bluff,” said Frank Pfenning of CMU. “AI that can do that is a great step forward, with numerous applications. Imagine a smartphone that’s able to negotiate the best price on a new car for you.”

Although the win is momentous for AI, the system can only play in a more basic version Texas Hold’em called ‘Heads Up poker’ where there are only two players involved at any time.

Any more players and the calculations become too complex for the machine to process, however, it is only a matter of time before it is capable of winning in the more standard versions.

Libratus’ successes have brought to the foreground the question of whether the rise of the robots could soon displace humans, but this fear could be overblown. The key is to work with new technology, rather than against it.