A poker robot is currently taking on top professional poker players in a 20-day gaming session at Rivers Casino, Pittsburgh. The event, titled “Brains Vs. Artificial Intelligence: Upping the Ante”, pits four expert players against Libratus, a computer program designed by researchers at Carnegie Mellon University.

Brains Vs Artificial Intelligence

During the event, over 120,000 hands of Heads-Up No-Limit Texas Hold’em will be played with the winner receiving a prize of $200,000. Currently the robot is beating the expert poker players – Jason Les, Dong Kim, Daniel McAulay and Jimmy Chou – hands down.

In the first day of the tournament, Libratus won $81,716 compared with the human players’ $7,228. The robot then doubled its lead on the second day and continued to outfox the humans on the third.

Liberatus poker robot

Poker poses a difficult challenge for AI as it requires a machine to make complicated decisions based on incomplete information. The robot also has to contend with bluffs, slow play and other clever tactics.

The main problem for artificial intelligence is that in poker, players cannot see their opponents’ hands. This means that, in contrast to most board games, not all of the information in the game is available to them.

Board games and robots

Robots have successfully beaten human players in games such as chess in 1997, Jeopardy! in 2009 and with the board game Go in 2016 but poker has posed a challenge too far – until now.

The research team at Carnegie Mellon University said: “Since the earliest days of AI research, beating top human players has been a powerful measure of progress in the field”.

Libratus poker robot has been impressing observers with its human-like poker play, which includes bluffs such as making very small wagers and huge over-bets.

Libratus analyses poker rules

The machine takes its own cards into consideration when deciding whether to bluff, unlike its predecessor robot from 2015, called Claudico, which was defeated by poker players due to weaknesses in its bluffing system.

Libratus learned and developed its knowledge by analysing poker rules, not by trying to copy human players. This means it has developed sound mathematical strategies and has not picked up any weaknesses.

The robot spent around 15 million hours on the Bridges computer at the Pittsburgh Supercomputing Centre. It continues to train each evening – not needing to sleep, unlike its human opponents!

Applying the AI to other situations

Although Libratus is currently being used as a poker robot, its algorithm can easily be applied to many other situations that involve incomplete information. The programming could in the future be used in cybersecurity, military strategy and business, and may also be applied to medicine.

Games are good way to test AI before using it in other circumstances. Simulated environments modelled on the real world can be useful as it gives researchers chance to observe the behaviour of robots in real time.

Last year, Facebook and Google announced its desire to build a robot that would be pitted against humans playing Starcraft. This is another difficult game for AI to master due to the advanced planning and decision-making needed.

Deepstack poker robot

Libratus is not the only poker playing robot to put real players to the test. A team of at the University of Alberta have developed DeepStack, a program which has played around 45,000 hands of poker against many different players, beating them all easily.

Deepstack’s intelligence includes analysis of the playing history of its opponent on order to second-guess their decisions.

The current “Brains Vs. Artificial Intelligence”poker tournament will continue until 30th January. But at the moment, it looks like the Libratus robot will claim a comfortable victory over the human players.

The future of AI in gaming

We think it is interesting that robots are getting so good at playing poker. Could a robot sneak into a tournament in the future looking like a human and outfox all the players, taking the tourney prize for its human developers?

How will artificial intelligence affect poker and other games in the future?

Let us know your thoughts in the comments.

 

Leave a Reply

Your email address will not be published. Required fields are marked *