AI’s triumph over a game of poker

The matter of artificial intelligence – or AI – is one that has long been an exciting topic for tech engineers and enthusiasts. Furthermore, in recent times, we have actually started to witness real developments in this field, making AI a reality instead of a science fiction fantasy. AI has been pitted against human intelligence in a number of different tasks, situations and games to find out how it will react to these challenges, with one of the latest examples of this testing being a game of poker in which AI was playing against human beings.

Earlier this year, an AI named Libratus was subjected to a pretty intensive test when
it was put up against a number of the finest and most successful poker players in the
world during a tournament lasting for 20 days. The human players who pitted their
wits against Libratus were Daniel McAulay, Jimmy Chou, Jason Les and Dong Kim,
and the tournament saw them playing against the AI via 11-hour sessions in front of
computer screens at Pittsburgh’s Rivers Casino. The chosen form of poker was
Texas Hold’em, but the results did not make easy viewing for those worried that one
day technology like AI will render us all redundant. Libratus comfortably defeated all
four of these top poker pros, claiming chips that amounted to winnings of $1.7 million.
Perhaps wisely, the human players were not competing for real money on this
occasion, but the results were a stunning vindication of AI technology. Libratus is the
brainchild of Carnegie Mellon University boffins and proved astonishingly flexible –
able to change strategies each day to best its opponents. This left the, perhaps
somewhat bruised, poker pros complaining that the boffins were adjusting it after play
ended each day, but there is no evidence to support this.

In case this was not enough of a chastening experience for humanity, a second AI –
this time named Lengpudashi – went on to repeat this triumph over a run of exhibition
poker games held in China during April. This time the competition lasted for five days,
but the results were equally comprehensive – as Lengpudashi won £230,000. The
human players that took it on included Yue Du, who has enjoyed success at the
World Series of Poker, but the other players were investors, computer scientists and
engineers, rather than pros. They tried to outthink the AI by deploying machine
intelligence and game theory, but this proved no more effective than the strategies of
the poker players in the earlier match-up.

While this doesn’t mean that the game is up for us humans when it comes to poker, it
does indicate that AI may well play a major role in the game in years to come –
whether it becomes part of the set up at online casinos or in some other form.