A new AI just beat professional no-limit Texas Hold ‘Em players for the first time

Could professional poker eventually become a battle of the bots?

A new artificial intelligence created by Carnegie Mellon researchers named Libratus just won at no-limit Texas Hold ‘Em poker after a 20-day tournament. The AI played against top poker pros like Daniel Macaulay, Dong Kim, Jason Les and Jimmy Chou, topping them by $1.7 million USD. This is the first time an artificial intelligence has bested top-ranked human players in this game, which is complex, requires long-term strategy and, among humans, a certain degree of intuition.

In no-limit Texas Hold ‘Em, the aim isn’t necessarily to win each hand, but to win the most money by the end of the tournament — which means employing cunning strategies that were previously believed to be unique to humans. The technology behind Libratus could bring AI breakthroughs in the financial sector, among other industries.

Libratus is the second attempt from researchers Professor Tuomas Sandholm and grad student Noam Brown, who challenged top human players with a similar AI two years ago and lost. In both attempts, the researchers may have altered the AI’s behaviour between the days of competition, but as WIRED points out, the human competitors also strategized with each other at the end of each day in an attempt to pull out a win for humankind.

Image credit: fielperson via Pixabay

[source]WIRED[/source]