Poker bots have shown, once again, that artificial intelligence is better than at least some humans at playing No Limit Texas Hold’em.
The latest man vs. machine showdown between one of Carnegie Mellon University’s poker playing computers took place recently in Hainan, China. Although not a challenge in the same way that the Libratus match was when it took place earlier this year, it was another chance for the researchers behind the all-conquering poker bot to strike again.
Fronting Team Dragon, a group of six Chinese poker players, was Shanghai venture capitalist and WSOP bracelet winner Yue Du. With more than $800,000 in live earnings, Du is no slouch at the poker table, but even he couldn’t outsmart Libratus’ relation, Lengpudashi.
As per the rules of the challenge, Lengpudashi and Team Dragon were set to play 36,000 hands between April 6 and April 10. Despite being an exhibition match, there was $290,000 on the line for the outright winner and Lengpudashi had ready to play.
Poker Brain Does It Again
Like Libratus, the Lengpudashi software was connected to the Bridges computer at the Pittsburgh Supercomputing Center. This combination of AI software and hardware is capable of performing calculations unlike any poker bot before it and it’s this processing that earned machine another impressive win.
When the chips had fallen and all was said and done, Lengpudashi won by $792,000 in virtual money. That win not only left the Team Dragon without the $290,000 in real cash, but proved once again that AI is now more advanced that it’s ever been.
Back when Libratus beat a team of four US poker pros, it made international headlines. Because the predecessor to Libratus, Claudico, had lost to Doug Polk, Bjorn Li, Dong Kim, and Jason Les by $732,713 back in 2015, few gave the computer a chance heading into the January 2017 showdown.
Computers Solve the Hold’em Riddle
Old adages that No Limit Hold’em was too complex for computers to play effectively against highly skilled players dominated the pre-match talk. But just a few days into the 120,000 hand battle, Libratus was able to show that it was a markedly better player than Claudico.
In the end, despite the best efforts of Jason Les, Dong Kim, Jimmy Chou and Daniel McAulay, Libratus won by a margin of $1,766,250 in virtual chips. Following the result, Noam Brown, a Carnegie Mellon graduate who worked on the poker bot, said that Libratus develops its own strategies from scratch.
“We give the AI a description of the game. We don’t tell it how to play. It develops a strategy completely independently from human play, and it can be very different from the way humans play the game,” Brown told Wired back in February, 2017.
The latest result might not go down as a “scientific” result on account of the small sample size, but it’s certainly shown once again that AI is more than capable of holding its own at the Hold’em tables.