Libratus Offspring to Prove AI Win Wasn’t Luck in $290K Match

The legacy of Libratus is set to make an impact on Asia thanks to an exhibition match between Carnegie Mellon University’s artificial intelligence (AI) software and a team of Chinese players.

Libratus offspring Lengpudashi to take on a team of Chinese poker players in a $290,000 exhibition match. (Image: thefintechtimes.com)

When poker bot Libratus beat four US pros back in January 2017, it made the headlines around the world and that’s led to an invite from Team Dragons.

Fronted by Shanghai venture capitalist and WSOP bracelet winner Yue Du, the team will take on an AI poker bot named Lengpudashi between April 6 and April 10.

Another Chance for AI to Shine

Although the bot isn’t the same one that triumphed in the man vs. machine challenge earlier this year, it will be powered by the same AI software. Lengpudashi will be connected to the Bridges computer which is housed inside the Pittsburgh Supercomputing Center.

As well as the change of opponent, Professor Tuomas Sandholm, the lead researcher behind Libratus, has stressed that this isn’t a scientific experiment. When Libratus played Jason Les, Dong Kim, Jimmy Chou and Daniel McAulay at Rivers Casino in Pittsburgh, it did so over the course of 120,000 hands.

This sample size was large enough to give AI researchers a clear insight into the proficiency of the software and how AI is working to solve complex situations at the poker table. For the Lengpudashi vs. Team Dragons showdown, the competitors will play just 36,000 hands.

“This is an exhibition, not a match, challenge or competition. We are running a relatively small number of hands, so this is not a scientific experiment like the Brains vs. AI competition in January,” Sandholm told scs.cmu.edu.

Plenty of Potential Winners

Despite being an exhibition match, a $290,000 winner-takes-all prize will be up for grabs. This could be great news for the human players as the smaller sample size may mean that a rush of positive variance could help overcome any skill disadvantages they might face.

Beyond the potentially positive implications for the five members of Team Dragons, the match could prove useful for both AI and poker. If Lengpudashi can win and show that January’s result wasn’t luck, then it will further enhance the credibility of AI. This, in turn, could help further research in a variety of industries.

Closer to home, a win for machine could help to legitimize poker as a skill game in China. As it stands, poker in China operates in a grey area. Gambling is illegal in China, except for select regions, which makes hosting tournaments in the country difficult.

The same is true for online poker. Although players in mainland China can access international sites, the government has been known to randomly blacklist operators.   

Although a showdown between a poker bot and five players is unlikely to instantly change these dynamics, it may help to prove that poker isn’t gambling in the same sense as blackjack and the like. If that happens, then it could still end up being a win for man regardless of the result of how the match plays out.