Artificial Intelligence beats top pros at six max

It's time to welcome our robot overlords as once again an AI has beaten some of the best poker players in the world quite convincingly.

In 2017 Carnegie Mellon University Artificial Intelligence programme 'Libratus' famously beat a number of top poker players at heads-up, prompting a wave of mainstream media attention. 

It's happened again, as the University, in partnership with Facebook, has now developed an AI called 'Pluribus' which has beaten five top players at six max No Limit Hold'em. Facebook research scientist Noam Brown created the AI in partnership with Tuomas Sandholm from the previous Libratus project. 

Among the top players Pluribus beat were Linus Loeliger, Darren Elias and Chris Ferguson. Ferguson will always be a controversial choice given the Black Friday Full Tilt scandal but he has been associated with Carnegie Mellon's AI research for more then a decade. 

In 2017 Libratus caught the AI world's attention because by beating the best at heads-up poker, it displayed 'general intelligence', a much more reflexive form of learning compared to that needed to beat a game like chess. General intelligence is needed because there are so many unknowns in poker. By beating top players at six max, this is considered another big step for AI, as with five opponents there are so many more variables than heads-up. 

Pluribus was beating the five humans at a rate of $5 per hand at $50/$100 stakes over 10,000 hands. Perhaps most alarmingly it only required $150 worth of cloud computing resources to perform at this level. A Facebook spokesman said they would not be outsourcing the software. 

If you are interested in this sort of thing the full research paper is worth a read. 

Does this concern you for poker and life? Let us know in the comments: