Computers have dominated games like Chess and Go being able to outplay even the highest caliber of players. But that's because Chess and Go can become predictable, that is, the AI can compute all possible scenarios from the first move to the last because all information is available to them.
However, in games where there is incomplete information like poker, it will be a lot more difficult to come up with an algorithm that would enable the AI to play with elites. Until researchers from Carnegie Mellon University, Tuomas Sandholm and Noam Brown, developed an AI that beat six elite poker players.
For AI researchers, poker presents a better model of the real world. Rarely in life do situations involve just one winner and one loser, or scenarios in which information is fully available. By improving an AI’s ability to deal with hidden information in multi-participant scenarios, computer scientists are dramatically expanding the domains in which AI can be used.
To create a system capable of proficiently playing six-player no-limit Texas hold’em poker, Brown and Sandholm employed a grab bag of strategies, including new algorithms the duo developed themselves. Before the competition started, Pluribus developed its own “blueprint” strategy, which it did by playing poker with itself for eight straight days.
Apart from the blueprint strategy, the AI also adapted its gameplay to be more balanced and unpredictable. One surprising thing about Pluribus' strategy was that it effectively employed strategies that even elites tend to shy away from or consider weak strategies.
(Image credit: Michal Parzuchowski/Unsplash)