Artificial Intelligence in Poker?

In the past, Poker has been a regarded as a game of psychology and reading opponents. When I first started learning how to play, this was how I was taught. I was taught to look at other players’ habits: to notice what types of hands they would play, when they liked to check, when they liked to bet, and when they liked to bluff. And although this is still a huge part of Poker today, the game is advancing. As some may know, Poker is also a game of mathematics and odds where players today are able to know their expected value and their probability of winning with certain hands. In addition to this, some players have also decided to adopt the GTO (Game Theory Optimal) playstyle which is a more mathematical and balanced approach to the game; players like Daniel Negreanu have begun to incorporate this strategy into their play style already. So as I thought about how much more strategic and mathematical Poker has become, I though about AI and how effective it would be in Poker. Today, I will be talking about the applications of AI in poker and an example of a Poker bot that has already been developed.

Upon further research on this topic, I found that in 2019 researchers at Carnegie Mellon, along with Facebook AI, created a poker bot that was able to beat six professional players. According to the article on Carnegie Mellon’s news website, the AI, known as Pluribus, was able to implement some very interesting strategies while it played. In some video examples shown, Pluribus was able to set up traps and call bluffs while mixing up its strategy, making it unpredictable – similar to GTO. In addition, the bot was trained through self-play meaning that it played millions of hands against copies of itself which allows it continuously refine its strategy and learn from its mistakes.

I find the applications of AI in poker to be extremely interesting as there are even more ways that it can be used in the game. In addition to what has been shown from Pluribus, Poker strategy also includes things like pattern recognition, real-time adaptations, and even consistency which I believe AI will be able to do better than humans. One interesting point with AI and Poker, is that sometimes after losing a lot of money and hands, players can start to go on tilt, playing more emotionally and therefore more irrational. With AI, the bots play consistent as they have no emotion and will continuously make its decisions based on logic and math instead of emotion.

As AI continues to develop, I’m very interested to see the development of its applications in Poker – a game I’m loving more and more as I play. Although there could be potential ethical implications of AI in Poker as well – with players having the ability to maybe use AI to gain an advantage in online tournaments – I think it also presents an opportunity to understand the game and even elevate player’s strategy as well.