Is there an AI for poker?
Artificial Intelligence & Machine Learning
Why Will Machine Learning Become the Ultimate Poker Player?
As the world of artificial intelligence (AI) develops, it will reach insurmountable heights. News about computers beating top professional game players has increased in recent years, from checkers to chess to poker (Brown and Sandholm 418). As poker can be more complex than other card games, it is interesting to delve into the AI characteristics that have allowed it to surpass human ability to play. It can therefore become argued that AI will become the ultimate poker player. When it inevitably surpasses human skill level on a consistent basis.
Poker is a card game that has a long and rich history. The origins of poker can become traced back to the 10th century in China. Where a game called the “domino cards” became popular. This game migrated from Persia and subsequently to the Islamic world, where it found the name: “As Nas.”
In the early 19th century, poker began to take on the form that is recognizable today. It was introduced to the United States by French and Persian sailors, who played a version of the game called “poque” on Mississippi riverboats. The game quickly gained popularity, and by the mid-19th century, it had spread to the entire country.
During the American Civil War, poker was a popular pastime among soldiers on both sides of the conflict. The game was also played in saloons and gambling houses. And it soon became associated with the Wild West.
In the late 19th and early 20th century, poker became a more organized and formalized game. The first poker tournaments became held, and the first poker books became written. The World Series of Poker, which is now the most prestigious poker event in the world, was first held in 1970.
In the 1970s and 1980s, poker experienced a resurgence in popularity, thanks in part to the spread of casino gambling and the rise of televised poker tournaments. The game has continued to grow in popularity in the 21st century. Moreover, thanks to the advent of online poker and the popularization of Texas Hold’em. Now considered the most popular version of the game.
There are a few components that comprise AI’s status as the ultimate poker player.
Firstly is AI’s ability to learn information. Moreover, AI that are coded to adapt to their circumstances are more flexible to learning information. Brown and Sandholm note that Libratus – the AI that beat top poker players – computes its moves by using game-theoretic strategies in abstraction, subgame-solving during the game, and self-improvement on that strategy as the game goes on (418-419). AIs that have a similar player strategy then are enabled to distinctly identify poker moves and what routes they should take to succeed in alternative outcomes.
Secondly, AI’s ability to calculate outcomes at any time during a poker game and act based on those hypothetical situations, indicates it has an inherent ability to “remember” the information about a certain game. As AI is not typically coded with “forgetfulness,” its perfect ability to “remember” also means it has perfect information about a current game and prior games as well. AI is therefore able to capture a wholehearted perspective of the situation at hand and integrate it into its outcome predictions.
AI’s ability to “remember” also extends to additional information its allowed to receive regarding common strategies.
The term “poker face,” which denotes a players’ apathetic appearance to their hand, is an example of this. AI could become coded to consider poker face presentations in a game. As such, its perfect information would become supplemented. So that the AI would be able to compute emotional information like displays of emotion, or lack thereof, from other players.
When considering the impacts of AI surpassing human gaming ability, there are some concerns regarding AI’s abilities. If AI surpassing human gaming ability becomes a more common occurrence, it is worrying that humans are able to code something more strategic than themselves. It highlights that there is incongruence between strategy and emotions – as that is the state of being human. Exploitation of this potential weakness then opens an avenue of AI coding exploration that could have far-reaching implications outside of just playing a game of poker.
AI’s coding makes it flexible based on what it becomes coded to do. Adaptation, remembrance, and identifying all aspects of a situation are what culminate to allow it to surpass human ability. These core features of AI are integral to understanding its role in gaming settings.
Written by Yueyuan Xu
Works Cited
Brown, Noam, and Tuomas Sandholm. “Superhuman AI for Heads-up No-limit Poker: Libratus Beats Top Professionals.” Science, vol. 10, no. 1126, 2018, pp. 418-424.
