Can A Computer Beat The Best Poker Player?

Can A Computer Beat The Best Poker Player?

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 be argued that AI will become the ultimate poker player when it inevitably surpasses human skill level on a consistent basis. 

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 it has been allowed to receive regarding common strategies.
An artificial example showing how a reachability plot can become used to find clusters with the OPTICS algorithm.

The term “poker face,” which denotes a players’ apathetic appearance to their hand, is an example of this. AI could be coded to consider poker face presentations in a game. As such, its perfect information would be 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 is 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. As such, it can be stated that they will contribute to AI’s rise as the ultimate poker player. 

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.

Can A Computer Beat The Best Poker Player?