It wasn’t very long ago that Artificial Intelligence played next to no role in online poker. But more and more frequently, AI shapes the game – from strategy development and fraud detection, to delivering players an increasingly immersive, personal experience.
The development of poker bots such as Libratus, which defeated human players in the no-limit Texas hold’em tournament in 2017 – and Pluribus, which in 2019 beat six-player no-limit hold’em against elite human professionals – is strongly indicative.
AI is transforming decision-making
If you play online poker, chances are that either AI has already been integrated into the experience or it will be soon. Poker is a great example of how much the play and the experience of a game can be changed by technology such as AI. Using AI technology to create gaming environments specific to a particular player puts the player’s user experience on a steriod-fueled path toward personalisation that humanly devised games can only dream of. In addition, the games get better at developing a human player’s decision-making skills.
Indeed, without intervention by AI, online poker would have never been able to solve the prevalent problem of cheating before poker sites existed. Clearly, poker players have finally learnt their lesson about the importance of integrity in gaming, and the role of regulation in ensuring its consistency.
Obviously, with advancements in AI, each of these possibilities may occur in numerous different ways of training players and poker over time. We could see AI introduced in training platforms to provide more complex testing environments for players to hone their analytical and strategic prowess. Within poker itself, players could become better analysists of opponents’ quantitative strategies, acting quicker and more decisively under the gun.
AI opponents are challenging human players
While AI poker bots have upped the ante and disrupted conventional playing techniques, they are not yet ready to completely replace intuition and skill. It would require vastly more processing power at present to build an AI that can win regularly in round-robin style go/no-go high-stakes tournaments, never mind multiple-hand home games enjoyed by hobby cash-game players, except for those who turn to poker as a full-time profession, and even there they have only recently achieved that goal.
That is, while poker is easily solvable using a game tree in the abstract, full tables with changing stack sizes and variable player behaviour preclude AI programs from predicting who will fold and who will call every time and on every street of a hand before it is dealt. Even with these constraints, different players play differently, and so AI needs to account for this.
To address these issues, poker AI programs often include distinct player personalities for users to choose from. They can be tight or loose; aggressive or passive. And the strategy for any one Avatar will change on the fly, based on what real humans are currently doing in the game for a tougher, more realistic poker experience.
AI is enhancing the user experience
As play continues for longer periods, AI seems to be having an effect on poker players’ playing experiences. Some apps can adjust the game for each individual player according to that player’s own level of play and taste in the game, thus stimulating interest and motivation to keep playing and getting better at the game. Poker players are similarly kept interested and motivated by the constant challenge and potential for improvement.
Programmed by Carnegie Mellon University and Facebook AI to play six-player no-limit Texas Hold’em – a poker subset with lots of imperfect information and bluffing and variable player behaviour – Libratus outperformed four players, each considered one of the world’s best, between January and February 2017. Courtesy of facebook.com/AI.This success demonstrates not only that AI can create a good solution for such imperfect-information games with variable player behaviour, but also that Libratus is a good solution for such games.
This is not ‘the end of human players’; it seems most likely that AI will improve our understanding of the game, and provide us with training tools to play it better by refining our own strategies.
AI raises ethical concerns
Because poker ultimately depends on interaction with other humans, it provides an ideal ‘sandbox’ to test artificial intelligence (AI). While first-generation programs like Chinook and Polaris helped define the field – both in terms of demonstrating the AI’s ability and its limitations – most of the aforesaid industry problems remain unsolved, and it seems that players of all skill levels deserve a fun game.
Online poker players are acutely concerned with AI cheating. In light of this, sites often have very prohibitive measures against the use of unauthorised AI, not only to preserve the integrity of the game, but also so that players know they are competing fairly.
However, industry members will also face the issue that it may take creating a newcomer-friendly poker format for the game to truly reach a larger demographic. It will take a lot of work to make sure that all the relevant parties and regulators – human and otherwise – are protected.
