Why does artificial intelligence like poker?
In many cases, artificial intelligence (AI) developers have a special interest in seeing how their creations can handle playing a poker game. Unlike other strategy games such as chess or Rubik's cube, poker has an important additional psychological component which translates into player intuition. The intuition involved in the success of each decision is the component that AI programmers want to replicate.
We have already talked in other posts about the victories of Libratus, the Carnegie Mellon University AI that won Jason Lee in the Rivers tournament of the Pittsburgh casino bordering on perfection with a win rate of almost 100% in the 120000 games played.
But we shall continue speaking of AI because we live in a technological era and more and more AI projects are being developed, projects such as Cepheus from the University of Alberta in Canada or AlphaGo from Google's DeepMind company are increasingly present in blog news thanks to their achievements.
As far as poker is concerned, AIs plays basically two modes, the "limit hold'em" played by machines like Libratus and the "no limit hold'em" played by the most advanced machines since it is a much more complicated game modality and has more variables. The increase of variables and uncertainty means that the programs can not calculate with such precision the possible outcomes of the game, this is result of the intuitive reasoning behind poker. AI can not always calculate this variable with accuracy, the machine does not possess the intuition of the player that inclines him to make one more or another.
How can a machine be intuitive?
Researchers develop algorithms based on game theory also called incomplete information game so that their creations can takethis information based component of the game into account.
This theory was developed in the 1950s, as part of the mathematical models, a science applied to study interactions of conflict and cooperation in a given situation. This theory based on strategy and probability, was used throughout the cold war to anticipate a possible enemy attack and is based on the rational resolution of conflicts. These mathematical models are used today in fields as diverse as biology, economics, sociology, artificial intelligence and computer science.
The decisions and calculations made by a poker player, are a variants of this type of calculations and encompass all the possibilities that can be faced in a game, through the psychological and rational study of the possibilities of their rivals.
This way, machines are able to adapt to the peculiarities of each item and calculate their probabilities taking into account the uncertainty in their calculations. As machines learn managing a huge volume of information, the training process is of vital importance. This is the case of DeepMind which increases its knowledge by playing online games of all kinds against humans. The machine receives feedback on human behavior to understand how humans process information and make decisions.
Operators and bots
The development of AI alerts operators who strictly prohibit the use of bots in their rooms. All interaction must be done by a human and this kind of software his forbidden. The operators have measures of collecting player´s data: they control the IP, analyze clicks patterns and the game of its players.Everything that is necessary to detect the possible bots which may be manipulating the game.
We can be sure that in the future we will keep hearing about the advances of artificial intelligence but for the moment poker players have no greater enemy than their human rivals. We will have to keep training and improving!