Before going further, you’ll need to understand the concept of game theory. Game theory is basically a branch of applied mathematics. In-game theories (How Game Theory Strategy Improves Decision Making), there are different available tools with the help of which different situations are analyzed. There are parties in-game theories mostly referred to as players and the decision they have taken are interdependent. This is a kind of playing chess in which the turn of one player is associated with the future strategy of the opponent player.
So, a complete strategy, decision, possible movement, and formulation of strategy basically come under the banner of game theory. Game theory is basically a framework with the help of which one can understand the strategy of the opponent player. With the help of optimal game theory, the players come to a point where optimal decision-making is reached and the strategy of the player becomes strong enough so that it can play in any situation with any kind of player. This concept is very important from a technical point of view for those researchers who are willing to work in the field of machine learning, deep learning, AI, or NLP. In this article, the association of game theory with AI and NLP will be discussed.
Game Theory in AI
The main aim behind the game theory is to understand the outcome of the optimal solution of the opponent player. There is another field which is an inverse game theory with the help of which the strategy of the opponent player and strategy can be explained.
With the help of game theory, the AI agents can design with the help of which the opponent environment can be created in a game (Game Theory in Artificial Intelligence | by Pier Paolo Ippolito | Towards Data Science, 2020).
This can be explained by an example like when you are playing a game against the computer there are different AI agents who are pre-defined inside the game are playing against us.
The caliber of the AI agent depends on the difficulty level selected initially and with the passage of time and with the training data set this AI agent gets better and the difficulty level is enhanced. The use of game theory in AI is a very vast field. Game theory is mostly used in the training of GANs, multi-agent system, and deep reinforcement learning.
In artificial intelligence, there are different types of categories in which different games are placed. These categories are perfect information games, imperfect information games, deterministic games, and non-deterministic games or chance moves games (Artificial Intelligence | Adversarial Search - Javatpoint, 2019).
A brief explanation of these categories is as follows:
-
Perfect Information games, in board games such games are those in which the AI agent is responsible to have all the information of complete board and the agent can see all the moves of the opponent player and make decisions accordingly. Example of such games is chess, checkers, go, etc.
-
Imperfect Information Games, are those games in which the agent does not have the necessary information and is not aware of what going on next. Example of such games is card games, bridge, tic-tac-toe, etc.
-
Deterministic games are those games in which there is a strict pattern to be followed and a proper set of rules for the games. The rules are properly defined and there is no random algorithm associated with them. Examples are chess, Go, and tic tac toe.
-
Non-deterministic Games: In such games, there is no proper set of rules in the games. The events that can be occurred are mostly unpredictable. Such games mostly include the chance and luck of the user as well. Such games required very high-quality AI agents and their training have a large data set so that a competitive game environment can be obtained. Examples are a monopoly, checkers, etc.
Figure 1. Game Theory and Prediction explained (The Basics Of Game Theory, 2020)
Game Theory and NLP
In in-game theory, another approach is neuro-linguistic programming comes into play. It is basically a psychological approach with the help of which different strategies are being analyzed and apply accordingly depending on the nature of the opponent player. In NLP, there are basically thoughts, languages, and behavior of the opponent.
All of these strategies are combined with each other in order to get an optimal and specific outcome of each step taken by the opponent player. (Craft, 2001). Neurolinguistics is basically how our brain works but there is no significant research carried out on whether such technique really works or not. It is basically a psychological technique used by magicians as well.
The pro players of card games like bridge, rung, etc. also used such techniques to find the actual situation of the opponent player but such technique is yet to be implemented in computer or mobile phone games.
Figure 2. A brief description of NLP (The Universal Self-Help Keys of NLP - Neuro-Linguistic Programming | The Whole Parent, 2020)
From the figure shown above a complete understanding of NLP is stated.
As it is basically the combination of three techniques one is neuro, the other is programming, and in the last linguistic. In programming, there are the thoughts of a person which are stated in words.
The ideas and all the organization of the thoughts of the individual who is doing programming for the AI agent so that it can be trained accordingly with situations and environment that can be happened later on in the game.
The second thing is neuro, in which the thinking process of the user can be compiled. In this portion, all the senses of the person are combined and compiled, and later on, these thoughts are embedded in the programming section.
This portion is yet to accomplish in the simulation environment as it required a large force and large computational power. The last part is linguistic which is basically the psychological part. In simple words, the linguistic portion contains all the behaviors, languages, eye contact, and such kind of things to get possible outcomes from human behavior.