Neural networks and machine learning in games: how and why they are used

In 1950, computer scientist Alan Turing published a test to see if a machine could think. The wording of the test is as follows: “A person interacts with a computer and a person. By answering questions, he must determine with whom he is talking - with a person or a program. The purpose of the program is to mislead a person into making the wrong choice.”


Games are one area where it is easy for a machine to hide a person behind an interaction interface. Neural network based systems can play offline and online games. Technology is at such a level that the machine passes the Turing test, and a person cannot distinguish the machine from a live player. In this case, a program that imitates partners in multiplayer games is called a bot. We have examples of creating bots for computer card games.


How it works
Action
In offline games, the action is performed by a robot. Online - a character or a virtual player.

Perception
The system recognizes objects using machine vision. Cards in hand and cards on the table.

"Thinking"
Using mathematical algorithms, probability theory, machine learning mechanisms, artificial neural networks, or a combination of these methods, the machine decides what to do in the current situation, chooses which move will be optimal.

The mechanism is divided into three components - perception, "thinking" and action.

How it is used in gaming
The system, using machine vision, determines which cards are on the table and, based on the situation, builds a game algorithm. Sometimes systems based on neural networks can not only play, but also chat, completely imitating human behavior.