Modern computer games often use artificial intelligence (AI) algorithms to drive certain elements of the gameplay. This is particularly relevant to those genres where decision making is being repeatedly done many times over a single session, for example, when playing strategic games.

The author of this video used a StarCraft II API client for Python 3. This library contains functions specifically created for programmers developing StarCraft II bots.

It is difficult to play against a capable machine learning algorithm. But what would happen if we made another algorithm to compete against the original one that was made by the game developers?

Bots – or automated player agents – are nothing new, and they existed since the dawn of video games. But in the past they were being made using a hard-coded logic. What is different when their functionality is augmented using AI?

The following video shows one experiment where a machine learning bot is built with aim to beat the StarCraft 2. This is, of course, not an easy task, but a very interesting one, and material presented here could be useful for those who wish to learn the principles of Artificial Intelligence in a highly visual and fun way: