Machine learning in video gamesAugust, 11th 2017
Artificial intelligence is traditionally quite poor in video games, developers content themselves by imagining as many as possible situations and adequate answers...
And if they are good enough then they will arrive at a satisfying result, neither too simple nor too insurmountable, which will make the player's experience pleasant and enjoyable.
With the emergence of machine learning techniques, it is possible to offer much more interesting AI. For example, reinforcement learning, an area of machine learning, has the principle of leaving the AI autonomous (and random) in order to reward positive actions and reprimand actions without interest. The AI educated that way fulfills its objectives with various possibilities without solutions previously defined . The developers then have just to define the possibilities of actions and the objectives to be fulfilled, the rest is up to the machine to learn it, in particular by training.
We can find an example of learning by reinforcement with Charles an algorithm allowing to make autonomous driving on GTA V, check it out!