This is a small experiment in Artificial Intelligence.

I made a game of Flappy Bird in Unity and made an evolving neural network that learns how to play the game.

• Each generation has children based on their best scorers.

• Each child can have a random mutation.

• This eventually gives us offspring that are really good at the game.

• The map is generated and changes from run to run, so the neural net is not remembering the map, but using the distance between the next obstacles, the shape of the obstacle and the speed to determine whether to press space/jump or not.