DeepMind beat a world champion of Go, so now it wants to try its hand at a much more popular game. Sorta.
Google says it’s used DeepMind to teach an ant to play football — that is, if you consider the beige four-legged ball an “ant” and the act of manipulating a black-and-white ball into a red space “football”.
The description may be somewhat exaggerated, but the development is cool: the project shows DeepMind’s reinforcement learning abilities. Through a combination of trial-and-error and deep learning, the AI quickly learns the mechanics of kicking the ball into the goal.
Engadget points out that this technology could be used for manipulating animal-like robots and help them manoeuvre in the world with much less training. Just imagine what this kind of advance could have done for the robot that ran away. [Google, via Engadget]