AlphaGo made DeepMind famous. But before the London-based company built the neural network that beat the best living player of the oldest continuously played game in history, it had to master Atari. Games like Breakout are pretty simple for humans to figure out: Move the paddle, bounce the ball, break the bricks. But to a computer, all those shapes and colors are gibberish. DeepMind tackled the problem using an approach called deep reinforcement learning.