Google's latest acquisition has demonstrated its artificial intelligence's ability to learn to play video games.

The AI algorithm, developed by London-based, Google-owned company DeepMind, uses reinforcement learning, selective memory, and other elements of behavioural science to analyse the pixels and inputs of a game and work out how to play it.

Terrifyingly, the software was able to effectively master Breakout after four hours of play, going from moving its paddle randomly to developing advanced strategies at high speeds.

It outperformed a professional game tester by over 1300 percent.

Other games from the 49 Atari 2600 titles learned include Pong, Ms. Pac-Man, Space Invaders, and Montezuma's Revenge.

The next step for the DeepMind team is to apply it to more complex games like Quake and StarCraft, and hopefully not thermonuclear war.