A machine has learned how to play 49 classic Atari 2600 video games, including Space Invaders and Breakout, according to a UK study.
The artificial agent was only given information about the pixels and the score but managed to achieve more than 75% of the human score on half the games, rivaling the capabilities of many professional game testers.
The study, published in the journal Nature, helps open the way to building artificial intelligence systems good at overcoming challenging tasks from scratch.
Google DeepMind researchers Demis Hassabis, Vlad Mnih, Koray Kavukcuoglu and David Silver developed an artificial agent called a deep Q-network (DQN), which combines reinforcement learning with deep neural networks.
The games at which the DQN excelled were varied, from side-scrolling shooters to boxing and 3D car-racing games.
Beating a human professional chess player was once seen as the gold standard in Artificial Intelligence (AI) research.
This has been done and the AI focus has shifted to harder, more real-world problems.
Bernhard Schölkopf, of the Max Planck Institute for Intelligent Systems, says this research with video games may be a better model of the real world than chess.
NOW WATCH: Briefing videos
Business Insider Emails & Alerts
Site highlights each day to your inbox.