All Sections

Google’s AI learns to beat games on its own

The day has come when the conspiracy theorists can finally feel justified in fearing the robot apocalypse! For Google has created a computer which can play games without any instructions!

Okay, so it’s not quite self-aware human-hating robots, but what DeepMind Technologies has developed is quite remarkable. The Google-owned British artificial intelligence company has created a programme which can play, and eventually beat, old Atari games by learning as it goes along and improving its own performance. The programme was designed to demonstrate how AI can work out how to complete complex tasks without being given any guidance from a human.

The company showcased the breakthrough in a simple fashion, by releasing two GIFs, one showing the AI playing Breakout for the first time (and doing pretty miserably at it) and then a few hundred attempts later, playing like a boss. DeepMind claims that the AI has taught itself how to play a further 48 games on the Atari.

The theory behind it all is that if an AI can learn how to play games, then it can also perform trickier tasks like learning how to drive a car. Demis Hassabis, former game developer and brains behind DeepMind said: “If this can drive the car in a racing game, then potentially, with a few real tweaks, it should be able to drive a real car”.

Google acquired DeepMind, a company which was originally funded by Tony Stark wannabe Elon Musk, CEO of Tesla Motors, for an estimated $400 million in the hope that AI could be used to improve a number of Google products, from the fabled driverless car right down to Google Now, with Hassabis quoted as saying “Imagine if you could ask the Google app to complete any kind of complex task: ‘OK, Google, plan me a great backpacking trip through Europe!”