YOUR top score is about to be trounced. Google has developed artificial intelligence software capable of learning to play video games just by watching them.
Google DeepMind, a London-based subsidiary, has trained an AI gamer to play 49 different video games, beating a professional human player鈥檚 top score in 23 of them.
The software isn鈥檛 told the rules of the game 鈥� instead it uses an algorithm called a deep neural network to examine the state of the game and figure out which actions produce the highest total score.
Advertisement
鈥淚t really is the first algorithm that can match human performance across a wide range of challenging tasks,鈥� says DeepMind co-founder Demis Hassabis.
The software performed best on simple pinball and boxing games, but also scored highly on the arcade classic Breakout, which involves bouncing a ball to clear rows of blocks. It even managed to learn to tunnel through one column of bricks and bounce the ball off the back wall, a trick that seasoned Breakout players use (Nature, ). 鈥淭hat was a big surprise,鈥� says Hassabis.
聯It scored highly on arcade classic Breakout and even managed to learn a trick seasoned players use聰
Google won鈥檛 say why it鈥檚 interested in the software, but it could be used to analyse large datasets and to improve the ads the company serves up.
This article appeared in print under the headline 鈥淐hampion gamer鈥�