The British company DeepMind, bought a year ago by the US giant Google internet, was based on an algorithm and has combined various computer learning techniques with mechanisms inspired by the functioning of neuronal networks.
It was how you got your ‘artificial intelligence agent “, dubbed DQN, learn to play 49 classic video games in Atari 2600 console, from the experience of memory.
An article, published today in the journal Nature says that the created system, which operates at a level comparable to a professional human in-game, is able to discover the object of the game and master it in addition to more information than the images that appear on the computer screen and the score.
The authors of the computer system point out that the application of artificial intelligence to video games is no more than a power demonstration of their algorithm, whose use can be generalized to other types of media or industries.
The co-founder of DeepMind, Demis Hassabis, explained at a press conference in London that the artificial neural network created by your team is of a different nature from other machines, such as Deep Blue, a computer that defeated in 1996, the world champion Russian chess Garry Kasparov.
No comments:
Post a Comment