The artificial intelligent robot plays the electronic game, surpasses the human to become the most powerful player

February 25, "Nature" magazine published a Google DeepMind team about its research and development of artificial intelligence robot to play electronic game article. The article said that the intelligent robot in the "Fishing Derby", "Highway", "Robot Tank", "Kung Fu Master" and other 29 games outstanding performance. Among the more than 10 games, the star gunner and the crazy climber, they surpassed the most human players. The system is designed to integrate the deep neural network and enhance the learning operation of two brain-inspired intelligence, so called "extraordinary artificial agents."

Hassabis, a robotic inventor DeepMind, has shown the robot's ability to play games at Harris Casino, and soon Google bought the technology company 650 million dollars in January 2014. April 2014, the robot for the first time in the classic Atari "playing bricks" in the public debut video game, within half an hour it mastered the principle of the game, one hours later after 200 innings, the robot became a gifted veteran: every three or four times to lose a ball! After 300 innings, it actually won every inning, surpassing the human super player!

Of course, the DeepMind team developed the artificial intelligent robot in the video game beyond humans is not the first, i.b.m. "Deep Blue" computer has also defeated the world Chess champion Kasparov. But computer scientist Zachary Mason (Manson) points out that the "deep blue" computer only needs to consider hundreds of steps in a chess-limited game, which can be achieved by the computer's perfect memory and powerful computing power, and that it takes thousands of steps to play Atari games, This makes the DeepMind artificial intelligence system more difficult to deal with. In addition, "Deep Blue" is to advance the chess all the movements and rules of the design, and DeepMind invented the robot can be used in general code to play multiple games.

The robot inventor has begun working with satellite operators and financial institutions to try to make it possible for intelligent robots to eventually "play" their own data sets and learn about weather and oil futures forecasts. In the short term, the invention team plans to design an intelligent robot capable of playing video games that were developed in the 90 's. At present, the robot can not be like humans to play the game is more need to plan or path more complex games, such as "Miss Bean", "private investigator" and "Montezuma Revenge." The inventor put forward a kind of solution idea, namely let artificial intelligence robot make bold decision, dare to take risks.

Computer scientist Manson is not optimistic about the DeepMind team's plans. "According to current research plans, they achieve the goal of StarCraft 5-10 years," Call of Duty may take 20 years, artificial intelligence control unmanned aircraft real-time combat may be 50 years. (Colinlin Compilation)