A robotic hand with "human-like" fingers was able to solve the Rubik´s cube in about 3 minutes.
The device, led by artificial intelligence, is the first "robot" to manage this achievement without specially designed, where he trained himself.
The robot´s hand was designed in a way that could be used to manage other things, where I learned to solve the Rubik´s cube through a trial and error technique, known as "learning reinforcement."
OpenAI has taught artificial intelligence to control automatic hand, developed by Shadow Robot.
"It starts from not knowing anything about how to move the hand or how the cube will react if you press the sides or the faces," said researcher Peter Wilinder.
AI then scores on a successful move, such as turning the cube or rotating its face.
The researchers said it was difficult to control artificial intelligence, because of the number of simultaneous contact points between the hand and body, which often enter the solution of the "Rubik" cube.
However, artificial intelligence was able to learn to correct errors, such as accidentally rotating the cube too much.
Just like a human, the time it takes a robot to solve a Rubik´s cube depends on how difficult it is in the first place.
Researchers hope the robot will be further trained to do many common tasks.
