The beauty of rock-paper-scissors is that it equalizes the odds of success among the players, like a coin toss, but still provides the illusion that there’s some agency involved. (Your rock-paper-scissors strategy is the best strategy, of course.) But what if someone rigged the system and cheated, by somehow knowing an opponent’s every hand draw?
A robotic hand built in a University of Tokyo lab does just this. It has demonstrated 100 percent accuracy at beating a human opponent in rock-paper-scissors (which in Japan is called janken). High speed cameras allow the robotic hand to recognize whether its opponent is forming a rock, paper or scissor hand shape before that shape is completely formed, and quickly compensate by forming the superior gesture.
To do this, the robot takes advantage of humans comparatively slow visual processing time. It takes a person about 60 milliseconds to change his hand position, and humans can follow visual events on the order of 30 to 60 milliseconds. The robot, however, squeezes in just below that cut off, recognizing the human opponents gesture and flashing its winning motion in about 20 milliseconds.
More from Smithsonian.com: