Sony's Robot Ace Beats Elite Table Tennis Players

Ace uses a robotic arm with eight joints and multiple cameras placed around the table to follow the ball's position and spin. The camera system can detect how the ball is spinning within milliseconds by focusing on the ball's logo. Ace learned its skills through about 3,000 hours of practice in computer simulations. Sony's president of AI, Michael Spranger, said the company wanted the robot to win through better tactics and decision-making rather than simply hitting the ball faster than any human could. The robot's speed and reach were kept similar to those of a skilled human player.

The development builds on decades of using games to test AI, from chess to video games, but moving a machine into real-world physical competition is a major step forward. Spranger called the past year a "ChatGPT moment for robotics." However, some experts have raised concerns. John Billingsley, who began researching robot table tennis in 1983, argued that Sony's multi-camera system gives Ace an unfair advantage over human players, who have only two eyes. Players also found it unsettling to compete against a machine that shows no emotion and feels no pressure.