Google's robot plays table tennis like a human (video)

In a recently published article, the Google DeepMind Robotics team introduced a robot capable of competing in table tennis at a human level, as reported by TechCrunch.

During testing, the robot managed to defeat all beginner-level players and won 55% of matches against intermediate players. However, despite these impressive results, the robot lost all matches against professionals. Overall, the system won 45% of the 29 matches played.

The main issues with the system include a delayed reaction to fast balls, as well as difficulties with backhand play, receiving high and low balls, and evaluating the ball's spin. The developers suggest exploring advanced control algorithms and hardware optimizations to overcome latency limitations that hinder the robot's reaction time to fast balls.

The DeepMind team is confident in the potential of their development, viewing it as a small step toward the long-standing goal in robotics — achieving human-level performance in many useful real-world skills. The developers plan to continue working on creating versatile robots capable of performing a wide range of useful tasks while interacting skillfully and safely with humans.