Back in the Paddle

Google's DeepMind has shown off an AI-powered robot that can beat the average player at a game of table tennis.

According to an announcement by the company on X-formerly-Twitter, "it’s the first agent to achieve amateur human level performance in this sport."

A video shows the robot wielding an extended table tennis paddle, confidently aiming the ball to throw off its human opponent.

But when it'll be able to take on the likes of China's national team, which recently won gold at the ongoing Olympics in Paris, remains to be seen. While Google's engineers can seemingly keep up with the robot, a professional in the sport will be far harder to beat.

Amateur Hour

Since table tennis "requires years of training for humans to master," the sport is a "valuable benchmark for advancing robotic capabilities, including high-speed motion, real-time precise and strategic decision making, system design and enabling direct competition with a human opponent," according to DeepMind's official documentation.

Robots have already taken on a number of different human sports, like one developed by students at the University of Electronic Science and Technology of China that can challenge players in badminton.

DeepMind's invention isn't even the only robot that can play table tennis, either. A company called OMRON Robotics released the seventh generation of its table tennis robot, dubbed FORPHEUS — which the company claims can teach humans how to play the game — back in 2022.

Scientists at DeepMind used AI algorithms to "adapt to various opponents by tracking their behaviors and playing style — such as which side of the table they tend to return the ball to."

Based on 29 matches with human opponents of different skill levels, the robot performed as well as what DeepMind is calling an "intermediate amateur."

But when it comes to a more advanced player, the company admitted that the robot has plenty of limitations, including "reaction speed, camera sensing capabilities, spin handling, and the paddle rubber."

"Whilst the robot lost all matches vs. the most advanced players it won 100 percent matches vs. beginners and 55 percent matches vs. intermediate players," DeepMind noted in its documentation.

For a casual bout of table tennis among friends, that's a reasonable success rate. But DeepMind has some ways to go until its robot can take on the pros.

More on robots: After Just Ten Months, Amazon Says Its "Security Guard" Robot Will Become a Useless Brick


Share This Article