Watch Google DeepMind’s Ping Pong Robot Take on Humans

Ping pong seems to be the sport of choice when it comes to tech companies showing off their robotic wares. Japanese company Omron, for example, made headlines a few years ago with its ping pong robot, which could comfortably hold a match with a human player while showcasing the company’s sensor and control technologies.

And with artificial intelligence (AI) is advancing by leaps and bounds, and we are now starting to see more advanced robot ping pong players that could soon give even the best human players a run for their money.

Let’s take as an example the impressive effort of the company’s engineers Google DeepMindIn a new paper titled “Achieving a competitive human-level table tennis robot,” the team said they have created a “true human-level amateur player” that combines artificial intelligence and an industrial robotic arm with a bat attached.

The video (above) shows the AI ​​robot making quick decisions and executing a variety of backhand and forehand shots. Notably, it is also capable of hitting backhands with light spin, demonstrating the ability to read the spin of the ball and adjust the way it strikes the ball accordingly. It is also able to handle balls coming at it at high and low speeds from all parts of the table, as well as balls coming at it from significant heights after bouncing hard off the table. It is truly impressive.

“Achieving human-level speed and performance on real-world tasks is a guiding light for the robotics research community,” the researchers say in the paper. “This work takes a step toward that goal by presenting the first trained robotic agent that achieves amateur-level human-level performance in competitive table tennis.”

During a series of test matches, the robot won 100% of the games against beginners and 55% of the games against intermediate players. However, it is clear that it has room for improvement, as it lost all of its matches against advanced players. Overall, the robot won 45% of the 29 games it played.