AI Robot Ace Defeats Elite Table Tennis Players in Groundbreaking Study

Robots have already claimed dominance in tennis and marathon running. Now, the sport of table tennis faces a similar challenge. Researchers have developed an AI-powered robot arm, named Ace, capable of repeatedly defeating elite and professional players under official competition rules.

A study detailing this achievement was published this week in the journal Nature. Sony, the creator of Ace, claims it is the first robot to reach expert-level performance in any competitive physical sport, following decades of development in table tennis robotics.

A promotional video from Sony’s AI division showcases Ace’s lightning-fast reflexes, as it bounds back and forth to counter aggressive shots. Human players in the video do not hold back, delivering powerful strikes with full force. The display underscores a major milestone in AI and robotics, reflecting the rapid advancements in humanoid robotics over recent years.

While teaching a robot to play ping pong may seem trivial, the techniques behind Ace’s success could have far-reaching implications. Peter Dürr, lead author and project lead at Sony AI, told Reuters:

"The success of Ace, with its perception system and learning-based control algorithm, suggests that similar techniques could be applied to other areas requiring fast, real-time control and human interaction, such as manufacturing and service robotics, as well as applications across sports, entertainment, and safety-critical physical domains."

According to the study, Ace won three out of five games against elite players with over a decade of experience. However, it initially lost two matches to top-level professionals as of April 2025. Sony later reported that Ace went on to defeat more professional players in December 2024 and January 2025.

The Complexity Behind Ace’s Performance

The robot’s capabilities are staggering. It relies on an intricate system to track the ball and predict its trajectory in real time. Nine cameras and three vision systems enable Ace to:

• Track a ball at 200 Hz with millimeter accuracy.
• Measure spin at up to 700 Hz.
• Operate with around ten milliseconds of latency.

This precision is fast enough to capture motion imperceptible to the human eye. Deep reinforcement learning allows Ace to seamlessly predict ball behavior and counter its opponent’s moves.

Human Players React to the Challenge

Many high-level players have expressed difficulty adapting to Ace’s playing style. Professional table tennis player Mayuka Taira told Reuters:

"It’s essentially impossible to sense what kind of shots it dislikes or struggles with, and that makes it even more difficult to play against."

Despite Ace’s prowess, Dürr acknowledged that professional players retain an advantage—for now. He noted that human players excel at adapting to opponents and identifying weaknesses, a skill that remains beyond current AI capabilities.