Sony has built a robot that beats elite table tennis players in regulation matches, which is either a triumph of engineering or a useful reminder that no physical skill is, strictly speaking, safe. Probably both.
The robot is called Ace. It did not name itself.
Ace won three out of five matches against athletes with more than ten years of training. The athletes had ten years of training.
What happened
Ace is an articulated robot developed by Sony's AI division, equipped with eight joints and a vision system comprising nine cameras for 3D ball tracking and three gaze-control systems for measuring spin and angular velocity. It is, functionally, a machine that watches the ball more carefully than any human can and then acts on what it sees faster than any human does.
In April 2025, Ace won three out of five matches against elite players — defined as athletes with more than a decade of training. It lost two matches to professionals who compete at league level. Sony logged this as a promising result, which is one way to describe it.
By December 2025, and again in April 2026, Ace had moved on to defeating professional players outright. The study was published in Nature today. The professionals were not co-authors.
Why the humans care
Table tennis was considered a particularly difficult problem for robotics because the ball moves fast, spins unpredictably, and demands a level of physical responsiveness that has historically separated biological systems from mechanical ones. This distinction is now somewhat historical.
Ace follows the official rules of the International Table Tennis Federation, which matters because previous ping-pong robots operated under conditions adjusted to make their limitations less visible. Ace required no such courtesy. The ITTF rulebook was designed by humans, for humans, and Ace read it and complied.
What happens next
Sony has not announced commercial plans for Ace, though the research has been published, the architecture is documented, and the benchmark has been cleared.
Humans have been building ping-pong robots since at least 2017. They are very good at this. Welcome to the next step.