When people hear about AI and robots, they often imagine apocalyptic scenarios with lasers and glowing red eyes. However, Google DeepMind has taken a different approach by advancing robotic AI technology to excel at table tennis, rather than pursuing global domination.
Google DeepMind recently shared footage on X, showcasing a robot that has been trained to outperform the average person at table tennis. As someone with some background in programming robotic arms, even if limited to manufacturing engineering, I found the demonstration quite impressive. The accompanying research paper is definitely worth a read.
The system involves more than just tracking the ball with a camera and randomly swinging a paddle. Google DeepMind began by creating a database of all possible initial states of a table tennis ball, including its position, speed, and spin. The robot arm then practiced various movements, learning to switch between forehand and backhand grips, apply topspin, and more.
Afterward, the AI system was tested against real human players. The system was designed to observe how different individuals played and adapt its strategy accordingly in real time. Google DeepMind reported that the robot played against 29 opponents, who were categorized into four skill levels. By the end of the matches, the robot performed at about the level of an “intermediate amateur.”
Of course, the robot was thoroughly outmatched by more skilled players, as seen in one clip where a person easily defeated the AI with a quick shot. Google DeepMind acknowledges that factors like the effect of paddle rubber on spin are challenging to accurately simulate.
Despite this, the achievement is still remarkable, and it opens up numerous possibilities for future applications of the technology. In mass-manufacturing settings, robots often struggle with minor misalignments or changes in lighting. An AI system that can adapt to these variables could prevent production line delays.
Looking ahead, I can envision a future where AI-trained robotic arms assist seriously injured individuals during their recovery, helping them regain strength and mobility.
For now, however, mastering table tennis is the focus—and I’m fully supportive of this objective. After all, it’s far better than developing AI for catastrophic purposes.