It is an advance that until a few years ago seemed like fiction.
And a video from the Hong Kong University of Science and Technology, where researchers transformed the Unitree G1 into an almost natural basketball player, if I can describe it that way. It is an advance that until a few years ago seemed like fiction. The small 1.32 m athlete was already known for kung fu moves and withstanding flying kicks from adult humans.
But the scene shared by doctoral student Yinhuai Wang marks new territory there. The G1 drives the ball, changes direction, fakes and makes shots while a human tries to block it. Yinhuai Wang described the demonstration as the first real-world basketball block made by a humanoid and even joked that he had become the first person to block one of these robots.
Unitree stands out again.
The skill did not arise from isolated tricks, but from a system called SkillMini
, the approach feeds robots with human-ball co-movement data, allowing it to learn complete patterns from dribble to spin, fine control to shot. The same control policy absorbs multiple skills and creates smooth transitions between them, even when the data set does not contain exactly those steps.
The more varied the repertoire of movements analyzed, the broader the G1's motor vocabulary becomes, the result appears on the court, a robot that responds to pressure, adjusts the body, protects the ball and executes a shot without hesitation. The demonstration captures a moment of change in humanoid robotics, revealing how much the sector has evolved in a field where Unitree has become a protagonist amid growing competition.
The Internet reacted with a mix of fascination and irony.
There were those who praised the exception over the fundamentals, who joked that robots will do everything except load the washing machine and who dreamed big of entire robotic sports leagues, the video, however, leaves a deeper sensation, when a humanoid begins to master the subtleties of a game that depends on rhythm, time and instinct. Something changes in the collective imagination about what machines can learn.
