China’s Robot Games: Humanoids Outpace Usain Bolt, Yet Their Tweezer Skills Leave Me Even More Impressed.

This year’s new events might be too straightforward for human athletes, but they are crafted to challenge the cognitive abilities of robots as much as their physical strength. These events involved tasks like connecting cables to sockets and carefully placing trash in a bin within a simulated bedroom.

Tellex notes that training robots to interact with objects in complex settings is significantly more challenging than getting them to run or jump. Advanced manipulation is akin to a holy grail for robotics, with several US startups striving to achieve this through innovative AI technologies.

Fine motor skills present a major challenge for robots because they must contend with unpredictable factors such as friction and slippage. Humans possess a finely tuned sense of touch, an inherent grasp of physics, and exceptional adaptability at high speeds. Robots, however, lag in these areas.

“This illustrates Moravec’s paradox,” Tellex states. “Activities that appear spectacular, like performing backflips, are actually less complicated than tasks like connecting a cable.”

A significant caveat regarding the events is that some images and videos indicate that humans were remotely controlling some robots, implying that the bots may not be fully autonomous. Nevertheless, the ability of robot hands to pick up tiny objects using tools—even under human supervision—remains quite impressive.

A humanoid robot takes part in the Hotel scenario cleaning service event on day two of the 2025 World Humanoid Robot Games at the National Speed Skating Oval.

Video: Lintao Zhang/Getty Images

The full schedule for the games highlighted additional events aimed at testing manual dexterity, such as “block building” and “tweezer bean picking.” These challenges seem to reflect China’s ambition to automate tasks in manufacturing and warehouses, like wiring in automotive frames or assembling electronic components.

Galbot participated in several contests, including a Supermarket Scenario Competition, which entails stocking shelves in a retail environment. The company views these competitions as opportunities to enhance commercial robot capabilities. “Galbot will keep pushing forward humanoid robot abilities by tackling increasingly difficult scenarios,” stated Yvonne Yuan, a company representative, in an email. “The goal is to transition from structured situations to more unpredictable real-world applications where robots must learn to understand, adapt, and collaborate with humans.”