Bots on the ground: China leads humanoid race

In Beijing this week, a humanoid robot named Tiangong Ultra ran 100 meters in 9.39 seconds, beating Usain Bolt's world record. It then crashed into a wall. Other robots at the World Humanoid Robot Games caught fire, fell over, or failed in other spectacular ways. The games—featuring robot boxing, powerlifting, and football—were meant to showcase progress, but they also revealed how far the technology still has to go. China is now the undisputed leader in humanoid robotics, but the machines remain expensive, clumsy, and largely useless for everyday tasks.

Why humanoid robots are so hard to build

Teaching a chatbot to converse is relatively straightforward: you feed it millions of pages of text from the internet. Teaching a robot to pick up a glass of water without shattering it is a different order of problem. As Economist China business editor Don Weinland explains, humans perform countless physical tasks without conscious thought—knowing how much pressure to apply, how water sloshes, how to adjust grip. Robots must be trained to understand all of this from scratch. The scale of the challenge is staggering: researchers estimate that collecting enough real-world data to make humanoid robots truly capable would require roughly 100 million hours of training data—more than 11,000 years' worth. "Suffice it to say that we are really far away from collecting enough data to mak