In China, people can earn 4,000–6,000 yuan (US$600–900) a month by working as robot trainers. They repeat everyday chores so that robots can learn to do the same jobs at home.

Workers pick up cups, open bottles and wash dishes many times a day. Sensors are fixed to their wrists and fingers, so every small movement is recorded while they follow instructions on a screen.

Monitors watch how many actions each worker does. Employees usually record about 160 movements per shift; the best workers do more than 260. All the recorded moves become training data for robots.

The job is physically tough. Liu Yun washed dishes and scrubbed tables up to 200 times in one shift, leaving her exhausted. This shows why the work can be demanding even though the pay is modest.

The data from human workers is the key that lets future robots learn everyday tasks. Without these recorded movements, robots could not know how to help at home.