Companies Now Pay Workers to Record Household Tasks for Robot Training Data

Firms are compensating workers to wear head and chest cameras while performing routine chores—scrubbing dishes, folding laundry, pouring drinks—to build first-person video datasets for training humanoid robots. The hyperspecific collections capture thousands of variations: hands pouring water without spilling, cleaning motions on different surfaces, precise grip adjustments. This shift from laboratory-controlled demonstrations to real-world human movement reflects how companies intend to train robots for domestic environments at scale.