A startup called Shift is offering to clean New Yorkers' homes for free — with one condition. The company wants to record everything its cleaners do: scrubbing dishes, wiping counters, dusting furniture, mopping floors. Shift plans to expand this service to London and other cities, according to a report by Robert Hart at The Verge. The footage will be used to train robots to handle domestic chores.
This represents a broader trend in the AI industry. While chatbots and image generators were trained on data easily scraped from the internet, robots face a tougher challenge. They must understand three-dimensional space, motion, friction, and the countless variations of physical objects — tasks humans handle instinctively but that machines struggle to master. Folding laundry, picking up an apple, or pouring water have proven surprisingly difficult for roboticists to program.
The physical world is much harder to collect data from than digital content. High-quality training footage is becoming a critical bottleneck for companies developing physical AI systems. Shift isn't alone in this approach. Recent reporting revealed that India's home services platform Pronto has similarly been using clients' homes as a source of AI training footage, suggesting this model is gaining traction across the industry.