The New Invisible Labor Behind Humanoid Robots Is Being Filmed in Apartments Around the World

Humanoid robotics has found a new kind of gig work

The race to build humanoid robots is generating a labor market that looks less like a factory and more like a distributed content platform. MIT Technology Review reports that Micro1, a U.S. company based in Palo Alto, has hired thousands of contract workers across more than 50 countries to record themselves performing everyday chores with iPhones mounted on their heads. The footage is then sold to robotics firms training humanoid systems.

The workers are not annotating text or reviewing chatbot outputs. They are turning their bodies into data collection instruments. Folding laundry, washing dishes, cooking, making beds, and moving through ordinary domestic routines are being captured from a first-person perspective so machines can learn how humans manipulate physical environments.

The arrangement illustrates a broader shift in robotics. Just as large language models were improved by training on vast quantities of internet text, many robotics researchers now believe humanoids can improve by training on massive datasets of movement and task execution. The core intuition is simple: if a robot is supposed to act in the human world, it needs far more examples of what human action looks like in practice.

Why this work is spreading globally

Micro1 has recruited workers in countries including Nigeria, India, and Argentina, places with large pools of tech-literate young people seeking income. The pay can be attractive by local standards. One worker identified by the pseudonym Zeus, a medical student in Nigeria, told the magazine he earns $15 an hour. In a strained economy with high unemployment, that is meaningful money.

But the work is also repetitive and strange. Zeus described spending hours ironing clothes and carefully keeping his hands within the camera frame. The labor requires discipline and physical mimicry, but not much creative or technical control. It sits at the intersection of digital piecework and embodied performance.

That combination is revealing. AI labor has often been described as hidden cognitive work: labeling, moderating, ranking, transcribing, and correcting. Humanoid training expands the category into physical domestic life. Workers are no longer just helping machines understand language or images. They are teaching them how to inhabit space.

The ethical questions arrive quickly

MIT Technology Review notes that the jobs raise difficult issues around privacy and informed consent. That is unsurprising. Recording inside homes creates obvious questions about what else enters the frame, how environments are represented, and what workers fully understand about the downstream uses of their data. A dataset of chores may sound mundane, but domestic space is intimate. Capturing it at scale for commercial model training changes its meaning.

There is also a question of power and value distribution. The data is becoming a critical input for companies racing to commercialize humanoids, including firms such as Tesla, Figure AI, and Agility Robotics, as the article notes. Yet the workers producing that input remain contingent, globally distributed, and largely anonymous. They are paid for the act of recording, not for any long-term upside in the systems their labor helps make possible.

None of that means the work is necessarily exploitative in every instance. Some workers may see it as flexible, comparatively well-paid, and preferable to local alternatives. But it does mean the robotics boom is not only a story of advanced hardware and foundation models. It is also a story of labor arbitrage, data extraction, and the outsourcing of machine embodiment to people far from the companies selling the future.

What this trend shows

• Humanoid robot training increasingly depends on large-scale real-world movement data.
• That data is being collected by contract workers filming household tasks in their own homes.
• The model raises immediate questions about privacy, consent, and how value is shared.

The striking thing about this system is how ordinary it looks. The future of humanoid robots is being built not only in research labs and venture-backed factories, but in apartments where workers wear phones on their heads and perform chores for the camera. That is an important correction to the mythology of automation. Before robots can imitate domestic life, someone still has to do the dishes first.

This article is based on reporting by MIT Technology Review. Read the original article.