The Hidden Gig Economy Teaching Robots to Fold Your Laundry

The chatbots of the last few years learned by reading the internet — trillions of words already sitting on servers, waiting to be ingested. Humanoid robots can't do that. To fold laundry, wash dishes, or move through a kitchen, a robot needs data about the messy physical world, and that data largely doesn't exist yet. As MIT Technology Review reported in April 2026, the race to create it has spawned something unexpected: a sprawling global gig economy of people filming themselves doing chores.

Why robots need a different kind of data

Virtual simulations can teach a robot acrobatics, but they struggle to model real-world physics accurately enough to teach the subtler skills — how gently to set down a glass, how to grip a thousand different household objects at the right angle and force. For robots to work in factories and homes, the prevailing bet is that real-world data, however time-consuming and expensive to collect, is what's needed. Investors clearly believe it: they poured over $6 billion into humanoid robots in 2025.

The new global workforce

That demand has created jobs in an unlikely form. Data companies like Scale AI, Encord, and Micro1 are recruiting armies of "data recorders" across more than 50 countries — including India, Nigeria, and Argentina — where tech-savvy young people mount iPhones on their heads and film themselves folding laundry, washing dishes, and cooking. Even DoorDash has paid delivery drivers to record themselves doing chores. The pay can be good by local standards, but the work is unusual and the questions it raises are serious.

The privacy and consent problem

Capturing the physical world in this detail means capturing homes, routines, and the people in them. As the reporting notes, this raises real questions about privacy and informed consent — what exactly is being recorded, who owns it, and whether the people filmed in the background ever agreed. It echoes earlier waves of hidden human labor behind "automated" systems, from content moderation to AI data labeling, often performed wherever labor is cheapest.

What it means for businesses

For most organizations, humanoid robots in the workplace are still a few years out — but the trend is worth tracking for two reasons. First, the sheer capital flowing into physical AI signals that robotics is maturing fast, and long-term operational planning should leave room for it. Second, the data-sourcing story is a reminder that impressive AI capabilities often rest on large amounts of human work and real-world data — with cost, privacy, and ethical dimensions that don't disappear just because the end product looks autonomous.

Source: "The gig workers who are training humanoid robots at home," MIT Technology Review, April 1, 2026.