Buyers are pouring cash feverishly into fixing this problem, spending over $6 billion on humanoid robots in 2025. And at-home knowledge recording is changing into a booming gig economic system around the globe. Information corporations like Scale AI and Encord are recruiting their very own armies of information recorders, whereas DoorDash pays supply drivers to movie themselves doing chores. And in China, staff in dozens of state-owned robotic coaching facilities put on virtual-reality headsets and exoskeletons to show humanoid robots learn how to open a microwave and wipe down the desk.
“There may be a whole lot of demand, and it’s rising actually quick,” says Ali Ansari, CEO of Micro1. He estimates that robotics corporations are actually spending greater than $100 million annually to purchase real-world knowledge from his firm and others prefer it.
A day within the life
Employees at Micro1 are vetted by an AI agent named Zara that conducts interviews and critiques samples of chore movies. Each week, they submit movies of themselves doing chores round their properties, following a listing of directions about issues like protecting their arms seen and shifting at pure velocity. The movies are reviewed by each AI and a human and are both accepted or rejected. They’re then annotated by AI and a group of a whole bunch of people who label the actions within the footage.
“There may be a whole lot of demand, and it’s rising actually quick.”
Ali Ansari, CEO of Micro1
As a result of this strategy to coaching robots is in its infancy, it’s not clear but what makes good coaching knowledge. Nonetheless, “you might want to give tons and plenty of variations for the robotic to generalize properly for fundamental navigation and manipulation of the world,” says Ansari.
However many staff say that creating quite a lot of “chore content material” of their tiny properties is a problem. Zeus, a scrappy pupil residing in a humble studio, struggles to document something past ironing his garments daily. Arjun, a tutor in Delhi, India, takes an hour to make a 15-minute video as a result of he spends a lot time brainstorming new chores.
“How a lot content material [can be made] within the house? How a lot content material?” he says.
There’s additionally the sticky query of privateness. Micro1 asks staff to not present their faces to the digital camera or reveal private data similar to names, cellphone numbers, and start dates. Then it makes use of AI and human reviewers to take away something that slips via.
However even with out faces, the movies seize an intimate slice of staff’ lives: the interiors of their properties, their possessions, their routines. And understanding what sort of private data they may be recording whereas they’re busy doing chores on digital camera might be tough. Critiques of such footage won’t filter out delicate data past the obvious identifiers.
