Traders 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 firms like Scale AI and Encord are recruiting their very own armies of knowledge recorders, whereas DoorDash pays delivery drivers to movie themselves doing chores. And in China, employees in dozens of state-owned robotic coaching facilities put on virtual-reality headsets and exoskeletons to show humanoid robots the way to open a microwave and wipe down the desk.
“There may be numerous demand, and it’s rising actually quick,” says Ali Ansari, CEO of Micro1. He estimates that robotics firms at the moment are spending greater than $100 million every year to purchase real-world knowledge from his firm and others prefer it.
A day within the life
Staff at Micro1 are vetted by an AI agent named Zara that conducts interviews and evaluations samples of chore movies. Each week, they submit movies of themselves doing chores round their houses, following a listing of directions about issues like conserving their palms 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 crew of a whole bunch of people who label the actions within the footage.
“There may be numerous demand, and it’s rising actually quick.”
Ali Ansari, CEO of Micro1
As a result of this method to coaching robots is in its infancy, it’s not clear but what makes good coaching knowledge. Nonetheless, “it’s good to give tons and plenty of variations for the robotic to generalize nicely for fundamental navigation and manipulation of the world,” says Ansari.
However many employees say that creating a wide range of “chore content material” of their tiny houses is a problem. Zeus, a scrappy pupil dwelling in a humble studio, struggles to report 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 dwelling? How a lot content material?” he says.
There’s additionally the sticky query of privateness. Micro1 asks employees to not present their faces to the digicam or reveal private data reminiscent of names, telephone 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 employees’ lives: the interiors of their houses, their possessions, their routines. And understanding what sort of private data they could be recording whereas they’re busy doing chores on digicam may be tough. Evaluations of such footage won’t filter out delicate data past the obvious identifiers.
