I spent a week recording myself doing household chores for money. Who is the robot now?

Me, no He is just a human being. I am a channel of reality, and a means of messages. I have a knife in my hand and cut an organic cucumber, curving it iPhone Strapped to my forehead it can hold all 10 fingers. Toss the slices into the salad bowl and finish recording. Somewhere, a little robot is a little smarter.

This was me for an entire week last month as I collected data from my apartment while teaching humanoids How to clean dishes, fold laundry, and pour drinks, among other menial tasks. If robots will live with us one day and… Help around the houseThey need to be developed Fine motor skills. I did my homework with pride (I don’t usually contribute to large datasets when I have my jockstrap down). I was happy to make some money too.

First-person videos, shot with a camera mounted on a person’s head or chest, are a growing need as more companies try to build robots and improve their AI models. Although the Internet is full of reproducible videos, highly specific clips — such as thousands of close-ups showing hands pouring water into a cup without spilling it — can be crucial for fine-tuning machines to excel at real-world tasks. This style of recording, which the industry calls selfish data, is in such high demand that some… Investor estimates Leading companies will purchase hundreds of millions of watches from third-party suppliers over the next few years.

“I want every person on the planet to record themselves washing dishes,” says Avi Patel, the 22-year-old founder of the data collection marketplace Clied. “This will create a robot so you never have to wash dishes again.” Selfish data collection is already growing in countries like India, where freelancers generally roam around $125 per month on averageand first-person video vans can offer similar prices.

As interest grows, more data collection companies are looking to expand into the U.S., like DoorDash’s standalone to-do app that launched earlier this year. And before long, a lot Gig workers in the United States He may start actually catering to make ends meet, in addition to eating outside at room temperature.

Fortunately, I already had it Smartphone The head mount is in my possession from the tester DoorDash Tasks app. My impression, even at the time, was that detailed video data was the dystopian future Party workBut I wanted to understand this growing industry better. Since the to-do app isn’t available in California, where I live, I signed up for three other platforms: Kled, Luel, and Waffle Video.

The money I earned was minimal. I basically trained the robots for almost free and didn’t make a dent in the $2,500 a month San Francisco rent I split with my partner. But the carts had one unexpected advantage: my apartment had never been this clean.

The moment the clade erupted It came when Patel published A Video on X Earlier this year, part of the company’s extensive archive of video data was showcased. Soon the clip had been viewed more than 4 million times, and data hoarders began blowing up Patel’s phone. “Every founding model and major laboratory has reached out to me and asked for data,” he told me.

The bot’s training data is just part of what Kled collects from its more than 300,000 users — the startup mostly pays people to upload their entire camera roll as training data for the AI. Patel has seen early adopters stick with the gig in Malaysia, and there’s a “Special Tasks” section to help promote video submissions. Users select, from a list, the chore they want to photograph and then capture the content directly through the app. No hourly rate is listed for these; Each is categorized as low, medium, or high, with no specific range. (The company says that in about a month, the update will include pricing for many tasks, but not all.)