I'm Training the Robot That Will Take My Job. For $15 an Hour.
Who's buying this? Tesla. Google. Figure AI. They use it to teach humanoid robots how to move like humans - a technique called "imitation learning."
The person teaching a robot to wash dishes isn't an AI engineer. It's a gig worker in a developing country, earning $15 an hour.
50 Countries, Thousands of Workers - The Scale of Robot Training Data
A US startup called Micro1 built this pipeline. They've recruited thousands of workers across 50 countries - Kenya, the Philippines, India, Brazil. These workers strap iPhones to their heads and film themselves doing chores. Folding laundry. Cooking meals. Mopping floors.
Scale AI has already collected over 100,000 hours of footage. The robotics industry is pouring more than $100 million a year into this kind of data.
By the numbers, it's a massive industry. But there's something odd about the whole arrangement.
They Have No Idea What They're Actually Selling
$15 an hour. In Kenya, where the average wage is $1-2 an hour, that's decent money.
The problem lies elsewhere. These workers have no idea what they're actually selling. From their perspective, it's a simple side gig - film yourself doing housework. But what they're really producing is the core training data for a $25 trillion robot market. How many of them understand that distinction?
And there's another layer. They're filming inside their homes with a camera strapped to their heads. Family photos on the wall. Kids' toys on the floor. Medication on the counter. The contents of their fridge. Their entire private lives, captured in high definition. Where is this data stored? Who has access? How long is it kept? No one's telling them.
The Side Hustle That Kills the Day Job
It gets more uncomfortable.
Who are these workers in their day jobs? They assemble parts in factories. They clean hotel rooms. They move boxes in warehouses. The filming is a side hustle. They do it because their main paycheck isn't enough and $15 an hour is hard to turn down.
Now here's the part that should make you pause. The robots being trained by this footage - what jobs will they replace first? Factory assembly. Hotel housekeeping. Warehouse labor. The very jobs these workers depend on for a living.
Their side hustle is feeding the machine that will eventually eat their main income.
"It's Their Choice, Though"
At this point, you might be thinking: "Well, nobody forced them. It's their choice."
Fair enough. The pay is decent. The work is easy. No one held a gun to anyone's head.
But flip it around for a second. Imagine you're in Kenya, making $2 an hour at a factory. Someone offers you $15 an hour to just strap a camera on and do household chores. Would you say no?
I wouldn't. And that's exactly what makes this so uncomfortable. Calling it a "voluntary agreement" when the person barely has a choice. Saying the deal is fair because they signed the paperwork.
We're Not That Different, Honestly
Every day, we hand our search history to Google, post our lives on Instagram, and feed our viewing habits to YouTube. We have no idea what happens to that data either. The only difference is we don't even get $15 for it.
Well, there is one other difference. We're not worried about losing our jobs to the thing we're feeding. Not yet, anyway.
Key Takeaways
① Information asymmetry - Workers think it's a simple filming gig. In reality, they're producing core training data for a $25 trillion robot market
② The side hustle paradox - Factory, hotel, and warehouse workers film footage as a side gig, unknowingly training the robots that will replace their main jobs
③ Choice without a choice - When you're making $2 an hour, turning down $15 isn't really a decision you get to make
But one thing is clear. Knowing about this structure and not knowing - those are two very different things. Behind the AI we use, behind the robots heading to factory floors, there's a person with an iPhone strapped to their head, filming their daily life. Earning $15 an hour. With no clue what they're actually building. Training the very technology that will one day replace their livelihood.
Next time you see a headline about how amazing AI is, just ask one question: The person who made the raw material for this technology - where are they now, and what are they doing?
📌 Sources: The Information, Bloomberg, MIT Technology Review (2026)
댓글
댓글 쓰기