Indian Workers Train AI Robots via GoPro Videos in Bengaluru Data Farms
In the bustling tech hubs of Bengaluru, a silent revolution is unfolding not through lines of code, but through the lens of GoPro cameras strapped to workers' foreheads. Thousands of Indian laborers are now employed in what are termed "hand movement farms," where they spend grueling ten-hour shifts meticulously performing tasks like folding towels and stacking boxes. To an outsider, this might appear as an eccentric art installation, but in reality, it serves as the essential raw material powering the global humanoid robot revolution.
The Process: Capturing Human Movement for AI Training
Imagine a young worker standing in a sterile room with a GoPro mounted on their head. They reach into a basket with one hand, shake a piece of fabric with both hands, fold it precisely three times, and then place it in the left corner—all within a strict sixty-second timeframe. This is not an isolated scene; it is replicated thousands of times daily across India. These videos are meticulously recorded and dispatched to advanced AI laboratories in the United States, where sophisticated neural networks analyze every subtle nuance to teach machines the intricate movements of the human body.
The scope extends far beyond towel folding. Workers are also capturing a diverse array of tasks, including sorting utensils, crumpling paper, and performing other mundane activities. From Tesla's Optimus robot to Figure AI's cutting-edge prototypes, every major tech company developing humanoid robots requires vast datasets of real-world movements to master complex functions like cooking, plugging cables, or doing laundry. Collecting this data in high-wage countries such as America is prohibitively expensive, leading firms to turn to the cost-effective labor markets in India.
Objectways and Micro1: Leading the Data Labeling Industry
Objectways, a Bengaluru-based data labeling company founded by the young entrepreneur Dev Mandal at just twenty years old, is at the forefront of this burgeoning market in India. With over 2,000 employees—half engineers and half annotators—the firm produces hundreds of such training videos daily. Another key player, Micro1, employs locals across India, Brazil, and Argentina, paying them to wear augmented reality glasses while performing daily chores, thereby capturing even more diverse movement data.
The Irony: Training Their Own Replacements
Many of these workers may be unaware of a stark reality: the videos they are creating are training the very robots designed to eventually replace their jobs. From domestic maids to industrial cleaners, daily chores are expected to be fully automated by robots in the coming years. Videos of Indian workers recording tasks for robot training have allegedly gone viral online, sparking widespread concern among AI skeptics about the future implications.
While the exact purpose of these videos might not be explicitly confirmed, the existence of such data factories is undeniable. As one user on X poignantly wrote, "The workers are training their own replacements in real time, on the job, getting paid to do it. The most honest description of this is also the most uncomfortable one." Another added, "The most cold-blooded AI story you'll read today: pay workers to teach AI their jobs then replace them with the AI they trained they are literally filming their own replacement."
A third commentator highlighted the deeper issue: "Every shortcut they've perfected over the years… now a training label. That's not just a job shift, it's institutional knowledge being quietly extracted. The uncomfortable part may not be the camera, but what happens after, & this is the current shift in many pro-AI sectors globally."
The Human Cost of Robotic Labor
The global robotics market is projected to reach $88 billion by 2026 and soar to $218 billion by 2031, with investors pouring over $6 billion into humanoid robots in 2025 alone. While this represents a massive income boom for tech giants, for the workers creating these videos, it often translates into silent labor exploitation.
In India's data farms, compensation typically ranges from $230 to $250 per month for full-time shifts, with part-timers earning between $120 and $140. However, the physical and mental toll far outweighs the financial returns. Workers endure eye strain from mounted cameras, wrist fatigue from repetitive motions, and the psychological burden of monotony. These are just some of the harsh repercussions faced in these dataset factories, highlighting the human cost behind the robotic labor revolution.
