In Karur, Textile Workers Wear GoPros to Train Robots

At a Textile Factory in Karur, Workers Had GoPros Strapped to Their Foreheads

There’s something odd about a textile factory in Karur, a small city in Tamil Nadu, India. It looks like an ordinary workshop—people sticking on labels, ironing cloth bags, folding clothes—except that around eight of the workers are doing their jobs with GoPro cameras strapped to their foreheads or smart glasses on their faces. An AFP reporter witnessed the scene firsthand.¹

One woman working there is 28 years old and quit her job as a teacher after giving birth. What she does now is simply film herself for 6 to 7 hours a day as she folds laundry, washes dishes, and tidies up her child’s toys—her ordinary routine, captured on camera. In exchange, she earns an income. But this footage is being used to train robots that will one day do her job for her.

Watching this, I couldn’t help but think back to the piece I wrote last issue about China’s brain-computer interfaces (BCI).² China’s state-backed industrial push and India’s video-data harvesting are different businesses, but they invite the same question: who ends up capturing the value that the technology creates? A state-supported tech company and a worker paid to film herself end up profiting from the future in very different ways.

---

China Is Building Brain-Signal Technology Into an Industry

As I covered in the last issue, China has set its sights on the underlying technology for reading and decoding human neural signals.

In July 2025, seven Chinese government bodies—including the Ministry of Industry and Information Technology—jointly released a blueprint for cultivating the BCI industry. It’s a plan that pursues core technology development, standards, and industrial ecosystem-building all at once. The goals are concrete, too: achieve breakthroughs in electrodes, neural chips, and signal-decoding algorithms by 2027, and cultivate 2-3 “world-class” companies by 2030. In December 2025, Shenzhen announced a brain science industry fund and early-stage investment agreements totaling ¥1.16 billion (~$160 million).³

The targets for cultivation include hardware like electrodes and chips, as well as signal-decoding algorithms. It’s a policy designed to help domestic companies secure core technology and intellectual property. But we should distinguish between the government supporting an industry and the state directly owning every piece of technology.

Of course, this path has its limits. Invasive technology that implants electrodes in the human brain requires clinical validation and ethical review, and the market is opening slowly. China’s BCI market is estimated at just over $500 million even in 2025. But the direction is clear: China intends to build technology for handling neural signals into a domestic industrial capability.

India Is Harvesting First-Person Video Data at Scale

The business introduced in Karur, a town in Tamil Nadu, India, is different. What’s being collected isn’t brain signals but scenes from everyday life as seen by ordinary people, and the equipment isn’t surgical electrodes but secondhand GoPros and smartphones. This business earns revenue by fulfilling data orders from overseas clients and others. That doesn’t mean India’s entire AI industry operates this way.

The company collecting data in Karur is Objectways. It started in 2019 as a data annotation⁴ company and has since expanded into robotics data. It has offices in India and the US, counts Fortune 500 companies among its clients, and works with machine-learning platforms like Amazon SageMaker. Its CEO is originally from Tamil Nadu but now lives in the US and is in his 50s.¹

What they’re collecting is “egocentric data.”⁵ Put simply, it’s human daily life itself, filmed from a first-person point of view — folding clothes, making coffee, washing dishes, even cleaning the toilet. The hand movements and field-of-view footage captured this way become the basic raw material for training humanoid robots. The idea is to use footage of people picking up and moving objects to teach robots how to perform similar motions. The type and volume of data needed varies by model and task. Morgan Stanley has projected that more than 1 billion humanoid robots will be in use worldwide by 2050.

The video claims the industry expects “500 million hours of data per day.” I find that figure hard to take at face value. The source is unclear, and physically capturing 500 million hours of footage in a single day would require tens of millions of people filming simultaneously. It’s better read as rhetoric expressing the industry’s ambition than treated as a verified fact. What is confirmed, though, is this: according to AFP’s reporting, workers earn about 250 rupees (roughly $2.6) per hour of footage.¹ Another worker featured in the video said they receive an additional 10,000 rupees a month as a camera-wearing allowance.

In this business, workers are paid for the act of filming. The revenue earned from providing that data to clients is divided according to the contract between the collection company and the ordering firm. What rights or compensation workers hold over the reuse of their footage is a matter that depends on the contract, and news coverage alone can’t settle who owns every piece of video.

Companies that build the tech vs. workers who make the data

As I build go-to-market strategy, I keep coming back to the same question: “Who in this value chain actually holds the defensible asset?” Applying that question to these two businesses makes you look at who ends up owning the fruits of development.

China’s policy of fostering BCI (brain-computer interface) development supports domestic companies in securing the technology and the IP. The workers in Karur earn income from filming videos, but whether they earn anything further when that data gets reused later is a separate question entirely. Still, this one case doesn’t let us cleanly split the world into “China owns the technology, India just supplies labor.” Both countries have a wide range of tech companies and workers.

India’s IT services industry faces a similar question. Software services exports grew to over $200 billion in FY25.⁶ Of that, the repetitive development and support work is a plausible candidate for AI automation. But that doesn’t mean the entire industry lacks its own technology or data. Companies differ in how much weight they place on client relationships, domain knowledge, and proprietary solutions — and their capacity to adapt differs accordingly.

Some investors see this shift as bad news for India’s growth outlook. In 2026, foreign capital pulled a record-breaking $30 billion-plus out of Indian equities.⁷ Coverage of the exodus draws comparisons to the capital flowing instead into semiconductor firms in Taiwan and Korea. Still, you can’t pin the entirety of that capital flight from Indian equities on AI competitiveness alone. One investor went so far as to call India an “anti-AI play” outright. There’s a counterargument, though. India is home to over 1,800 Global Capability Centers (GCCs)⁸ employing roughly 2 million people, and 80% of new GCCs list AI and machine learning as a top priority. From this angle, India looks like a hub for AI enterprise application — fine-tuning, in industry parlance. But GCC work isn’t limited to model fine-tuning. The knowledge built while developing applied technology and products can itself become an asset. What matters is who ends up owning those results, and what career paths and compensation are left for the local workforce.

Immediate Income Versus Long-Term Labor Conditions

For workers in Karur, India, this job is an immediate income opportunity. Alongside the longer-term question of data rights, we need to look at what life choices this work currently offers.

One woman in the video introduced herself as the first in her family to graduate from college, saying her wages support her parents’ health and her child’s education. Being able to earn income without leaving her hometown is, for her, a meaningful opportunity. We can’t generalize this one experience to the situation of all Indian women, but it helps us understand why a worker would choose this job.

At the same time, filming work requires its own set of conditions. Decisions need to be made about what footage to collect and how much, whether workers can refuse to film certain things, how to protect the privacy of people around them, and what compensation applies for additional tasks. Workers also need to be told that the data they contribute to robot training could later be used to automate their own jobs.

The conditions of this work aren’t determined by data rights alone. Wages, voluntary consent, safety and privacy, and career advancement opportunities all need to be considered together. Nor can we assume that every filming job will disappear the moment a robot finishes training — new tasks and environments may continue to require fresh data. Still, there’s no guarantee that today’s jobs will last, so workers need to be able to choose with full knowledge of the risk.

Oswarld’s Lens

Honestly, I take a cautious view of India’s narrative that “population equals competitiveness.”

What I’ve often observed while working with data is that sheer volume alone doesn’t produce a lasting competitive edge. In the process of gathering data, you need to accumulate technology and customer relationships, and be able to reinvest those gains. I think the real question is whether video-labeling work gives workers not just wages, but also the chance to build new skills and move into better jobs.

This isn’t to say China’s path is the right answer. State-led development of neural signal technology has its own problems—privacy, control, research ethics. What I’m proposing is a criterion for judgment. When looking at any country’s AI strategy, if you first check which technologies and data that country actually owns—rather than just the volume of work being done—you can much more quickly gauge whether that work leaves workers with a lasting asset or not.

Closing

China’s BCI push and India’s video-data business both raise the same question: who ultimately benefits from technological progress? In Karur, filming work provides immediate income, but whether it also guarantees long-term compensation and data rights is a separate matter that needs scrutiny. I think we need to look not just at the scale of the AI industry, but at the terms under which workers actually participate in it.

So the next time you see headlines about some country’s “AI rise,” I’d encourage you to look past the scale and speed, and check who actually owns the technology and the data.

If you were weighing whether to take on this kind of filming work, what conditions would you check first? Wages, the scope of data use, the right to refuse filming, future career opportunities — let me know in the comments which of these matters most to you.

---

📨 If this perspective was useful, please pass it on to a colleague

---

The draft looks accurate and complete. One fix needed: the wage figure should be in Indian Rupees (₩ is Korean won), matching “250 rupees” (250 rupees) in the source.

References & Further Reading

Primary sources

• AFP, “The Indian workers training AI robots to take their jobs”, 2026.6. As published on Al Jazeera ··· the primary source for the GoPro labor and ₹250 (~$3) hourly wages at the Karur factory, and the ObjectWays information.
• Seven Chinese ministries, “Implementation Opinions on Promoting the Innovation and Development of the Brain-Computer Interface Industry”, 2025.7. TechCrunch analysis ··· an article explaining China’s policy push to cultivate its BCI industry.
• Nasscom, “Technology Sector Strategic Review 2025”. India Business Trade summary ··· the basis for the figures on India’s software services export volume.

Background

• Ruchir Sharma, “India is a loser in the AI race” interview, 2026.5. Business Today ··· gives an investor’s-eye view of why foreign capital is leaving India for Taiwan and Korea.

Worth reading alongside this issue

• INLEVEL9 Letter: the China BCI edition ··· covers the direction of China’s BCI development. It connects directly to this issue’s comparison of industrial policy.

---

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.

📝 Glossary

Footnotes

1. ObjectWays / Karur camera labor: A case AFP reported firsthand in June 2026 at a textile factory in Karur, Tamil Nadu. Workers wear GoPros and smart glasses on their foreheads to film everyday movements, and the footage is sold as training data for humanoid robots. ↩ ↩² ↩³

2. BCI (Brain-Computer Interface): Technology that reads neural signals from the brain and connects them to a computer or machine. It splits into invasive types, where electrodes are implanted in the brain, and non-invasive types, which measure signals from outside. ↩

3. Shenzhen Brain Science Industry Fund: Announced in Shenzhen in December 2025 with a total scale of ¥1.16 billion. The Shenzhen Guangming District government announcement puts this at roughly $160 million. It shouldn’t be read as a single national fund dedicated solely to BCI. ↩

4. Data Annotation: The work of labeling data — images, video, audio — with tags like “this is a truck, this is a person” so AI can learn from it. It’s the same process that lets self-driving cars distinguish objects on the road. ↩

5. Egocentric Data: Data shot from a “first-person point of view.” Filmed at human head and eye level, it’s training material that helps robots directly imitate human hand movements and field of vision. ↩

6. Software Services Exports (FY25): Covers India’s fiscal year from April 2024 to March 2025. The RBI survey estimate — which excludes revenue from overseas subsidiaries — puts this at $204.7 billion, a different scope than total tech-industry exports. Reference: an Indian state investment agency’s summary of RBI statistics. ↩

7. $30 billion in capital outflows: The amount foreign investors pulled from Indian equity markets in 2026 — already surpassing 2025’s full-year record, making it the largest outflow on record. ↩

8. GCC (Global Capability Center): A self-operated hub that multinational corporations set up in countries like India. It represents work once outsourced through BPO arrangements being brought “in-house” instead. India hosts more than 1,800 such centers, employing roughly 2 million people. ↩