These workers help train robots who could one day steal their jobs - Houston Public Media

TEXT START: KARUR, India — Three men stride across a cavernous textile factory floor of Texworth in the southern Indian city of Karur, handing GoPros and smart glasses to workers, including 40-year-old Shoba Kirubakaran.

The Dissection

The report documents a labor-to-capital conversion pipeline. Workers perform tasks; cameras capture their embodied procedures; firms package those movements as training data; AI and robotics companies acquire a replicable control layer. The worker becomes wage laborer, sensor, and raw material simultaneously.

The article’s emotional center is exploitation. Its structural center is ownership: workers generate the data, while companies acquire the system that may eliminate the need for their labor. The training phase is not separate from automation. It is automation’s construction phase.

The Core Fallacy

The article treats robot cost, power consumption, and unreliable infrastructure as meaningful counters to obsolescence. Under the Discontinuity Thesis, they are lag defenses. Engineering friction can delay deployment; it cannot preserve the mass-employment circuit indefinitely.

“Humans are cheaper” is a temporary price snapshot, not a permanent law. Hardware costs can fall, energy systems can improve, and trained capabilities can be deployed in richer or more reliable markets. India’s delay would not protect Indian labor from the global consequences.

The second fallacy is the assumption that retraining can absorb the displaced. The report offers no mechanism capable of creating hundreds of millions of economically necessary roles once machines perform the underlying tasks. Retraining workers for a vanished demand structure is administrative theater, not recovery.

Hidden Assumptions

• Human labor will remain cheaper than machines after scale, learning, and capital investment.
• Physical work will retain a durable human moat because deployment is difficult today.
• Automation will occur only where the data was collected.
• New occupations will absorb displaced workers at comparable scale and income.
• Workers will retain ownership or bargaining power over the data extracted from them.
• Government institutions can react faster than competitive automation spreads.
• Farming or informal work constitutes a viable replacement rather than subsistence fallback.
• The fact that replacement is not immediate means it is not structurally underway.

Social Function

Partial truth functioning as transition management and ideological anesthetic. The report correctly exposes coercive data extraction and the workers’ lack of informed consent. It also records the replacement pipeline with unusual clarity.

But its emphasis on current expense, electricity constraints, and future retraining makes the terminal process appear negotiable and distant. “Humans are cheaper” comforts the audience by turning a temporary bottleneck into a presumed defense. The result is a familiar social ritual: condemn exploitation, acknowledge disruption, invoke policy agility, and avoid confronting the ownership problem.

The Verdict

This is a pre-obsolescence autopsy performed while the body is still clocking in. The workers are already being converted from producers into training material for machines owned by someone else.

P1 is visible: human motion is being translated into machine capability. P2 is visible in the employer’s ability to record workers without meaningful explanation or consent. P3 has begun: the job is being transformed from a livelihood into a dataset and then into a replicable machine procedure.

The power crisis may postpone deployment. It does not reverse the mechanism. The decisive fact is not whether a robot replaces Kirubakaran next month. It is that her skill is already being extracted, standardized, and prepared for ownership by capital. The article’s future tense is camouflage. The replacement process has already started.