Pennies Per Microtask

The machines do not, in fact, learn by themselves. They are taught — by people, in shifts, for wages — and the question of who those people are, where they live, what they earn, and what they see on their screens has been one of the more revealing fault lines in the past two years of AI coverage. The promise has been narrated as automation; the reality has been narrated, when it has been narrated at all, as labor. Between those two narrations sits an entire global workforce: annotators in Nairobi rating slurs and abuse imagery, miners in Sulawesi pulling tin out of the ground, content moderators in Manila absorbing the worst the internet produces so that the chatbot in your classroom answers politely. The promise rises; the workers are mentioned, briefly, then receded from view.

This is the arc the column traces this week. From late 2024 through the present, the conversation about AI's hidden labor has cycled through three distinct registers: a phase of discovery, when "invisible workforce" became a phrase respectable institutions were willing to print; a phase of saturation, in which even the optimistic pieces conceded the existence of "ghost workers" and proposed reforms; and a phase of fading attention, in which the topic has not so much been resolved as simply absorbed into the background hum of AI commentary. The inversion is sharp — peak coverage in mid-2025, near-silence by late 2025 — and it is the kind of fade that deserves examination. What changed: the conditions, or the willingness to look at them? The arc suggests the latter.

At the close of 2024, the rhetorical posture was tentative discovery. A December 2024 piece from the International Labour Organization, The Artificial Intelligence illusion: How invisible workers fuel the "automated" economy, opened with a sentence that would have been unprintable as policy language two years earlier: AI's success "hinges on an invisible workforce performing low-paid, precarious tasks under challenging conditions." The framing was important because the ILO is not a campaign group; it is a tripartite UN body whose audience includes the ministries that write labor codes. In the same window, AI's Hidden Cost: Environmental and Social Toll of Rapid Growth translated the same concern into the register of environmental reporting, and The AI Dilemma: Powering the Future or Fueling Our Fears? framed the question as a sectoral reckoning. The probe counts for the quarter — two optimistic pieces, four critical — register the shift; the topic had moved from the margins of tech criticism into mainstream business and policy outlets.