When we speak of AI, we think in abstractions: tokens, neural networks, job loss, the bubble that seems to run the whole economy. These are the words that describe an industry that, seemingly overnight, has "revolutionized" how we think, relate to one another, and produce value.

But behind every abstraction is something material. Marx calls this “commodity fetishism”, where the commodity’s value appears intrinsic to it, obscuring the human labor and social relations that bring the commodity into existence. Journalists and academics have debunked this myth again and again with the vast network of material resources underpinning AI’s development. There are data workers who manually label pictures into a “tree” or a “car” to power autonomous vehicles, or paragraphs into “suicide”, “bestiality”, or “child pornography” to train safe model outputs with under $2 of hourly wage. There are data centers a fifth the size of Manhattan that raise electricity prices and pollute the water of marginalized communities. Behind every GPU chip–the purchase of which makes a few companies the richest in the world and fuels the AI bubble–is raw mineral extraction, dug from the lands of Congo by artisanal miners of unequipped men, women, and children. The militia violence and child labor behind key minerals that fuel our technologies go unnoticed.

The public is made aware of these supply chains, yet our understanding remains limited to a few high-profile cases. Due to proprietary concerns, the unregulated market economy, or intentional maneuvering using complex subcontracting, nondisclosure agreements, and more, much of the industry behind AI is conveniently opaque, avoiding investigation and public accountability. This covers up a story of AI colonialism: the demands from the global elites accelerate the exploitation of workers and lands in the Global South, perpetrating a pattern of violence that is concealed from the public eye. We need to make these industries transparent, to reveal the vastness and interconnectedness of such a network, to find solidarity among sites of exploitation and resistance, and to empower the public to hold the industry accountable.

Our first prototype, “Where are the data workers behind AI?”, is an investigative attempt to make the data annotation and labeling (DAL)  industry transparent. With an original dataset, the map moves beyond anecdotal reporting of labor practices in the DAL industry to systematically documenting data workers’ locations for each data labeling company, thereby producing an empirical examination of worker distribution in the DAL industry. 

The map layers worker locations against the DAL platforms they work for and the customers those platforms serve, each layer indicated with different shapes and colors. Their distribution on the map testifies to AI colonialism, where data labeling demand from the Global North fuels and accelerates systematic exploitation of workers in the Global South. 

Behind every node is a real worker, real socioeconomic conditions of a country, and a neoliberal world order that extends past any single company. The map wishes to highlight these human stories alongside this vast global network. The story of all nodes in the map is tied together in the “Explore the data worker story” feature.

A more detailed description of the map and its limitations is documented in the About and Methodology pages. Maneuver to the map to learn more. The data worker map is the first in the series, “Power Mapping AI’s Materiality”. A second prototype, “Where are the critical minerals behind AI”, is forthcoming.

We want to acknowledge the many influential projects that informed this work, including but not limited to: Empire of AI; Ghost Work; Data Workers’ Inquiry; Data Work Landscape; Billy Perrigo’s reporting; and so many more. Given the rich landscape of scholarly and public work, we hope our map can provide empirical backing to existing reporting, together telling the data worker story. 

We hope this is a long-term project that moves beyond informing the public and also empowers folks to locate (and potentially disrupt) the material supply chain behind AI. Any collaboration, questions, feedback, and critique are highly welcome. 

Please email ai-materiality-map@proton.me for communication.