Atlas of AI

In The You On AI Encyclopedia

Crawford's thesis is that AI is not intelligent, not artificial, and not merely software. It is a "material, computational, and infrastructural regime" whose operations depend on specific forms of extraction, labor exploitation, and environmental damage that the industry's self-presentation systematically obscures. The book is organized as an atlas — a geographic mapping — precisely because AI's reality is geographic and material, not merely computational.

The book documents cobalt mines in Kolwezi where children work in conditions of extraordinary toxicity; lithium brine extraction in the Atacama Desert that has depleted aquifers indigenous communities depend on; rare earth processing in Inner Mongolia that has produced lakes of radioactive sludge; content moderation centers in the Philippines where workers absorb the psychological violence of screening the internet's worst content; and the training-data harvesting that has captured the intellectual commons into corporate proprietary systems.

Kate Raworth

For Raworth's framework, Atlas of AI provides empirical ground. The embedded economy insists that economic activity be understood as materially embedded in the biosphere. Crawford's atlas is the detailed cartography of that embedding for AI specifically — the visible evidence that the ecological ceiling is being transgressed not incidentally but systematically, by design choices that prioritize throughput over sufficiency.

The book has become foundational for the AI critique that integrates political economy, labor analysis, and environmental reckoning. It stands alongside Shoshana Zuboff's Surveillance Capitalism and Jaron Lanier's Who Owns the Future? as core texts of the structural critique of the AI industry.

Origin

Crawford is a research professor at USC Annenberg and senior principal researcher at Microsoft Research. She co-founded the AI Now Institute at NYU. Atlas of AI emerged from nearly a decade of ethnographic and archival research tracing AI's material substrate across six continents.

Key Ideas

AI is material. The book's central thesis: AI operations are physical events with geographic, ecological, and labor consequences.

Extraction at every layer. From minerals to data to labor, AI operates through extraction structured by existing power asymmetries.

Classification as politics. Crawford documents how classification systems in AI models encode specific political judgments disguised as technical decisions.

Atlas methodology. The geographic framing is substantive: AI's reality cannot be understood without cartography.