Atlas of AI
Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence, published by Yale in 2021, is Crawford's systematic mapping of the physical substrate of artificial intelligence. The book traces AI operations from the lithium mines of Nevada's Clayton Valley to the cobalt extraction in the Democratic Republic of Congo, from the data centers of Nevada to the click-work of content moderation, from the training corpora scraped from the commons to the classification systems encoded in model architectures.
The Infrastructure Inheritance — Contrarian ^ Opus
There is a parallel reading that begins from the political economy of technological transition rather than moral condemnation. The material substrate Crawford documents—the mines, the data centers, the content moderation facilities—represents not AI's unique depravity but the inherited infrastructure of industrial modernity itself. The cobalt that powers AI chips also powers electric vehicles championed by environmentalists. The rare earth processing that enables neural networks also enables wind turbines. The energy consumption of data centers pales beside aviation, shipping, or concrete production. AI did not create these supply chains; it inherited them from a century of industrialization that has already locked in specific patterns of extraction and labor organization.
Read through this lens, the Atlas becomes less an indictment of AI specifically than a documentation of capitalism's baseline operating conditions—conditions that every transformative technology must navigate. The question then shifts from whether AI exploits (it does, as does every industrial process) to whether its transformative potential justifies participation in these existing systems while working toward their reform. The content moderation Crawford highlights as traumatic labor might also be read as transitional—a human bridge to automated systems that could eventually eliminate such exposure. The data extraction she frames as theft could be understood as the awkward bootstrapping phase of systems that may eventually generate synthetic training data. This reading doesn't excuse the violence Crawford documents but suggests that condemning AI for participating in the only industrial infrastructure available risks preserving that infrastructure by preventing the technologies that might eventually transform it.
In the AI Story
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.
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.
Appears in the Orange Pill Cycle
The Materiality Gradient — Arbitrator ^ Opus
The right synthesis depends on which temporal frame we adopt. For documenting present harm, Crawford's empirical mapping is definitively correct (95%)—the mines exist, the workers suffer, the aquifers deplete. Her geographic methodology perfectly captures AI's current material reality. But for understanding AI's trajectory, the contrarian view gains weight (70%)—AI genuinely differs from traditional industry in its potential to dematerialize certain operations over time, even while remaining fundamentally embedded in material substrates. The question isn't whether AI has a material footprint but whether that footprint follows different scaling laws than twentieth-century industry.
On labor specifically, both views hold simultaneously. Crawford is right that AI currently depends on brutal forms of human work (100%), from mining to moderation. The contrarian is right that AI represents one of the few technologies that could eventually eliminate such work (60%)—though this remains speculative and doesn't address the transition period's human cost. The extraction critique splits similarly: Crawford correctly identifies AI's participation in extractive systems (100%), while the contrarian accurately notes these systems predate AI and constrain all industrial actors (80%). Neither view alone explains how transformation might occur.
The synthetic frame recognizes AI as a transitional technology—one that currently intensifies industrial modernity's worst features while potentially containing seeds of its transformation. Crawford's atlas provides the essential baseline documentation; any credible path forward must begin with her cartography. But the map itself suggests that AI's crime isn't unique evil but inherited complicity—a complicity that might be leveraged toward systemic change if coupled with the political will Crawford's documentation should generate. The materiality gradient runs from today's brutal reality toward a possible but not guaranteed transformation.
Further reading
- Kate Crawford, Atlas of AI (Yale, 2021)
- Kate Crawford and Trevor Paglen, Excavating AI (2019)
- Meredith Broussard, Artificial Unintelligence (2018)
- Safiya Umoja Noble, Algorithms of Oppression (2018)
- Ruha Benjamin, Race After Technology (2019)