Kate Crawford’s Atlas of AI advances a decisive critique of artificial intelligence: AI is neither immaterial nor autonomous, but a planetary infrastructure of extraction sustained by minerals, energy, human labour, data capture, military agendas, and corporate power. Beginning with the story of Clever Hans, Crawford dismantles the fantasy that intelligence can be read from performance alone, showing how apparent cognition often depends on hidden cues, training conditions, institutional validation, and collective desire. This becomes the organising metaphor for AI itself: systems praised as intelligent are produced through vast material and social arrangements that are usually concealed. Crawford therefore rejects the narrow technical definition of AI and reframes it as a political-economic formation, a “registry of power” that optimises dominant interests while presenting itself as neutral computation. Her atlas method is crucial: rather than opening one “black box”, she maps interconnected terrains—lithium mines, data centres, warehouses, image datasets, classificatory systems, affect-recognition tools, and military infrastructures. A specific case is Project Maven, where Google’s machine-learning capacity was recruited for drone-video analysis, revealing the intimate relation between AI, warfare, object detection, and state violence. The conclusion is severe but generative: artificial intelligence must be judged not by abstract promises of efficiency, but by its material costs, its classificatory harms, and its consolidation of asymmetric power. To contest AI requires linking data protection, labour rights, climate justice, racial equity, and democratic limits on technological domination.
