Heat, Colonialism and the Geography of Artificial Intelligence

by Larry Lohmann

first published 20 August 2021

This draft article for the journal Digital Geography & Society tries to place the energy-intensive project of mechanizing interpretive labour known as artificial intelligence (AI) in the context of the longer trajectory of post-18th century industrialization and the capitalist appropriation of human and nonhuman work. Using the mechanization of recognition as a case study, it proposes viewing AI as a set of entropic, globally-distributed machines for labour exploitation linked to particular geographic patterns of colonialism and ecological fatigue. In particular, it examines the contradictory relations through which the ‘dead labour’ crystallized in recognition machine networks must constantly recruit and degrade enormous quantities both of the ‘living labour’ of human recognition and of thermodynamic energy to perform its repetitive predictive tasks. As with the 19th-century industrial revolution, the article argues, the geography of this process is better understood by treating energy not as a singular resource extractable from a variety of sources and then used up at various points of an essentially invariant landscape, but as the political reorganization of entropy gradients and exchanges across the borders of nonequilibrium systems.

The article’s purpose is to encourage activists and researchers to have more confidence in challenging the assumptions through which AI is usually defined and to provide a new perspective on the ‘digital natures’ it produces.