Sustainable AI : dreams of neutrality

Abstract

The discourse surrounding "sustainable AI" is shaped by dominant visual and linguistic metaphors that frame AI as a tool for controlling and optimizing ecological systems. While AI applications like DeepMind’s GenCast or IBM’s Green Horizon are promoted as solutions to environmental crises, they often obscure the technology’s resource costs and reinforce a technocratic vision of sustainability. By depicting Earth as a digital system that can be monitored and managed, these narratives risk oversimplifying ecological complexity, deepening the divide between humanity and nature, and fostering misplaced confidence in technological mastery.