Reimagining AI Perception. Cybernetics, Extractivism, and the DIY Turn

Abstract

This article explores the perceptual epistemologies underlying contemporary artificial intelligence (AI), tracing their origins to cybernetic theories of perception, learning, and feedback. While generative AIs inherit a bottom-up model of behavior formation from cybernetics, their current implementation is largely shaped by extractive logics – both in terms of data and hardware infrastructure. These logics obscure the perceptual mechanisms, training datasets, and material embodiment of AI systems, leading to multiple “black boxes”. As a counterpoint, the article highlights alternative approaches found in DIY culture and performance art, such as Gordon Pask’s Musicolour, Luca Pagan’s Multi-Node Shell, and the community-driven AI IAQOS by Salvatore Iaconesi and Oriana Perisico. These projects demonstrate the possibility of ethical, situated, and relational models of machine perception and agency. Ultimately, the article advocates for a reimagining of AI as a co-constituted, materially grounded, and participatory technology, capable of fostering meaningful interactions between human and non-human agents.