The Fallacy of Autonomous AI

Abstract

In the mainstream media, concerns are voiced about the potency of AI as a threat to humanity. Some of the academic literature that gives credence to that threat, does so in reference to posthumanism, where we find, besides conceptional tools for thinking technology in an indeterministic way, also an appeasement towards strong AI. This paper’s aim is to demonstrate that a possible threat by AI does not come from the alleged attainability of autonomous, cognising machines. Starting out from the conception of technology as a socio-material arrangement, if becomes clear that AI, like any other technology, is socially performed. What is more, supposedly autonomous machines run on algorithmic, linguistic, written code, as I show in an analysis of computer language as Derridean writing. As such, they are extensions of human cognition. To proclaim machines conscient and autonomous, is hence not just misleading per se, but disguises the human agency that uses the AI-autonomy as a proxy.