Nietzschean Language Models and Philosophical Chatbots: Outline of a Critique of AI

Abstract

Developers of the deep learning algorithms known as large language models (LLMs) sometimes give the impression that they are producing a likeness to the human brain: data-processing ‘neural networks’ are ‘taught’ to recognize patterns in language and then, based on this pattern recognition, create or generate new content in the form of natural, humanlike speech, writing, images, etc. The results have been unsettling to some; less appreciated are the metaphysical assumptions underlying the attribution of any meaningful agency whatsoever to an algorithm. In this essay, Nietzsche’s thoughts on the “seduction of grammar” form the basis of one possible critique of generative AI – a critique, moreover, which exposes our society’s current fixation with LLMs for what it is: a fetishization and humanization of new technologies.