The Algorithmic Unconscious

Abstract

The central hypothesis of this paper is that the concepts and methods of psychoanalysis can be applied to the study of AI and human/AI interaction. The paper connects three research fields: machine behavior approach, psychoanalysis and anthropology of science. In the “Machine behavior: research perspectives” section, I argue that the behavior of AI systems cannot be studied only in a logical-mathematical or engineering perspective. We need to study AI systems not merely as engineering artifacts, but as a class of social actors with particular behavioral patterns and ecology. Hence, AI behavior cannot be fully understood without human and social sciences. In the “Why an unconscious for AI? What this paper is about” section, I give some clarifications about the aims of the paper. In the “Unconscious and technology. Lacan and Latour” section, I introduce the central thesis. I propose a re-interpretation of Lacan’s psychoanalysis through Latour’s anthropology of sciences. The aim of this re-interpretation is to show that the concept of unconscious is not so far from technique and technology. In the “The difficulty of being an AI” section, I argue that AI is a new stage in the human identification process, namely, a new development of the unconscious identification. After the imaginary and symbolic registers, AI is the third register of identification. Therefore, AI extends the movement that is at work in the Lacanian interpretation of the mirror stage and Oedipus complex and which Latour’s reading helps us to clarify. From this point of view, I describe an AI system as a set of three contrasting forces: the human desire for identification, logic and machinery. In the “Miscomputation and information” section, I show how this interpretative model improves our understanding of AI.