Latent Spaces: What AI Art Can Tell Us About Aesthetic Experience

In her book L’art victime de l’estétique, Carole Talon-Hugon criticizes the identification that has been made between art and aesthetics since the Eighteenth century. In this context, what the author calls “aesthetics” or “aesthetic experience” is essentially linked to the sphere of sensitivity, which becomes more and more isolated from that of good and truth. Art – this is the author’s thesis – progressively loses its link to good and truth as it becomes something that must just affect the senses. In this context, the author recalls that a central role was played by the growing importance of colour, in a process that led from Impressionism to abstract art. Throughout this evolution, according to Talon-Hugon, the intention of painting increasingly became an attempt to restitute “the purely visible impression,” isolated from any judgement or identification. In this paper, this thesis will be confronted with a new type of art, the art made through artificial intelligence. In particular, the focus will be on some artworks made through a kind of AI called GANs (generative adversarial networks). If one considers these pictures, one may notice that they share with some early Twentieth century paintings the fact that colour becomes somewhat independent from the outlines and drawing. Outlines are very blurred and the identities of the objects are not certain. Nevertheless, this type of colour in GANs pictures doesn’t demonstrate a denial of the task of identifying things. On the contrary, it is precisely concerned with the machine’s very attempt to classify objects. Moreover, even if we recognize in the AI our own attempts to classify and comprehend the world (the Ai as an “uncanny mirror” of ourselves), the AI remains nevertheless something other than ourselves. Therefore, the aesthetic pleasure when faced with these pictures implies the expectation not only of a possible knowledge, but also of a new relationship with an other. This will allow some considerations on the very notion of aesthetics itself.