Pictorial Forms in Motion. AI Embodiment, Hallucinations and the Painting

Valentine Bernasconi, Ph.D.

Large Language models (LLMs) as well as generative models encapsualte cultural information presents in their training dataset. They can be used to describe and contextualize artworks. Additionally, digital images can be easily manipulated and transformed, blurring their original still and mute essence. Depicted characters can be animated to talk about the artwork where they are found, thus embodying and giving a voice to LLMs. Such artistic digretions of artworks become particularly interesting in confrontation with art historical practices as they are the receptacle of new interpretations, and question contemporary reading of artworks.

Following similiar artistic, computational and art historical research, the deepennig of the question of AI embodiement would expand my research interests on the discourse of the image and embodied experiences. Over time, the concept of the body has been redefined, along with its associative components and means of expression. As explained by Belting, every new iteration on the definition of the body requires a form of visualization: “Bodies are strongly shaped by their cultural history and thus never cease to be exposed to mediation via their visual environment.” Similarly to the image, the body is an "embodying of possibilities both conditioned and circumscribed by historical convention". The body brings its experience in the creation of art, but also in the reading of it. Generative AI for the creation of images raises the question of the absence of the body in the production of art, as well as encapsulated perceptual constraints in historical representations. Based on the idea that gender is socially constructed, as well as the concept of generous imitations, we can therefore address and confront contemporary representations of the gender with past ones. Main research questions would be revolving around gendered art production, whether embodied experiences are inherent to the creation of art, and the impact of historical contexts on their perception. How could we use AIs image production as clues towards the definition of the AIs body and the contemporary construction of gender? What do these images communicate to the spectators and how do they transform their experience of the world in relation to their gender? How can we read these expressions to educate and provide critical perspectives in our approach to art and gendered experiences?