A.I. Interprets A.I. Interpreting ‘Against Interpretation’ (Sontag 1966)

2023, 3 channel video with sound, 25 minute loop

An AI is made to visually interpret Susan Sontag’s seminal essay ‘Against Interpretation’, and then another AI surreally interprets those images back into language.

Sontag writes in ‘Against Interpretation’ about her dislike of critics over-interpreting works of art, how we read too much into content and meaning over just experiencing the work of art and it’s form. In this video however we have an AI nonsensically reading too much into Sontag’s words, this also has additional prescience since the generative AI is arguably (uninterpretable,) creating pure mimesis and form since it is devoid of any human artist’s intentionality, meaning or content.

The visuals are created with an image generating diffusion model with Sontag’s sentences as its raw prompts / inputs (open-source Disco Diffusion thanks to Somnai & Katherine Crowson). These images are then interpreted back into language using an image labelling algorithm (GPT2 & CLIP). These large pre-trained AI models were created using huge datasets of images and text taken from the internet representing a frozen snapshot of a biased section the internet a particular point in time. The re-interpretations are bizarre in how authoritatively and brazenly they seem determined on spreading disinformation.