One way of looking at the recent explosion of generative AI is to consider it as the logical end point of Post-Modernity. For many years we have been living in a world of intertextual remix. Everything is a reference to something else. Our culture seems to be a series of elaborate “in jokes” requiring intimate knowledge of innumerable references.
Deep Learning models rely on vast troves of training data – to extract the statistical regularities from a dataset requires many, many exemplars. GPT3 was trained on 10 trillion words, encompassing documents from all over the internet. The image generation model Stable Diffusion consumed 2.3 billion images to learn its skills.
Generative AI is automated post-modernity. Everything it produces is a synthesis of the training data – a blend of all it has ‘seen’ before.
I have touched on the subject of #manufacturedauthenticity a number of times. In some ways technology itself has been integral in many of the processes of recasting ideas into the skin of another mode, be it twitterbots attempting to be human, or pastiches of prior forms presented with the borrowed aura of nostalgia.
This booklet, consisting of 20 images paired with lyrical couplets, was created as a supporting artefact for the recent DRASS gig at the East Sussex Psychedelic Film Club event in Lewes. The images are AI generated Victorian Medical Illustrations, drawn from a Stable Diffusion model cross-trained on pornography.
While the intended use of the model is clearly for contemporary pornographic image generation, the underlying network retains the ability to form a wide range of far less salacious imagery. However, the model is inexorably tainted by the echoes of its Ludoviko-esque exposure to filth. The resulting images operate in an uncanny-valley world of historical plausibility, whilst triggering a touch of uncomfortable sexual transgression.
I have a few copies left over. Click below to buy one.