I have long held a fascination with #manufacturedauthenticity. Authenticity is what we seek out in art (and across our lives, more generally). Many of my previous works have investigated how machine generated interventions can appear as ‘authentic’ human action, if you inject them into the right places (and squint a little).
Gen2 is a video generation model from Runway which produces short snippets of video from a single image. For example, give it a picture of a ballet dancer, and it will (attempt) to make them dance. The technology uses its ‘knowledge’ of how things move to apply movement to novel images – in a similar way that still image diffusion models, like stable diffusion or midjourney, use their ‘knowledge’ of images to follow a text prompt.
I have always had a somewhat ambivalent relationship with the ‘art’ of text-driven generation. Obviously there is some agency in the crafting of a ‘good prompt’, but the relationship between the words and the output is still quite arcane and unpredictable.
For me, it is this very unpredictability which is exciting – I rarely use a text2image system looking for an exact and accurate depiction – it’s far more exciting to investigate the less-travelled areas of latent space in search of unexpected visual delights.
In this scenario, the Artists’ role is as curator, or tastemaker, rather than craftsman. There’s nothing wrong with that, but it is most certainly different than painting a picture by hand.
I am old enough to remember the rise of the DJ as celebrity – and feeling similarly bemused that the mere act of curating other people’s work could be considered an art. In retrospect, it seems inevitable. Not so much ‘seizing the means of production’ as ‘seizing the means of selection’.
Which brings us to this piece, Mon subconsient est baise (My subconscious is fucked): it comprises three short animated films, created entirely from the ‘mind’ of #gen2. Each began with an image generated by Stable Diffusion and thus my personal agency in their creation was minimal, beyond selecting the most interesting results. Stylistically, they remind me of the experimental animations I used to watch on Channel 4 as a child – animation as art, rather than kids cartoons. Experiencing them at the time, I remember the delightful cognitive dissonance of seeing such a familiar media used in such unusual and thought provoking ways.
As an homage to this, I created a manufactured historical framing for the film – lo-fi French language titles, the addition of film noise and an avant-garde soundtrack – to take the work away from it’s (very) 2023 origins and (re)place it into the past.
In doing so, the work appropriates a sense of authenticity from the viewer. Films that look like this were made in the past, therefore this must be from the past, and so it feels more ‘authentic’ than the more obviously signified contemporary video.
We are (potentially) at the apex of post-modernity, both culturally and technologically: generative text, image and video systems are beginning to inject fictions into our narratives. And as the models harvest more and more training data, they inevitably begin to ingest these fictions and poison their nascent grasp of truth.