GlitchesAreLikeWildAnimalsInLatentSpace!—BOVINE

GearBox Gallery is pleased to present this year’s winning Inner Room Call for Entries proposal, GlitchesAreLikeWildAnimalsInLatentSpace!—BOVINE, by Karin and Shane Denson.

Inspired in equal parts by glitch-art vernaculars, the chronophotography of Eadweard Muybridge and Étienne-Jules Marey, the cut-up methods of Brion Gysin and William Burroughs, and generative practices from Oulipo to Brian Eno and beyond, our ongoing series GlitchesAreLikeWildAnimalsInLatentSpace! stages an encounter between human imagination and automated image-making.

We begin by training AI models on a set of Karin’s glitch paintings, which are produced by subjecting wildlife videos (all shot by Karin) to aleatoric “databending” processes. The digital videos break and reveal their underlying protocols and infrastructures in the form of glitches and compression artifacts. Selected images are then painted by hand in acrylic on canvas. Fed back into the machine, deep learning models are prompted to generate new images in the style of Karin’s glitch paintings. But since contemporary (“diffusion”-based) AI models are effectively trained to eliminate “noise” and glitches, we are pushing them against their intended purpose—and the results are accordingly unpredictable.

Having obtained a new image, Karin translates it back onto canvas. In this case, we are working with a cow that the AI surprisingly, and humorously, generated in response to the prompt “glitches are like wild animals.” Shane then takes an image of the new painting as the “seed” for an AI-generated video. Produced in short bursts of animation and strung together, the video is then subjected to the original glitch processes that are responsible for its training data.

Because the video is broken, it plays differently in different software. When the various versions are combined, they fall out of phase with one another, producing a ghostly effect. The soundtrack is also produced through generative methods by computationally (mis)interpreting a jpeg of the original painting as an audio file. Custom software cuts and combines the video in real time and draws on Shane’s book Discorrelated Images to generate novel sentences.

Finally, the generative video serves as a kind of latent space for the production of new painted images. In this iteration of the project, Karin has chosen five frames from the video and painted a slice of each one as a panel of a pentaptych, thus capturing the video’s temporal progression in a spatial form that is equal parts digital chrono(post)photography, automated cutup, and human resistance to automation.

Through this recursive and collaborative process, which shuttles repeatedly between the digital and the physical, the invisible and the perceptible, we hope to open for viewers a space of glitchy imagination, if not insight, into the opaque black-boxed operations of the algorithmic media that are rapidly reshaping our visual cultures and environments.