Robobotanist and AI artist Demon Sanctuary’s (David Atlas) latest piece is a poetic, introspective interpretation of the soaring orchestral track ‘The End’. “Can we create self-sustaining machines?” wonders Atlas. “First we must make ourselves self-sustaining. To me this song touches heavily on this. It has a hopeful hopelessness about it.”
His film is made using a form of AI known as a Generative Adversarial Network, or GAN, comprised of two neural networks competing against each other and learning to generate new data from a training set. Atlas has trained his GAN on a visual storytelling dataset full of photographs of our natural world, resulting in these computer-generated morphing forms that look realistic but also a little off; that have an otherworldly uncanniness about them. “Essentially,” Atlas says, he’s “trying to trick a GAN into being creative.”
The resulting film is an orgy of conditional forms: of budding shapes bursting forth, hyperreal textures melting in and out of the screen and decomposing silhouettes in Earthy tones. As the song nears its crescendo, the screen splits in two and flowers bloom on both sides:
“Two flowers dance together but can never really meet after a storm; perhaps during which they had touched for a number of brief, almost violent moments.”
Atlas has designed an artificially conscious artist and taught it to dream up new lifeforms. To the sound of swelling strings, he shows us machine-made visions of things no person has ever seen before.
Furthermore, Atlas has also trained his GAN on a dataset of images of sea slugs and generated a suite of hyperbolic renderings of 3D “Ganflowers”, which he’s combined into a quadruped exclusively for The 1975: this can be viewed as a 3D object, and downloaded for 3D printing, on the exhibition microsite that opens on 6th June, 2020.
Demon Sanctuary responds to ‘The End’