Completion 1.0

Completion 1.0 is an installation that generates, thanks to an artificial intelligence, or recursive neural networks, an infinite, alternative and counterfactual history of photography. The installation is composed of two LED screens, a 3m3 modular aluminum structure and a sound source. A screen is placed on the floor and broadcasts the entire image-net, a dataset of more than 14 million images. It is a large part of the known world that scrolls for 7 years, a duration that makes it impossible to perceive it completely. The second screen is fixed in the aluminum structure and shows images generated automatically from the statistical cross-referencing of image-net. The images are surreal metamorphoses between heterogeneous ontological categories: plants, animals, techniques, minerals, etc., seeming to describe a world where things are always unstable, at the edge of their own identity and mutate in a continuous flow. They are realistic without being real. A synthetic voice describes each of these images, drawing on thousands of works from the history of photography, and seems to mimic the speech of an art critic. We listen to this narrative that goes off the rails, that seems to mock with irony the texts that accompany the works (label, text, etc.), that describes an imaginary off-field and that sometimes appears to be more articulate and relevant than certain analyses produced by human beings, sending back to us a troubling reflection of our cognitive and perceptive faculties. The software produces images that it tries to describe, the images being the residual memory of past photographs and the word, the distant memory of the aesthetic discourse of the last century. Gregory Chatonsky continues his investigation of the aesthetic, political and historiographic consequences of AI. From a metric architecture that forms his signature and that allows to assemble the whole of his past and future works, he proposes a manifesto work of a new inductive realism. If photography has largely determined our way of conceiving and perceiving reality since the industrial revolution, recursive neural networks allow, thanks to statistical induction, to produce new media from existing media. Feeding on photorealism, AI produces a disturbing counter-realism that seems to respond to the current tremors of truth (conspiracy, fake news, pseudosciences). This automation of resemblance in the context of a hypermnesia that has overwhelmed us for 30 years on the Web, forms the heart of the Complétion metaproject of which this is the first iteration. It is about repairing, completing, predicting, generating the whole of human history by proposing indefinite alternative versions of it.