COLLECTOR’S CHOICE - Mistaken Identity by Mario Klingemann

Mistaken Identity by Mario Klingemann represents a complex exploration of neural networks through the deliberate manipulation of their internal structures. This work consists of three videos created using generative adversarial networks (GANs) and neural networks. Exhibited at the ZKM in Karlsruhe during the Beyond Festival in October 2018, the triptych investigates how neural networks interpret visual information.

Mario Klingemann, Mistaken Identity, 2018, a video triptych

Mario Klingemann is known for his pioneering work in generative and AI art, with preferred tools including neural networks, code, and algorithms. His artistic practice reflects a systematic approach and curiosity about understanding complex systems. Klingemann often dissects systems such as neural networks, analyzes their components, and reconstructs them to explore patterns and to recreate and understand the system’s behaviors.

Portrait of Mario Klingemann

The approach of deconstructing and reconstructing forms to understand underlying structures and patterns has historical precedents in art. Classical painters such as Leonardo da Vinci conducted anatomical dissections to gain a deeper understanding of the human form. Similarly, Cubist artists such as Pablo Picasso and Georges Braque fragmented objects into geometric components, breaking them down and reassembling the forms to depict multiple perspectives.

Mistaken Identity, 2018 at ZKM in Karlsruhe during the Beyond Festival in October 2018

To understand how AI functions and perceives human forms, Klingemann developed a method called "neural glitch”. This technique involves deliberately introducing errors into fully trained GAN models. Initially, GANs are trained to generate realistic human faces. Once the models achieve a near-perfect state, Klingemann disrupts key neural components.

Mario Klingemann, Mistaken Identity – Chapter #1, 2018

These disruptions include small changes, such as altering, deleting, or exchanging the training weights, which impact the numerical values that determine how the model synthesizes images. These changes result in significant and often unpredictable alterations to the output.

Mistaken Identity, 2018 at ZKM in Karlsruhe during the Beyond Festival in October 2018

They affect both texture and semantic levels, changing the arrangement of facial features and altering finer details, such as skin tone or shading. These changes produce outputs ranging from slightly altered portraits to entirely abstract forms. This process represents how neural networks work, interpreting and perceiving human faces differently than humans.

Mistaken Identity, 2018 at Future U exhibition at RMIT Gallery, Swanston Street, Melbourne, Australia in 2021

The final result consists of three nearly two-hour-long videos, presented as a triptych for the first time at ZKM in Karlsruhe in 2018 and the Future U exhibition at RMIT Gallery, Swanston Street, Melbourne, Australia in 2021

The first chapter of the series, Mistaken Identity - Chapter #1, is part of the Seedphrase collection.

Previous
Previous

HISTORY OF GENERATIVE ART - Generative Music

Next
Next

HISTORY OF GENERATIVE ART - Metaverse