Flick_KA AI
2019
© Daniel Heiss; Photo © ZKM | Center for Art and Media
- Title
- Flick_KA AI
- Year
- 2019
- Category
- Computer-generated
- Installation
- Video
- Format
- Video Installation
- Material / Technique
- 4 flat screens, website, program: Generative Adversarial Network based on StyleGAN, programming language: Python 3.6
- Description
- “Who is born? Who is generated?”
Randomly selected portrait photographs appear on four flat screens. In an endless cycle, the viewer is repeatedly confronted with the question of whether the persons depicted were actually photographed or whether their portraits were generated by a computer program. To generate the artificial images, the artist used a machine learning technique that he applied to a photo collection of 60,000 self-portraits. This collection contains images of all ZKM visitors who have photographed themselves in the installation »FLICK_KA« by Peter Weibel and Matthias Gommel since 2007. On this extensive training database, a so-called “Generative Adversarial Network” (GAN) was trained, an artificial neural network that enables the generation of new artificial faces from the combined features of all the real persons. The viewers discover that they do not succeed with certainty in distinguishing real persons from artificially generated ones: a Turing test for faces that gives us an idea of how well computer programs can now deceive our perception.
Author
Daniel
Images
© Daniel Heiss; Photo © ZKM | Center for Art and Media