MODEL COLLAPSE

 

Model Collapse is a multi-facetted video installation developed in collaboration with Daans Heemskerk (Creative Programmer).

By programming new technology in facial recognition, detection and Machine learning, this installation fully autonomously creates DeepFakes of the spectators of the exhibition and places their appearance into highly personal VHS tapes from the early 90’s. Their presence in the exhibition effectively replaces the family members from my childhood. Simultaneously, all original material on the tapes is rerecorded, forever altering the only existing documentation of these past events. Model Collapse shows the effects of mass-surveillance and offers a glimpse into how any video images could just as easily be manipulated to be used as fake information.
In an ongoing loop, each time the video is displayed, new people are introduced into the VHS documentation, creating an ever changing selection of personal memories, that you, whether you wanted to or not, now became a part of.

Installation overview (Rijksmuseum Twenthe, The Netherlands)

Special thanks to:
Bram van Es (AI-Engineer)
Nasir Memom (Prof. NYU Tandon School of Engineering, Cyber Security Dept.)
Deborah Meibergen
Theo Blom

Generously supported by: