I have just returned from Berlin, where I visited an Ars Electronica exhibition devoted entirely to artists’ critical use of artificial intelligence.
Two works particularly caught my attention: The Normalizing Machine and VFRAME (Visual Forensics and Metadata Extraction).
The Normalizing Machine (2018), by Israeli artist Mushon Zer-Aviv , questions the algorithmic standardization methods of facial-recognition systems.
In this interactive installation, each participant is shown four previously recorded videos of other participants and asked to select the most normal-looking person among them.
The selected person is examined by algorithms that add their image to a database projected onto a wall reproducing the anthropometric charts of French criminologist Alphonse Bertillon, the father of the verbal portrait, which later provided a basis for eugenics.
It is astonishing to see our image broken down within seconds into measurements of eyes, mouth and ears and processed alongside hundreds of other participants.
The project addresses not only what society establishes as normal and how it does so, but how AI processes may amplify the discriminatory tendencies underpinned by old anthropometric theories centuries ago.
VFRAME: computer vision and human rights
The other project is VFRAME (2018), by Berlin-based US artist Adam Harvey. It is a set of computer-vision tools for human-rights researchers. In this project, Harvey is working with Syrian Archive, an organization dedicated to documenting war crimes and human-rights abuses.
The focus is identifying cluster bombs in videos recorded in war zones. These devices contain multiple explosives and are therefore defined as container weapons. One of Nazi Germany’s most horrific creations, they continue to be used in Middle Eastern wars. Their most intensive current use is in Syria.
VFRAME uses 3D modeling and digital fabrication combined with software that is still a prototype but includes tools capable of organizing, classifying and extracting metadata from 10 million video keyframes in under 25 milliseconds, identifying cluster bombs in those videos.
Its aim is to bridge the gap with state-of-the-art artificial intelligence used commercially, making it accessible and adapted to the needs of human-rights activists and investigative journalists committed to a future oriented toward peace and the protection of life.
Access the playlist with all the audio recordings of the Ouvir Imagens column broadcast on Rádio USP since March 2018 | Read all transcripts and posts.
Since September 2018, this column has also had a playlist on YouTube!
Visit and share it.
Image: VFRAME prototype for identifying cluster bombs, developed for Syrian Archive.