Fixation
Science on Screen: Shorts on Attention, Museum of the Moving
Image
Fixation, University of California, Los Angeles
2024
Hollywood rendering pipelines are increasingly used to create realistic images for commercial computer vision datasets. In Fixation, an imaging program developed by Meta to train eye-tracking algorithms relapses back into human cinema. The length of the film corresponds to the number of images necessary to train a model that outperforms models trained on real images; the audio track was scored using computer vision.
What Comes After Z?
2024
A freelance Ukrainian photographer narrates the movement of images through the circuits of war. As his pictures move into the stock of lighting environments that train computer vision systems, he reflects on the inadequacy of language to the sensuousness of the world. Told through 3D rendered subtitles, his story adopts the training data’s own garb.
[2.8, 9, 5, ∞]
2024
A camera peers from the window of a server farm as a tree blows across its view, interrupting its autofocusing algorithm.