Trevor Paglen

Trevor Paglen (born 1974) is an American artist and geographer, winner of a 2017 MacArthur “genius” fellowship, known for photographing classified military sites, launching a satellite as sculpture, and diving to undersea internet cables. Since the mid-2010s his central subject has been machine vision: what images look like when they are made by and for machines, and who gets hurt in the process.

ImageNet Roulette

In September 2019 Paglen and AI researcher Kate Crawford released ImageNet Roulette, a viral web app that labeled users’ selfies using the “person” categories of ImageNet — the canonical training dataset behind modern computer vision. The results were often racist, misogynistic, and absurd, by design: the project let the training set “speak for itself.” Within days, ImageNet’s maintainers announced they would remove more than 600,000 images of people from the database. The pair’s companion essay “Excavating AI” and their Fondazione Prada exhibition Training Humans made the case that such taxonomies are politics masquerading as engineering.

Paglen has continued the excavation in works like Faces of ImageNet and CLOUD, and in warnings that generative models inherit the same buried biases. His stance is investigative rather than prohibitionist — he uses machine-learning tools constantly in the studio — but the through-line of his career is that AI systems encode power, surveillance, and classification errors that fall hardest on the already marginalized.