ImageNet Roulette was part of a broader project to draw attention to the things that can – and regularly do – go wrong when artificial intelligence models are trained on problematic training data. ImageNet Roulette is trained on the “person” categories from a dataset called ImageNet (developed at Princeton and Stanford Universities in 2009), oneContinue reading “ImageNet Roulette”
Tag Archives: Facial Recognition
Kate Crawford and I were invited by Prada Mode to create an exhibition and cultural program at the iconic Maxim’s restaurant in Paris. We transformed the space into a story about the history of facial analysis and a reminder of the dark histories from which contemporary facial recognition systems emerge. Pages from 19th Century phrenologyContinue reading “Making Faces”
From ‘Apple’ to ‘Anomaly’ (Pictures and Labels)
For a 2019 commission in the Barbican’s Curve Gallery in London, I took a close look at the most widely-used “training set” used in AI – ImageNet, a database of over 14 million images organized into more than twenty-thousand categories. The installation was made out of approximately 30,000 individually printed photographs, showing the precarious relationshipsContinue reading “From ‘Apple’ to ‘Anomaly’ (Pictures and Labels)”
They Took the Faces…
Contemporary facial-recognition algorithms were first properly researched in the early 1990s. To conduct that research, computer scientists and software engineers need large collections of faces to experiment with and to use as performance benchmarks. Before the advent of social media, a common source of faces for this research and development came from mugshots of accusedContinue reading “They Took the Faces…”
I conceptualized this exhibition with Kate Crawford to tell a story about the history of images used to ‘recognize’ humans in computer vision and AI systems. We weren’t interested in either the hyped, marketing version of AI nor the tales of dystopian robot futures. We wanted to engage with the materiality of AI, and toContinue reading “Training Humans”
Behold these Glorious Times!
This video installation is composed of images from two sources. The photographic images in the video are parts of training libraries used to teach artificial intelligence networks how to recognize objects, faces, gestures, relationships, emotions, and much more. They are images designed to teach machines “how to see.” The second kind of images in thisContinue reading “Behold these Glorious Times!”
A standard technique in facial recognition software is to use an algorithm to create a “faceprint” of a given person and to use that faceprint to try and match a person’s face with photos. To grossly oversimplify, if you want to teach an algorithm how to distinguish a particular person (say Fanon) from a collectionContinue reading “Fanon”
Machine-Readable Hito & Holly
These two pieces are made out of hundreds of portraits of artist Hito Steyerl and sound artist and composer Holly Herndon that have been analyzed by various facial-analysis algorithms. Below each picture is the output of algorithms attempting to detect their age, gender, and emotional state. Other algorithms attempt to determine whether they are wearingContinue reading “Machine-Readable Hito & Holly”
It Began as a Military Experiment
Contemporary research into facial recognition technology began in earnest in the mid-1990s at the behest of DARPA, the Defense Advanced Research Projects Agency. The military wanted facial recognition to exist, so DARPA began funding researchers in computer science and computer vision to work on the problem. The military realized that to do facial recognition, researchersContinue reading “It Began as a Military Experiment”
This is an article that I co-authored with my friend and collaborator Kate Crawford, who directs the AI Now Institute at NYU. In the article we take a look at some of the bad assumptions and bad politics built into the architecture of the training data used in AI systems.