Overview

Consensus Portraits is a series of still images and videos that bring machine learning algorithms into conversation with one another in order to explore the reality these systems reproduce. These works explore that bias baked into machine learning systems meant to categorize people while also inviting a closer reading of these images and the digital material specificity/artifacting that belies the algorithms source material.

The first works in this series used images from This Person Does Not Exist which were animated using Deep Nostalgia. These synthetic moving portraits were then analyzed by the same facial recognition software I used in my Landmarks project. In this series, I was interested in the uncanny valley and the ability for algorithms to understand each other. The video at the top of the page is from that initial series.

The current in progress series is a set of still images created by asking the image generation software DALL·E 2 to create transwomen. These images were then fed into another algorithm tasked with identifying and categorizing faces. The resulting images are of overwhelmingly white figures whose bodies and surroundings belie the supposed neutrality of these systems. Three images from the series can be found below.