Skip to Content
Three square images are hung on a white wall. The images are computer generated images. The first has a sepia overtone and looks like the torso of a woman who is folding her arms across her abdomen. The second has black and white tones and looks like a glitched top hat with teeth near the bottom. The third is in color and looks like the face of an older white woman who is smiling. Each of the images have computer generated deformities.

A ______ Woman Smiling

Stephanie Dinkins

dye sublimation prints on metal, 2021

These images were created using a text to image General Adversarial Network (GAN) and are a result of the GAN process. The caption below each image is the textual description used to form the image. The GANs rely on data that is widely used and available. Dinkins started this experiment using the phrase “A Black woman crying.” She wanted to see if a Machine Learning (ML) system could use the images and corresponding tags it had at its disposal to reproduce the textual idea in image form. In late 2019, the artist found that the ML system had difficulty accurately producing the phrase as a recognizable image. The result created by the text to image GAN featured a pink figure in a black cloak abstraction.