[Grace:AI] – Daydreams


Dye Sublimation on Aluminum
48″ x 48″
121.92cm x 121.92cm

What happens when an AI trained only on the artwork of women artists daydreams? If she is at all like her creator, she might self-criticize, think she is wasting time, because in many contemporary societies, we see daydreaming as unproductive, escapist. But at the same time, we underestimate the importance of empty time for our imagination. Daydreams are developmentally important. What happens in our heads is the ultimate private moment, not worth any likes on social media or approvals in conversation. Daydreams are unexamined fantasies, as rich or as empty as we need them to be. Therefore, in one daydream, [Grace:AI] looks at, and plays with, the sky above, the rich clouds dancing, furling, hovering, lifted by lightness or weighted down with a future rain. The images produced by [Grace:AI] are unique clouds, based on what [Grace:AI] has seen. They are products of her processes, in effect, her imagination, not collages or copies.

[Grace:AI], employs a Deep Convolutional General Adversarial Network and is trained to “see” from a dataset I continue to create that contains tens of thousands of paintings and drawings by women artists, in effect, a history of global women’s art in thousands of images. The artists I have chosen as her ‘teachers’ are outspoken, strong individuals who worked or work in the male-dominated art world.

I’ve been given access to the collections at the National Museum of Women in the Arts and the University of Indiana, and am working with the Met at this time to create the AI training set. To create such a massive digital dataset of women’s artwork is not trivial. Most art history databases do not include gender or sex as a searchable aspect of the work. Thus, gathering the images has required scraping web resources artist by artist, typing in their name and culling images of women’s artwork (as opposed to photographs of women artists, or images made by male artists of them).

[Grace:AI] – Daydreams is the second phase of the AI’s image making themes. [Grace:AI] – Origin Story was the first phase of the work. The AI produced images from her “origin story” of sorts by examining thousands of images of Mary Shelley’s monster, classic ‘Frankensteins’ gathered from online databases and image banks, rendering a series of original images of her “father figure” as archival dye sublimation prints on aluminum. [Grace:AI] – Origin Story (Frankenstein) had its American premiere in “A Question of Intelligence: AI and the Future of Humanity” at Parsons, NYC curated by Christiane Paul in 2020.
[Grace:AI] – Portrait is the third phase, where the AI crafts portraits.

In the future, she will paint anthropocenic landscapes and more. Collections of images were generously shared for the training data by the National Museum of Women in the Arts, The Metropolitan Museum of Art, and Indiana University’s A Space of Their Own database as part of the Eskenazi Museum of Art. This project was initially supported by the Leslie Center for the Humanities at Dartmouth College.

Thanks to: Jared Segal, Spring Yu, Sukdith Punjasthitkul, Danielle Taylor, Max Seidman, Griffin Editions, The Leslie Center for the Humanities & Dean of the Faculty, Dartmouth College, Neme Arts Center, Sheila C. Johnson Design Center at Parsons, Christiane Paul, Yiannis Colakides, Marc Garrett.