AI Imaginaries: Generative networks
I used a commonly available generative adversarial network architecture to create a series of digital artworks. Keyword prompts, which the system uses to generate images, were selected using a random word generator.
This project aims to surface social and technical contingencies surrounding notions of 'randomness' and 'creativity' with respect to generative adversarial networks.
About
The prompt for this image was "Dune". The final image resembles an abstracted sand dune, with a black portion in the background evoking the creatures from Frank Herbert's 1965 novel.
About
The prompt for this image was "Newspaper". Resembling the Ukiyo-e style of Japanese art which flourished from the 17th through 19th centuries, the network has generated symbols that stand-in for logographic characters.
%20copy%205.jpg)
The prompt for this image was "Dune". The final image resembles an abstracted sand dune, with a black portion in the background evoking the creatures from Frank Herbert's 1965 novel.
The prompt for this image was "Strikebreaker". The final image depicts a mass of signage in an industrial setting demarked by railing and machinery.
.jpg)
The prompt for this image was "Fox". This final output can be seen as a commentary on common classifier systems: we can see fur, a snout, and legs – but certainly not a fox!
.jpg)
The prompt for this image was "Implication". This image is perhaps the most difficult to distinguish from the set, reflecting what might be representative of the network's inability to identify reference material related to abstract nouns.
%20copy.jpg)
The prompt for this image was "Shadow". The output includes a light source at the top of the frame, which casts the shadow of a figure across a background reminiscent of stone or a cobbled floor.
.jpg)