MoMA 2077
A speculative glimpse into the future of art where we’ve surrendered to Generative AI
MoMA 2077 presents a speculative glimpse into the future of art in a dystopian world where we’ve surrendered to Generative AI, abandoning the human touch in artistic expression.
MoMA Archival Data
For the archive data, I scraped the MAID (MoMA Archives Image Database). I spent around an hour downloading ~50,000 images from 1924 to 2024. This includes artwork, installation spaces, documentation, and events. In addition, each image was provided with a detailed caption/description.
Set dressing was used to make the installation to feel more tangible, but also highlight how easily it was to “steal” the work of artists, collections, and museums without permission.
MoMA 2077 model
Using fast-stable-diffusion and DreamBooth, I fine-tuned Stable Diffusion v1-5 into custom text-to-image models. This resulted in 15 models: 10 models (the 1920-2020 decades) fine-tuned on the MoMA collection and 5 models (the 2030-2077 decades) fine-tuned on the decade model before. Since the MAID Archive had photos of the history of the Museum, alongside the artwork inside, the model also generated images representing the a pseudo-history of MoMA.
Installation Development
The installation was broken into two parts: one being a way to view all 50k images of the MAID archive, and the other to view AI trash generated off of this data. Individuals could come up to the projected installation and toggle the Archive against AI-generated work, and use timeline controls to move throughout the years of MoMA and MoMA 2077.
Everything was assembled in TouchDesigner, where I took the two major archives and tied them together via switches, buttons, and timers. With a complex timer loop system, the shown four images to increment every four seconds; allowing for the installation to never be still, even when individuals weren’t interacting with it.
