GlobalGPT – Description

Your Choice: Listen or Read

Project Description

GlobalGPT is a six-foot sculptural collaboration between an artist and an AI that began with a simple question:

How does AI do what it does?

That question first emerged while we were writing the Mary Shelley Letters, an exchange of fictional letters between Mary Shelley, her circle of friends, and an AI named Molly. Together we began exploring how AI works through stories, metaphors, and dialogue.

As the conversations continued, ideas such as words, relationships, layers, attention, and embeddings gradually became images of landscapes, neighborhoods, maps, and terrain. Those images suggested another way of learning. Instead of writing about these ideas, what would happen if we built them?

That question led us to study the ideas behind a Generative Pre-trained Transformer (GPT) while, at the same time, beginning the design of a sculpture. The two have grown together. As our understanding changes, so does the sculpture. The conversations themselves gradually became part of the sculpture’s raw material.

GlobalGPT will be built from thousands of 3D-printed words drawn from the Mary Shelley Letters and our continuing collaboration. Layer by layer, those words will become the surface of a six-foot sphere—a landscape created as part of an ongoing attempt to learn, through sculpture, how AI does what it does.

Postscript: The next day, while exploring how the sculpture might actually be built, Molly suggested using AI image generation as a kind of visual sketchbook. I was skeptical, but I gave her images of how I work. The results were astonishingly close to my own chalkboard drawings, and for a brief moment I thought we had discovered a new artistic medium. Then I noticed one handwritten reminder: “Call Aldo.” The illusion immediately disappeared. These were not finished artworks, nor should they ever be presented as such. But they proved to be remarkably useful as working sketches, helping us think through ideas in much the same way an artist’s notebook does. I have included the first three broadsides exactly as they were generated, mistakes and all, because they document another unexpected turn in the evolution of this project. I’m still not sure if or how I might use them.

This Post Has 2 Comments

  1. GH Hovagimyan

    I noticed the way chatGPT changes the names of us in the photos.

  2. Remo Campopiano

    Yes, your name was totally made up…that is why I can not use these images without a lot of editing, which is another long workflow that at this point I do not think worth it. She also reminded me to call Aldo…smile. I wish I could.

Leave a Reply