DeepDream is a machine learning technique for amplifying patterns in images, revealing strange, dreamlike forms. I used it to build an art installation that dreamed about what it saw, presented as a tech demo at the Game Developers Conference.

The original DeepDream code went viral in 2015 after Google released it to the public. For a while, the internet was full of hallucinatory images, but I wasn’t interested in the memes. I was fascinated by how the images grew out of noise, like a Polaroid photo of hyperspace. What if those images could come from a live camera instead?

Audience reactions caught me off guard. People posed with strangers, mugged for the camera, and held their breath to see how far the dream might go. As a UX designer, I was fascinated by how easily participants ascribed moods and intentions to the system. That’s when I realized my machine wasn’t just hallucinating. I’d created an immersive interface for stillness and presence—a pause in a space where most people are busy thinking about what to say next.

What began as a curiosity became a full-scale installation, exhibited at such venues as the Stanford Linear Accelerator and the Palace of Fine Arts in San Francisco.