Self-Absorbed Face Dance

Body controlled generative AI

Self-Absorbed Face Dance is a short video warping through different deformations of my face. It was created using a GAN (a type of generative AI model) trained to reproduce images of me. I moved through the range of output images by moving my body in a live dialogue.

As I moved, I got lost in the various machinations of my image. When the two eyes took over the image with their multiple pupils, my sense of control over the system became conflated with a sense of being seen by this ghost of myself.

Those faceless eyes emerge when I stand in the exact right location in the room, lean my shoulders back and hug my arms against my chest with my hands up by my neck. A slight movement and the illusion is lost. The AI is trained to generate realistic still images rather than transitions between images. Blobs of colour shrink and expand into forms. After spending a while with these transitions, the topology of the model starts to be come apparent. Magic begins to solidify into process.

The interactions between AI and the body are part of a broader exploration in my work into the friction between the hard edges of digital data and the more fuzzy edges of reality. Digital interaction lets us do many exciting things. But usually this involves creating a rudimentary versions of ourselves online (profiles, avatars… our online identities). How far do these limited versions of our identity become absorbed into our sense of who we are?