Emerging Faces (2017)

In 2017, I had been working with robotic painting for twelve years. It was a very small field at the time, a niche that existed outside of the traditional art world. Over the years, I had begun to incorporate more and more machine learning techniques and computational intelligence into my processes, and the most recent additions then were neural networks, GANs (generative adversarial networks). What fascinated me most about GANs was that they function in a way that might very well have parallels to the functioning of the human imagination. The emergence of an image from a GAN feels like the emergence of a thought within one’s mind, a form that appears from an unknowable place.

I began to think a lot about what creativity actually means, and whether my machine system might have its own independent creativity. I wanted to test this out, so I trained a GAN that was able to recognize faces, combined this with a set of algorithms that evaluate aesthetic properties, and outfitted my robot with a camera to observe its strokes as it painted. I then created a set of conditions that removed myself from the process, instructing the system only when to initiate and terminate its work:

1. Imagine a human face and paint it on the canvas.
2. Stop as soon as you recognize a face.