Patrick Lichty

=Personal Taxonomies is a multi-year project exploring my attempts to use machine learning and GANs to investigate Chomsky's notion of Deep Structure, drawing on a large body of my own work to find the internal consistencies in how my mind operates. According to Chomsky, language arises from patterns rooted in the physiology of the human brain—a kind of innate predisposition to language. After encountering the work of neurologists seeking to reconstruct what a person sees by analyzing brain scans, I wanted to attempt something similar using machine learning systems and a large collection of my own paintings. If a machine learning system can identify the inner consistencies, or "patterns," in my mind by analyzing many examples of my work, perhaps I can cross-compare two sets of these AI-based analyses to reveal the "deep structures" of my brain as it creates.

Process

Over nearly two years, I developed an asemic (meaning-free) calligraphy style informed by my experience with Japanese and Persian writing. The work builds on Machine Drawing and Alien Calligraphy, a series of 40 asemic, calligraphy-based paintings I made at the Virginia Center for the Creative Arts in 2007.

I created these calligraphies on my iPad and selected 512 favorites. I divided them into two sets of 256 images and fed each set into Playform.io's freeform morph GAN, iterating 512 times. From each result, I took 256 images from the first output of each set and fed them into a freeform morph GAN again—selecting identical results from each set was essential, as any deviation would break the consistency. Finally, these images were cross-compared in the GAN, and the output was examined.

Questions

Does this methodology truly reveal my mind's Deep Structure? Does the project expose genuine patterns in my cognition? Could other methods yield better results? And what are the aesthetics of the work itself?