Spurious Memories
One intriguing characteristic of the human mind is its ability to be creative, that is the ability to generate an output that is not explicitly learned; to evolve memory through synthesis and noise, arriving at ideas and solutions that seem to come out of nowhere, that appear to be completely new.
Creativity, like consciousness or intelligence, is a fundamentally social and perceptual phenomenon. There is no list of ingredients, secret formula, or divine synthesis that gives rise to any of these qualities. A person, an animal, or a machine, is creative if we deem them to be so, if our culture has shaped us to believe it. Nonetheless, the structure of the biological brain and the process by which it evolved lends it the capacity for these qualities and the perception of them.
This project is concerned with the possibility of creating creativity. It proposes that by constructing electronic structures using processes and materials inspired by biology we can enable a creative capacity in machines.
Through utilization of principal component analysis and neural network techniques I have developed an autonomous face categorizing and generating software program. Through exposure to facial images over time the network develops a sense of what faces are, what components they are composed of, what makes them similar and different. It is then able to extrapolate from experience to create new facial images which embody its knowledge.
The software, written entirely using the open source language Python, appropriates algorithms traditionally used for surveillance and military purposes and gives them a new life as tools of imagination.
This project has two modes, an awake, interpretive mode and a reflective dream mode. In waking mode the software analyzes an input image and outputs an image from memory or imagination that is the closest it can conjure to the input image. This allows a facial recognition capacity or a facial imagining capacity, depending on input. When given ambiguous input images such as ink blots and clouds interesting and unexpected interpretations emerge. The underlying structure of the code is a neural network of vast connections, and the dream mode of the software highlights this form by traversing the network of imagined facial possibilities in a method akin to dreaming. The code begins with a self-generated random input image similar to the random neuronal firings which occur in the human brain prior to dreaming. This random input creates an image pattern which then inspires a traversal of neuronal links, a stream of consciousness dream state which is output for us to see as a sequence of images.
Like a living creature, each instantiation of the code is unique; though each may experience similar information, they apprehend it and remember it in their own anomalous way. Following from this, they interpret new and ambiguous stimuli divergently making each installation different from any previous. This project examines the core possibility of how it is that human beings come up with ideas which are new to them, and how this capability can be translated to the realm of machines.
Artist name:
Heather Dewey-Hagborg
URL: http://deweyhagborg.com/spurious/