My image-making process—supposedly an internal and personal voyage—questions how much of creative intention is constituted through the unperceived external mediation of experience. Is the artist, as Kurt Vonnegut has suggested, “a canary in a coal mine,” an entity unstuck in time looking for stability, the elusive composed?
In The Cargo Cult series (2020) I use the machine learning technology of “style transfer” to explore the cyclic evolution of transformational exchange between a manually (autographically) manipulated object and the computational representation of that object. I began this series with a sculptural reflection upon the socio-cultural phenomenon of cargo cults, creating an object meant to “call back” the cybernetic gods who fashion the world but detach from direct presence, leaving us unable to reproduce the magic we have once witnessed. This object is used as a canvas to be modified by computational convolution with samples drawn from the environments of its complicit appearance.
The two representations undergo an AI process of multi-style-transfer which interprets the previous state of the content in terms of updated hand-painted responses to the most recent computational output. Thus, a cycle of computational AI mediated perturbation is introduced into the artist’s process of expression and reflection, suggesting that a symbiotic relation with technology develops in media positioned as aesthetic apparatus.
The Cargo Cult sculptural component is used as image “content.” The object underwent successive physical transformations as the output images were in turn painted over autographically to affect the next cycle of algorithmic iteration and influence structural changes made to the content sculpture.
One of the multi-style sets used in the composition of the image. These “style” images represent a starting set, drawn from the “sampled world.” As the process evolves these samples are replaced with iterations of the computational output imagery itself; the output becoming the style in a self-reinforcing feedback that becomes increasingly self-referential while remaining stochastically indeterminate.
Iterations of style and content with style scale increasing from top to bottom and content influence increasing from left to right. The artist determines points of visual resonance and aesthetic branching in a wide affective palette.
Virtually endless variation is encountered in the process of AI mediated resonant self-reflection. Here four studies are shown. From left to right, the first is purely generated by the neural network from supplied input images, the second a digital composite of network output, the third a digital colouring of network output and the fourth an autographic drawing on paper over the same output image as the third. Such technologies both expand and contract the aesthetic horizon in that the range of possibility rapidly generated by AI technologies can overwhelm intent with noise. The artist learns to discriminate across dimensions of their work that they may not have been aware existed: Does one lose the self or find what was missing?
The image is scaled up and cut into tiles which will be mounted on canvas allowing for larger autographic gestural interaction. Alternately, several of these tiles might stand on their own as compositions for further development. The embedded influence of network training data coupled with the influence of artist chosen sources reveals unexpected echoes of source motivations, offering an affective serendipitous space of anticipatory aesthetics.
Proceeding beyond the machine, attempting escape from the implicit metaphor of computation, in this pair of images I enter an exploration of the introduction of human-scaled autographic gesture and media affordances in the cycle of AI processing. The computational neural media output in progress was scaled-up to 36″ x 36″ and transferred to canvas. This allowed for the insertion of aesthetic expression literally at body-scale, proposing that embodied knowledge must enter praxis as directly embodied, not as mediated samples. The process revealed a surprisingly different relation to the image space and the compositional clues perhaps embedded in it.
A texture fragment sampled from the work-in-progress revealing algorithmic reorganization and blending of organic and mechanical referents—a drawing of the silicon mind. The eye’s traverse across potential referents embedded in the neural surface promotes tacit cognitive awareness of a changing gestalt, inducing perceptual discontinuity often perceived as pareidolic phenomenon.