Generative Adversarial Nausea

A medium angle photo of four (out of 10 total in the installation) bright LCDs displaying a static image that loosely resemble human beings. The screen is attached to a microcontroller and hangs on the wall with four pushpins. A 12 volt power supply connects at the bottom of the microcontroller and then threads into the wall.A closeup photo of two (out of 10 total in the installation) bright LCDs displaying a static image that loosely resemble human beings. The screen is attached to a microcontroller and hangs on the wall with four pushpins. A 12 volt power supply connects at the bottom of the microcontroller and then threads into the wall.A closeup photo of a bright LCD displaying a static image that loosely resembles a person. The screen is attached to a microcontroller and hangs on the wall with four pushpins. A 12 volt power supply connects at the bottom of the microcontroller and then threads into the wall.A closeup photo of a bright LCD displaying a static image that loosely resembles a person. The screen is attached to a microcontroller and hangs on the wall with four pushpins. A 12 volt power supply connects at the bottom of the microcontroller and then threads into the wall.A wide angle shot displaying 10 bright LCDs, each with static image that loosely resemble human beings. Each screen is attached to a microcontroller and hangs on the wall with four pushpins. A 12 volt power supply connects at the bottom of the microcontroller and then threads into the wall.

 

(5 images. alt text included)

Generative Adversarial Nausea is an endeavor in categorical de-categorization in which I trained a neural network (a fundamental component of artificial intelligence) to learn the visual attributes of two distinct sets of images- one of hundreds of stock photo thumbnails of “nauseated” people and the other of a comparable amount of selfies I have taken the past four years living with my undetectable illness- in order to produce new images based on its learning. Though the two reactants in my machine learning experiment were flawed representations of nausea, my interest in combining them to make new images was two-fold. 

Firstly, I wanted to engage in categorical confusion; the hybrid images are no longer clear indicators of nausea (as the stock photos were trying to be) nor are they recognizable as a singular digital subject. Secondly, I wanted to employ a paranoid technology that tries to make sense and replicate its ingestion, to be contaminated by the same unease as its subject matter. The people who inhabit the stock photographs demonstrate nausea with exaggerated facial expressions and physical contortions. Their pain is specific enough to be tagged in the stock database as some conjugate of “nausea” and yet ambiguous enough to also stand-in for any number of bodily pains. This legibility as several co-existing, co-tagged states is key to their success or failure as images whose main purpose is paid reproduction and dissemination. Meanwhile, the subtle expressions and gestures found in my selfies, whose dissemination is rooted in my accrual of social, rather than economic, capital, belie the pain and uncertainty of my internal lived condition. Their categorical specificity lies in their reification of me as a singular subject, evidenced by the facial detection algorithms training constantly on social media platforms that surface innocently enough as white squares and suggested tags.

In fixing these 10 images on bright LCDs, I seek playful tension between the fixed state of the image and the possibility for its continued immersion back into a long and identity-altering training process. The title Generative Adversarial Nausea is a play on the acronym G.A.N. which stands for generative adversarial network. GANS are a relatively recent (2014) breakthrough in artificial intelligence that initiate learning through self-deception. A GAN is a two-part system in which one network (the “generator) creates new images based on what it learns from an original set of images it trains on, while another network (the “discriminator”) tries to determine whether these images are based in reality or synthesized. Managing an undiagnosed chronic condition yields a similar division of self into an experiencer and a diagnoser. 

 

Hiatal

Closer examination of the stock images used in my training process show humans trapped in exaggerated moments of pain while inhabiting highly aspirational apartments that seem to tell them to hurry their pain along (presumably so that they can return to being productive enough to afford occupying these spaces). There is no space for a mild nausea or mild investment in place. Hiatal, my 8-minute video, subverts these images by re-animating amorphous figures within them who deflate, hollow-out and slip into the cracks; simultaneously occupying and undoing subjecthood and place. They expose their aspirational surroundings as equally malleable backdrops that can be unveiled, blown away, and thrown up.


Hiatal. 2020. 7 minutes, 50 seconds. Digital Video.

Image Description: This video is comprised of five separate scenes, each taking place in the environment of a stock photograph: a couch in a living room in an exposed brick apartment, a large bedroom with a full-sized mirror and bright light from the windows, a clean kitchen in a bright white apartment, a bathroom consisting of a single white toilet, and a long hospital hallway with a single vanishing point. There is one figure in each of the scenes, and using the 3D software Blender, I attempted to turn each of these figures inside out. In the process, they deflate, twist, hollow-out and slip into the cracks of their environment. The sound emanating from two speakers hung on either side of the video projection is used to further bring the scene to life. The rustling of fabric describes the movements of the figures on screen, while radios, birds, clocks, and other voices, all describe things happening in the periphery.


Excerpt, Hiatal. 2020. Documented at Mason Gross Galleries, New Jersey.

Image Description: The video described above documented in an exhibition space where it is projected to fill the entire width of a wall. A one minute segment.

 

 

special thanks to Playform.io AI art platform.