-- Leo's gemini proxy
-- Connecting to sol.cities.yesterweb.org:1965...
-- Connected
-- Sending request
-- Meta line: 20 text/gemini
posted 2022-06-01 | afterworded 2023-09-04
why do i work with AI? first of all, because i can. but what made me keep at it, and keep reading about it?
GAN-generated images hold a sense of wonder, as if you were to wander into the library of babel and find a floor where all the books made some simulacrum of cohesion, rather than being mere linguistic white noise.
AI art offers a method of uncontrollable yet not randomized generative art. a multi-step process of translations from meaning into meaninglessness back into meaning again: i ask the machine to give me an angel (i type a meaningful prompt in english). the machine doesn't know what an angel is, but it does know what several people said it looked like (the dataset itself is a microcosmos of translation from word (and what words are chosen?) to image and vice versa). the machine outputs what i may or may not recognize as an angel or as something else, and i make my own meaning of it. or the machine outputs what i choose to interpret as what an angel is in its own world (that we also made in the sense of panspermia, that we sowed but do not control what sprouts). and now when i say angel i mean the same thing they meant when making the dataset, but also what the machine has showed me.
i keep translating asemia into meaning and meaning into asemia. i paint the machine angel as one would a still life, or a landscape. i paint as if from a photograph. i take the angel and paint it over and over from my imagination, no longer a generated image but a generated concept of its own.
i start from artbreeder and now my vocabulary is arbitrarily limited but it's all the same for the algorithm. i request a { monastery / maillot / altar / letter opener / scuba diver / jellyfish / stingray / sea anemone / bee eater / microphone / mask / space shuttle } and the math finds a way to quantify all these words-concepts into numbers and into a location on latent space and into a sequence of pixels it deems adequate. i look at the image and translate it into something else, i choose to call it simply an underwater building.
the meaningless output becomes part of a story / vocabulary / shorthand when I touch it with my human brain. i become a cyborg.
i write this in english so that i can translate it to portuguese.
rather than coming from rogue AI narratives and retroactively applying them to our reality, working with the AI of the present creates new stories grounded in our current, real struggles and issues — of mass surveillance, unsustainable growth, pursuit of automating humanity away from workers and consumers. away from people.
i work with AI because I can. because it is already a part of our world, society, economy, and so it is also a part of ourselves and our art. i work with AI because i'm a digital native and this is an endemic species and i need to understand my habitat. i work with AI because all the work and resources put into it can't be just for the sake of profit and surveillance by a capitalist system, it also has to be made useless and mystical and used for beauty and reflection. i work with AI because i cannot stand the push for hyperrealism. i cannot let the photos of the unreal become stale stock photos of a manicured Earth. i cannot let the AI be used as pure content mills for a quick NFT profit either, from mindless profit driven automation into mindless profit driven automated products. AI generated images are a vision of the future and i belong to the future and get to claim it for myself.
my work with AI started with an image in september 2020 and then a digital painting of said image in october. time flows differently for computers and it now seems like a lifetime ago.
february 2021: i create my first "gallery", a first person experience in unity 3D. you are placed in a small environment, surrounded by eight images generated with Artbreeder. the images constantly rotate to face you / the camera, in a sort of role reversal: rather than looking at passive artwork, the art looks at you, their gaze unavoidable.
april 2021: an abstract dollmaker composed of cutouts of Artbreeder-generated images and paleolithic Venus figurines. the ancient and the cutting-edge, all parts of the same whole of humanity, put into a nostalgic context of a digital native childhood and implications of constructing oneself.
june 2021: M??, an imaginary museum accessed only through its ephemera — a series of three leaflets, written asemically, one of the exhibits composed of generated images and paintings of them. dream artifacts.
november 2021: i learn to speak directly to someone's machine — i learn of colab notebooks and tap into the AI art machine, made by Hillel Wayne and based on Katherine Crowson's notebook. i create angels, traditional imagery for the traditional medium of lithographs, their ephemeral images (gone if not saved to a hard drive) to be eternalized in stone, the ancient and the cutting-edge once again. i only print them as faux-lithographs made with eyeliner pencil and oil paints but the intention is still there.
2022: i continue my work with this extension of my body. i explore compression, as the industry rushes and pushes for higher resolutions that burn through electricity as each pixel is carefully generated throughout the virtual epochs. i read 'braiding sweetgrass' and wonder how can i honor the human labor and the natural resources that go into every single pixel in this array picked by the machine. the algorithm is not a person or an animal, not sentient, intelligent or even alive, and this matters so much to me, but i create a fictional world where nevertheless all the data and metadata harvested (like plants) and living (like plankton) in the cloud becomes physical and concrete, because indirectly it is.
===
originally written for my thesis report, yes i wrote it in english knowing full well i'd need to translate it later, yes this most likely shows in the translated version, yes i wish i had less worms in my brain. anyway, figured i'd post it since it's already written and i'll need to change stuff for the final report regardless.
gee it's been a while hasn't it. i'll get around to converting my little thesis write up to gemtext... at some point. meanwhile, here it is on my website:
it was such a pain working on this subject in 2022. it was not a nice year to read academic papers and books on the subject of machine learning and then chill online and be bombarded by the worst takes imaginable. i genuinely wish "AI enthusiasts" actually did some work into understanding what they're supposedly into. but of course, it's all about the hype machine, and not the actual tech that exists. so they're all nice and chipper while me and other people interested on _ethical_ and _artistic_ synthesis get burnout. don't even get me started on copyright bullshit.
i miss working with fucked up synthetic images. i'm still working on dataplankton, and it's lacking in images, so i might get back to fooling around. collages too.
but the nature of being an artist is that as the time passes and you try new stuff out, you just accumulate more things you wish you could go back to, but days don't get any longer. especially with a job. oh well.
-- Response ended
-- Page fetched on Sun May 12 00:12:01 2024