Week 4 - "Negative weighting" and "shadow bodies." - Blog 1/10 (Hyejoo)

I recently learned of another algorithmic technique with AI art called "negative weighting." Essentially this finds the opposite of an input. For example, you input "dog" and the AI will find everything it understands to be the opposite of a dog. 

An AI illustrator (who goes by the moniker "Supercomposite"), using negative weighting, came across a grotesque, creepy, ghostly figure that emerged as the output of a string of certain prompts. The illustrator named it "Loab." Here's the article on it by Max Read (disclaimer: potentially scary/gory pics!––you can see Supercomposite's whole Twitter thread, with even scarier images, as well).

"Loab" reminded me of the concept of "shadow bodies" (mentioned in Gillespie). Even though shadow bodies refer to an actual human user, it's kind of like the negative weighting of a user: "these shadow bodies persist and proliferate through information systems..." They remain unseen in the deep ocean of machine learning datasets. 

This was an alluring, intriguing concept for me. Though, the macabre fantasy of "Loab," along with our dystopian imaginations of AI overpowering humans, etc. may all be the flip side of the same coin as fetishizing computers as all powerful, occupying a "higher plane of thought" (Finn 24) etc. As Max Read writes: "Loab's not a ghost––it's an embedding somewhere in the multi-dimensional latent space, strongly associated with gore, darkness, and dread." It's just strong associations... but it sparks our fantasies nonetheless.