RUE
QIAN
Introducing Unstable Infusion,
a collective AI image generator that subverts convential AIGC aesthetics.
The core of this project Algorithm Deformed is to de-center the singular mode of image generation, and give space and agency for more than one prompt, to see how the image would change when there's multiple participants, multiple layers of meaning, and multiple input of perspectives.
In this sense, generative AI is used less as a tool that encloses a vision, but more of a medium through which a diversity of visions could be expressed, the in-betweenness it creates and the ambiguity that it allows. 
Unstable Infusion, named after the generative AI tool Stable Diffusion, is a prototype that directly intervenes in its diffusion model. Diverging from automated, single-prompt denoising, it now invites a community of people to be part of the generation process, and have each of our prompts to influence every round of the denoising, disturbing predictions the algorithm makes to generate an image.
The outcome image might not conform to the "standard" depiction of the concept we'd like to generate. But it is precisely this “incorrectness” and deviation from the normative that shifts the authority from the algorithm to this collective process we’ve been through. It is a version of the concept that speaks to a variation of perceptions that is unique to each of the participants.


Here's a collection of the deformed concepts generated from Unstable Infusion:
...and more to be explored by you.


It is a deformation that doesn’t strive to put the humanities together and establish the integrity of a knowledge, but rather, a deformation that is a departure that moves away from dominant perceptual models structured by algorithms. It is a mode of knowing that relies on unknowing, erasing what an algorithm is defining as legitimate knowledge, in order to produce meanings of our own.


I'm currently developing this prototype into a public tool. Stay tuned for its release :)