Researchers on the University of Chicago have launched free “Glaze” software program that they are saying can thwart efforts by generative synthetic intelligence (AI) to copy an artist’s fashion.
The program makes tiny modifications to digital pictures that, whereas invisible to human eyes, act as a “style cloak” when they’re posted on-line, the staff behind the venture defined on their web site.
If generative AI finds a Glaze-guarded picture on-line, it’s prevented from appropriately analyzing and copying the fashion, the staff mentioned.
Glaze was created on the behest of artists outraged that applications like Midjourney and Stable Diffusion, schooled on troves of pictures out there on-line, might mimic their kinds on command.
“AI has been evolving too fast, and there must be some guardrails or regulations around it,” mentioned Shawn Shan, the doctoral scholar answerable for the venture.
“The goal of this is to push back from a technical standpoint.”
The staff behind Glaze labored with artists together with the illustrator Karla Ortiz, who’s among the many plaintiffs in a US courtroom case towards a number of companies with image-producing generative AI companies.
Researchers clarify how Glaze works
Photo Credit: Glaze Project
Â
“If Karla uses our tool to cloak her artwork, by adding tiny changes before posting them on her online portfolio, then Stable Diffusion will not learn Karla’s artistic style,” the lab staff mentioned.
“Instead, the model will interpret her art as a different style – for example, that of Vincent van Gogh.”
The creators of Glaze concede it isn’t a panacea, given how shortly AI evolves.
The hope, the staff mentioned, is that Glaze and related tasks will shield artists no less than till defensive legal guidelines or rules will be carried out.
Glaze has been out there free of charge obtain since March 15.
