University of Chicago researchers seek to “poison” AI art generators with Nightshade
University of Chicago researchers seek to “poison” AI art generators with Nightshade
Altered images could destroy AI model training efforts that scrape art without consent.
cross-posted from: https://lemdro.id/post/2614204 (!aistuff@lemdro.id)
Reminder that this is made by Ben Zhao, the University of Chicago professor who illegally stole open source code.
How did he illegally steal open source code? Isn’t it open source? Which license did the code use? How did he use it in a way that violated the law or the license?
Click the link and you'll find out.
The University of Chicago is such an unnoticed stain on human development. Their faculty fostered the aptly named Chicago School Of Economics theory of how to run a market. It's evolved over the generations since it is never actually able to predict anything. It's only ever used to retrospectively analyze things for the sake of a conservative narrative.
They're a nightmare factory manufacturing human suffering now. It's no surprise that another part of the school actively steals code while thwarting AI development. Nothing good can ever prosper because of them.
They’re still teaching their economic theories (Black-Scholes) in school. I have to do calculations pretending that the stock market has a normal probability distribution, and nobody has mentioned any problems with this (edit: I feel it has value as a model, but no one has mentioned limitations). I feel like the emperor has no clothes, and my class is about his fancy wardrobe.
There should be no need to mention these limitations. They sold be obvious to people using them, and of course there are academic discussions of the limitations of every model. I think they're just not being presented to you because they're well known.
Anyone who uses models should know that all models are wrong, but some models are useful.
know of any good articles to generally find out more about them?
This tool would be the first to allow content owners to push back in a meaningful way against unauthorized model training
-Ben Zhao
This statement is strange. This is far from the first paper attempting this. This isn't even this author's first attempt. A few months ago he contributed to the glaze paper attempting the same thing, marketed the same way.
yeah it's common to not acknowledge competitors that were there first. But I remember seeing posts about Gaze and if this is the same guy, why does he keeps claiming this is the first tool?
Perhaps because the previous ones keep not working.
Runs the image through an AA filter
Hmmm, seems the nightshade effect's gone now. Wonder why?
In a month this will be a hot new A1111 extension for hybridizing visual concepts.
This is the best summary I could come up with:
The goal is to help visual artists and publishers protect their work from being used to train generative AI image synthesis models, such as Midjourney, DALL-E 3, and Stable Diffusion.
The open source "poison pill" tool (as the University of Chicago's press department calls it) alters images in ways invisible to the human eye that can corrupt an AI model's training process.
Those with access to existing large image databases (such as Getty and Shutterstock) are at an advantage when using licensed training data.
But as the Nightshade team sees it, research use and commercial use are two entirely different things, and they hope their technology can force AI training companies to license image data sets, respect crawler restrictions, and conform to opt-out requests.
"The point of this tool is to balance the playing field between model trainers and content creators," co-author and University of Chicago professor Ben Y. Zhao said in a statement.
Shawn Shan, Wenxin Ding, Josephine Passananti, Haitao Zheng, and Zhao developed Nightshade as part of the Department of Computer Science at the University of Chicago.
The original article contains 656 words, the summary contains 179 words. Saved 73%. I'm a bot and I'm open source!