Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Nightshade: An offensive tool for artists against AI art generators

nightshade.cs.uchicago.edu

Post reply on HN
Print view View original

251–260 of 710 posts

Re: Nightshade: An offensive tool for artists against AI art generators

#251

Post by jdietrich »

throwoutway wrote: post #185
efitz wrote: post #172

This is the DRM problem again. However much we might wish that it was not true, ideas are not rivalrous. If you share an idea with another person, they now have that idea too. If you share words on paper, then someone with eyes and a brain might memorize them (or much more likely, just grasp and retain the ideas conveyed in the words). If you let someone hear your music, then the ideas (phrasing, style, melody, etc)…

I don't see the parallel between this offensive tool and DRM. I could, say buy a perpetual license to an image from the artist, so that I can print it and put it on my wall, while it can simultaneously be poisonous to an AI system. I can even steal it and print it, while it is still poisonous to an AI system. The closest parallel I can think of is that humans can ingest chocolate but dogs should not.

What you've described is the literal, dictionary definition of Digital Rights Management - a technology to restrict the use of a digital asset beyond the contractually-agreed terms. Copying is only one of many uses that the copyright-holder may wish to prevent. The regional lockout on a DVD had nothing to do with copy-protection, but it was still DRM.

Re: Nightshade: An offensive tool for artists against AI art generators

#252

Post by jwells89 »

squidbeak wrote: post #206

I really don't understand the anxiety of artists towards AI - as if creatives haven't always borrowed and imitated. Every leading artist has had acolytes, and while it's true no artist ever had an acolyte as prodigiously productive as AI will be, I don't see anything different between a young artist looking to Picasso for cues and Stable Diffusion or DALL-E doing the same. Styles and methods haven't ever been subject…

Imitation isn’t the problem so much as it is that ML generated images are composed of a mush of the images it was trained on. A human artist can abstract the concepts underpinning a style and mimic it by drawing all-new lineart, coloration, shading, composition, etc, while the ML model has to lean on blending training imagery together.

Furthermore there’s a sort of unavoidable “jitter” in human-produced art that varies between individuals that stems from vastly different ways of thinking, perception of the world, mental abstraction processes, life experiences, etc. This is why artists who start out imitating other artists almost always develop their imitations into a style all their own — the imitations were already appreciably different from the original due to the aforementioned biases and those distinctions only grow with time and experimentation.

There would be greatly reduced moral controversy surrounding ML models if they lacked that mincemeat/pink slime aspect.

Re: Nightshade: An offensive tool for artists against AI art generators

#253

Post by 23B1 »

Minor49er wrote: post #108
popohauer wrote: post #16

I'm glad to see tools like Nightshade starting to pop up to protect the real life creativity of artists. I like AI art, but I do feel conflicted about its potential long term effects towards a society that no longer values authentic creativity.

Is the existence of the AI tool not itself a product of authentic creativity? Does eliminating barriers to image generation not facilitate authentic creativity?

No, it facilitates commoditization. Art – real art – is fundamentally a human-to-human transaction. Once everyone can fire perfectly-rendered perfectly-unique pieces of 'art' at each other, it'll just become like the internet is today: filled with extremely low-value noise.

Enjoy the short term novelty while you can.

Re: Nightshade: An offensive tool for artists against AI art generators

#254

Post by 23B1 »

ufocia wrote: post #209
jamesu wrote: post #32

Long-term I think the real problem for artists will be corporations generating their own high quality targeted datasets from a cheap labor pool, completely outcompeting them by a landslide.

It will democratize art.

then it won't be art anymore, it'll just be mountains of shit

sorta like what the laptop did for writing

Re: Nightshade: An offensive tool for artists against AI art generators

#256

Post by ukuina »

wizzwizz4 wrote: post #82
ukuina wrote: post #78

Won't a simple downsample->upsample be the antidote?

How do you train your upsampler? (Also: why are you seeking to provide an “antidote”?)

I apologize. I was trying to respond to inflammatory language ("poison") with similarly hyperbolic terms, and I should know better than to do that.

Let me rephrase: Would AI-powered upscaling/downscaling (not a simple deterministic mathematical scaling) not defeat this at a conceptual level?

Re: Nightshade: An offensive tool for artists against AI art generators

#257

Post by 23B1 »

squidbeak wrote: post #206

I really don't understand the anxiety of artists towards AI - as if creatives haven't always borrowed and imitated. Every leading artist has had acolytes, and while it's true no artist ever had an acolyte as prodigiously productive as AI will be, I don't see anything different between a young artist looking to Picasso for cues and Stable Diffusion or DALL-E doing the same. Styles and methods haven't ever been subject…

[flagged]

Re: Nightshade: An offensive tool for artists against AI art generators

#258

Post by MattRix »

ufocia wrote: post #204
bakugo wrote: post #115

Earlier quoted context omitted.

A human artist cannot look at and memorize 100000 pictures in a day, and cannot paint 100000 pictures in a day. I am SO tired of this non-argument

A human artist does not need to look at and memorize 100000 pictures in any span of time, period. Current AI does. We needed huge amounts of human labor to fund and build Versailles. I'm sure many died as a result. Now we have machines that save many of those lives and labor. What's your non-argument?

The argument is that the humans producing the work should be willing participants. I don’t think that’s too much to ask for.

Re: Nightshade: An offensive tool for artists against AI art generators

#259

Post by xg15 »

peter_d_sherman wrote: post #158
xg15 wrote: post #90

Earlier quoted context omitted.

And then what? The scrapers themselves already happily ignore copyright, they won't be inclined to obey a no-ai.txt. So someone would have to enforce the standard. Currently I see no organisation who would be willing to do this or even just technologically able - as even just detecting such scrapers is an extremely hard task. Nevertheless, I hope that at some not-so-far point in the future there will be more legal gu…

>The scrapers themselves already happily ignore copyright, they won't be inclined to obey a no-ai.txt. Name two entities that were asked to stop using a given individuals' images that failed to stop using them after the stop request was issued. >Currently I see no organisation who would be willing to do this or even just technologically able - as even just detecting such scrapers is an extremely hard task. // Part of…

> Part of Image Web Scraper For AI Image Generator ingestion psuedocode:...

Yes, and who is supposed to run that code?

> Name two entities that were asked to stop using a given individuals' images that failed to stop using them after the stop request was issued.

Github? OpenAI?[1] Stable Diffusion?[2] LAION?[3] What do you think why there are currently multiple high-profile lawsuits ongoing about exactly that topic?

Besides, that's not how things work. Training a foundation model takes months and currently costs a fortune in hardware and power - and once the model is trained, there is, as of now, no way to remove individual images from the model without restraining. So in practical terms it's impossible to remove an image if it has already been trained on.

So the better question would be, name two entities who have ignored an artist's request to not include their image when they encountered it the first time. It's still a trick question though because the point is that scraping happens in private - we can't know which images were scraped without access to the training data. The one indication that it was probably scraped is if a model manages to reproduce it verbatim - which is the basis for some of the above lawsuits.

[1] https://www.theverge.com/2022/11/8/23446821/microsoft-openai...

[2] https://www.theverge.com/2023/2/6/23587393/ai-art-copyright-...

[3] https://www.heise.de/hintergrund/Stock-photographer-sues-AI-...

Re: Nightshade: An offensive tool for artists against AI art generators

#260

Post by jdietrich »

jamesu wrote: post #32

Long-term I think the real problem for artists will be corporations generating their own high quality targeted datasets from a cheap labor pool, completely outcompeting them by a landslide.

In the short-to-medium term, we're seeing huge improvements in the data efficiency of generative models. We haven't really started to see self-training in diffusion models, which could improve data efficiency by orders of magnitude. Current models are good at generalisation and are getting better at an incredible pace, so any efforts to limit the progress of AI by restricting access to training data is a speedbump rather than a roadblock.
Post reply on HN