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

Stable Attribution

stableattribution.com

Post reply on HN
Print view View original

251–260 of 365 posts

Re: Stable Attribution

#251

Post by otabdeveloper4 »

mrtksn wrote: post #46
goldemerald wrote: post #8

This is a great website, but not in the way the authors intended. Based on some of the examples they explicitly provided, it is clear to me Stable Diffusion creates novel art. Here's a random example https://www.stableattribution.com/?image=a2666aee-0a1a-411b-... I will admit this is a nice tool for verifying the creations of SD aren't pure copies, so I think it will be useful for a time. But as AI-generated images s…

The discussion over novelty is useless, all these legal structures like copyrights patents royalties and licenses are about creating business structures to allow artist or inventors be compensated. They were already imperfect in it and now suddenly a new technology drags their work into the wilderness with no stuctures for compensation. The images created by the AI are indeed novel but they feed on the work of people…

Don't worry about it too much. From now on, nobody will release their art for public access anymore. Even portfolios will be paywalled to prevent copyright theft being laundered through "AI".

Which means there will not ever be a Stable Diffusion 2.0 with new art styles. In a year or two the instantly recognizable style of Stable Diffusion will be seen as extremely lame and crass.

Re: Stable Attribution

#252

Post by mikewarot »

I would like to be able to take my 300,000 photos and throw them at something like at the embedded latent space behind StableDiffusion, to have it rate them, add keywords, etc. This would allow me to them put the top 1000 on Flickr.

Re: Stable Attribution

#253

Post by XorNot »

ertian wrote: post #47
version_five wrote: post #2

This appears to be just looking for the nearest neighbors of the image in embedding space and calling those the source data. This by definition would find similar looking images, but it's not strictly correct to call it attribution. To some extent all of the training data is responsible for the result - as an example, the model is also learning from negative examples. The result here may feel satisfying, but it's ove…

Yeah, it seems like it'll just as happily 'attribute' human-made images as well, which calls the whole thing into question. If it's really showing the images that stable diffusion has 'stolen from', and it'll do the same for humans, does that not mean people are equally guilty?

Yep straight up just uploaded a picture from my camera phone and it confidently found a bunch of strikingly similar images.

Re: Stable Attribution

#254

Post by minimaxir »

mikewarot wrote: post #252

I would like to be able to take my 300,000 photos and throw them at something like at the embedded latent space behind StableDiffusion, to have it rate them, add keywords, etc. This would allow me to them put the top 1000 on Flickr.

That is more for CLIP than Stable Diffusion.

Re: Stable Attribution

#255

Post by esskay »

This is a bit flawed. By providing a non-ai photo it just picks up similar photos and claims they were used to generate it.

It's a nice, but as I say, flawed concept.

Re: Stable Attribution

#256

Post by avereveard »

josephg wrote: post #206
avereveard wrote: post #174

Earlier quoted context omitted.

Training method and recall are of no consequences because copyright law doesn't deal with technology for the most part. There is a test to wether the image is stored or not which is interesting here for the topic at hand, but it's always in the context of distribution. The worst is that weight may be considered storage, but the point is... Law already covers that. Because it doesn't concern with technology, but with…

> law is technology agnostic, just adding a layer of redirection doesn't matter, because tm it cares about the input and the output not what happens in between. What makes you think that? The law cares about whatever lawyers decide to care about. There was a case a few years ago where (if memory serves) a black woman sued an insurance company for discrimination after the insurance company refused to provide her cover…

Insurance is not copyright and the case is not even the same subject matter.

And again that case is technology agnostic, discrimination law requires you to be able to provide proof that results are non discriminatory, law itself doesn't care that it was specifically a neural network, it only cares about the end result, the firm lost because it failed to provide required data about their decision process, not because it was using neural networks, that they used a neural network was irrelevant on its own, and it could have been fine if they baked explainability in it.

Re: Stable Attribution

#257

Post by vintermann »

version_five wrote: post #2

This appears to be just looking for the nearest neighbors of the image in embedding space and calling those the source data. This by definition would find similar looking images, but it's not strictly correct to call it attribution. To some extent all of the training data is responsible for the result - as an example, the model is also learning from negative examples. The result here may feel satisfying, but it's ove…

So it is the opposite of attribution. It just makes up a plausible-looking story of attribution and passes it off as the truth. If you passed it a hand-painted image from 1850 which was not in SD's training dataset, it can happily declare that it was inspired by someone's piece from 2019. There's no actual causal inference going on.

It has more in common with using an AI language model as a "bullshit generator" than it has with human attribution. Which isn't a coincidence, since nearest neighbour similarity searches ARE a type of machine learning, just a much simpler one than SD.

Anti AI people who are upset about attribution, should learn the technology and try to create an actual attribution model: one that has a notion of causality, which could say who influenced who.

Causal model requires causal assumptions, but you can probably get far with the simple assumption that "works from the future do not influence works from the past".

Re: Stable Attribution

#258

Post by jfoster »

vintermann wrote: post #257
version_five wrote: post #2

This appears to be just looking for the nearest neighbors of the image in embedding space and calling those the source data. This by definition would find similar looking images, but it's not strictly correct to call it attribution. To some extent all of the training data is responsible for the result - as an example, the model is also learning from negative examples. The result here may feel satisfying, but it's ove…

So it is the opposite of attribution. It just makes up a plausible-looking story of attribution and passes it off as the truth. If you passed it a hand-painted image from 1850 which was not in SD's training dataset, it can happily declare that it was inspired by someone's piece from 2019. There's no actual causal inference going on. It has more in common with using an AI language model as a "bullshit generator" than…

You can literally take a photo right now, upload it, and it will find similar photos from the SD dataset.

Re: Stable Attribution

#259

Post by stale2002 »

ForestCritter wrote: post #154
bioemerl wrote: post #94

Earlier quoted context omitted.

I don't get the assumption that there should be. The machine was trained on publicly available hard that was already free. Why do people think they need to be compensated for something they put up online for free? They don't have the right to not allow people to learn from it, that's just never been a part of copyright.

It is expected of any professional artist to have an online portfolio if they are serious about generating work. Scraping my portfolio that I put out because I want to generate revenue is a shitty thing to do. Also 'trained' is a complete misnomer. AI is fed images made by humans, labelled and tagged by humans, categorized by styles defined by humans, then plotted, copied, traced by a program written by humans. Along…

> Art is made with physical motor skills, time and effort

Apparently it isn't. Now a computer can do it.

> that requires no effort of aquiring the physical skill needed to make the art.

So what? This doesn't matter.

> If you want to physically copy my art, go for it

This is the same thing as if a computer does it.

Re: Stable Attribution

#260

Post by wiz21c »

Attribution is half of the problem. I'm not happy at all with people putting my code behind the closed wall of AI to regurgitate some other code inspired from it...

I say : my code as in "the text of my program", not "the idea of my code" (I have no problem with people re-using my ideas as they are mostly ideas that I have learnt from someone else).

Post reply on HN