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

111–120 of 365 posts

Re: Stable Attribution

#111

Post by mrtksn »

bioemerl wrote: post #97
mrtksn wrote: post #66

Earlier quoted context omitted.

Artist function is not limited to the act of producing the creation. Their value comes from creating the methods of exploring ideas or feelings and not from the output itself. Guernica is not invaluable because no one else can draw it and Picasso is not infamous because drew the best lines. Just to be clear, I think art made with AI also can have value because it's a tool after all. My concern is, it breaks the busin…

> My concern is, it breaks the business for the people that feeds on and that's not OK. Why not? This is a website that's filled with programmers, almost all of our wealth is built on the skulls of jobs that once existed, and those programs were built without the consent of the people whose jobs were automated away, often through observing and rewriting the processes they used to do. The standard, this has been going…

We automate people's jobs, not taking their intellectual properties and build machines that can churn endless versions of it. The people put out of work by computers no longer work and they were paid in full for the work they did, unlike the people who developed intellectual property on their own dime hoping to be compensated later only to find out that their work was copied and distributed by the computer people.

Re: Stable Attribution

#112

Post by astrange »

kmeisthax wrote: post #68
fwlr wrote: post #45

Uploading works by real human artists gives you a batch of results that resemble a reference board (mood board, inspiration board, etc) the artist could have been looking at while creating their original work of art. Obviously it’s not the actual reference board, the only way to get that is to ask the artist yourself, but it sure looks like what you’d expect their reference board to look like. This site is grift, of…

The AI does NOT build or use reference boards for specific prompts. The only reference it has is the prompt itself, which gets distilled down into a list of 512 numbers, each one of which the AI associates with a particular image feature (or set of features). The only reference material it has is the training set. Most images are not actually retained in the model; but certain statistically significant or repeated im…

> Most images are not actually retained in the model; but certain statistically significant or repeated images will be.

This isn't true. The distribution of images seen by the model is not the distribution of images in LAION; the base models were trained with deduplication (or tried to) and custom ones could've been forcibly overfitted. So "repeated" as seen in clip-front doesn't prove anything.

Re: Stable Attribution

#113

Post by mrtksn »

astrange wrote: post #108
mrtksn wrote: post #101

Earlier quoted context omitted.

They put it online for free for human consumption. Because the tech is very new, old concepts and laws don't cover it and that's not the point. The assumption that there must be compensation comes from the capitalist society that we live in. Switching to Communism or something else can be a solution to not directly pay people who do works and still have them around.

Capitalism isn't when you pay for things, it's when there's capital you can sell.

The capital of the artists is their intellectual property, that's the point.

Re: Stable Attribution

#114

Post by paconbork »

whatshisface wrote: post #12

To actually accomplish something like this purports to be (the linked tool only searches for similar images and doesn't tell you anything about how information ended up inside the model), you could try removing individual images or sets of images from the same artist from the training dataset to see what outputs the resulting model would lose the ability to create. It would be expensive to do that for more than a few…

Sounds a bit like differential privacy

Re: Stable Attribution

#115

Post by goldemerald »

esperent wrote: post #83
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…

> Based on some of the examples they explicitly provided, it is clear to me Stable Diffusion creates novel art. You're jumping to a conclusion that the data doesn't warrant, I suspect because it's a conclusion that suits you. They're doing something like a "reverse image search" from an AI generated image over the original dataset and returning a few examples that have a high degree of similarity. There's no guarante…

Using pixel-space or latent-space distances to measure if a generative model has simply memorized the training data is a common evaluation metric in ML literature. If the website is using the full training set of LAION 5-B used to train Stable Diffusion, then I find it extremely convincing that SD has generalized to the data distribution it was trained on, and does not simply regurgitate other people's artwork. If there is a way to find more similar images, I have yet to see it.

I think this website is evidence SD is a novel image generator, and rarely creates infringing images (at least in the way we thought of infringing before these kinds of AI).

Re: Stable Attribution

#116

Post by josephg »

wokwokwok wrote: post #98
social_quotient wrote: post #86

Earlier quoted context omitted.

So the parent post to yours said he’s taking piano lessons. Let’s suppose he becomes great. And let’s suppose he writes his own music. How is Taylor Swift supposed to get paid by him from his early lessons which have undoubtedly played a key role in not only his love for music but the song he writes today?

You are fundamentally not addressing the issue. Previously, someone was being paid to produce something. Now, they are no longer being paid and the business can get the same thing “for free”. You can argue the same issue about out sourcing programmers to south east Asia. Is it good or bad? “…but how do we compensate the programmers who lost their jobs?” It’s just economics right? …but saying that doesn’t address the…

> People. Out of jobs.

> You think, maybe, it’s fair for the people being replaced to feel a bit upset about it?

I think it makes all sorts of sense that people are upset about stable diffusion.

And, people being upset isn't a good reason to change the law or outlaw the new technology.

Today the word "luddite" is a slur. But luddites were real people who made the same argument you're making. In their case, they were a secret group of textile workers so upset by the introduction of the mechanical loom that they went around sabotaging equipment.

Suppose we wound the clock back and the luddites won a legal battle and successfully outlawed the mechanical loom in the UK. Knowing what we know now, would you support a law like that to protect the jobs of textile workers in the 19th century? I sure wouldn't - it would have decimated the economy of the UK and stunted innovation.

Thats the danger of using the law to protect the status quo. Sometimes the status quo needs to change to make room for what comes next, regardless of how painful that change is. The law doesn't exist to protect your business model.

[1] https://en.wikipedia.org/wiki/Luddite

We're at the verge of the second industrial revolution. I have no idea how it shakes out, but I don't think clinging desperately to the old ways of doing things will be a winning strategy in the long run.

Re: Stable Attribution

#117

Post by poxrud »

I find the original images that were used to generate the art to be much more beautiful and emotion evoking than the strange and lifeless AI images. Maybe it’s just my subconscious but for me something always seems off with the AI generated art.

Re: Stable Attribution

#118

Post by mattbee »

clbrmbr wrote: post #89
mattbee wrote: post #44

I like this because they are trying to show how AI is a copyright laundry. I can see other commenters picking apart its method of heuristically guessing at source images from training data. That obviously won't be accurate, or a full picture, but I wonder if it would convince a judge. An interesting challenge for these heuristics would be to take the picture under test along with its prompt, retrain the model without…

Unless there are imitation Kincades in the training set labeled with the name Kincade, I would expect the model to fail to produce anything relevant. It’s a strange situation. Certainly a human painter is free to study copywritten works closely and then produce art in the style of the other artist, provided the resulting work is “in the style of” but does not copy a specific painting. So why is a human SD-artist not…

If there were a legally sound way to trace inspiration from a copyrighted work through the brain into another work, lawyers would be all over it! They try often enough with music.

The line will surely get a lot less fuzzy with AI models.

Re: Stable Attribution

#119

Post by qwerty456127 »

Is there still any chance for a human to come up with something new an AI can't? Haven't all the possible ideas ans/or "sub-ideas" already been implemented and fed into "AIs"?

I suspect the humanity is omniscient and it's only problem is lack of possibility to keep everything it knows in a mind simultaneously and connect the dots. An "AI" seems to be a solution to this problem.

Even an individual human could be almost omniscient if he could simultaneously put everything he has ever knew/thought/seen into his working memory.

Re: Stable Attribution

#120

Post by minimaxir »

kmeisthax wrote: post #68
fwlr wrote: post #45

Uploading works by real human artists gives you a batch of results that resemble a reference board (mood board, inspiration board, etc) the artist could have been looking at while creating their original work of art. Obviously it’s not the actual reference board, the only way to get that is to ask the artist yourself, but it sure looks like what you’d expect their reference board to look like. This site is grift, of…

The AI does NOT build or use reference boards for specific prompts. The only reference it has is the prompt itself, which gets distilled down into a list of 512 numbers, each one of which the AI associates with a particular image feature (or set of features). The only reference material it has is the training set. Most images are not actually retained in the model; but certain statistically significant or repeated im…

> The AI does NOT build or use reference boards for specific prompts. The only reference it has is the prompt itself, which gets distilled down into a list of 512 numbers, each one of which the AI associates with a particular image feature (or set of features).

Minor technical clarification:

For SD 1.X, CLIPText encodes a prompt and passes a (77, 768) [edit: up to 77] matrix to the core UNet. For SD 2.X, OpenCLIP passes a (77, 1024) matrix.

The black-box interaction between the text-encoded matrices and the UNet is how Stable Diffusion "learns" text associations.

Post reply on HN