this post was submitted on 26 Jan 2024
428 points (83.0% liked)

Technology

60044 readers
2831 users here now

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related content.
  3. Be excellent to each another!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, to ask if your bot can be added please contact us.
  9. Check for duplicates before posting, duplicates may be removed

Approved Bots


founded 2 years ago
MODERATORS
 

We Asked A.I. to Create the Joker. It Generated a Copyrighted Image.::Artists and researchers are exposing copyrighted material hidden within A.I. tools, raising fresh legal questions.

top 50 comments
sorted by: hot top controversial new old
[–] [email protected] 181 points 11 months ago (3 children)

We asked A.I. to create a copyrighted image from the Joker movie. It generated a copyrighted image as expected.

Ftfy

[–] [email protected] 52 points 11 months ago (12 children)

What it proves is that they are feeding entire movies into the training data. It is excellent evidence for when WB and Disney decides to sue the shit out of them.

[–] [email protected] 114 points 11 months ago (7 children)

Does it really have to be entire movies when theres a ton of promotional images and memes with similar images?

load more comments (7 replies)
[–] [email protected] 59 points 11 months ago* (last edited 11 months ago) (1 children)

I think it's much more likely whatever scraping they used to get the training data snatched a screenshot of the movie some random internet user posted somewhere. (To confirm, I typed "joaquin phoenix joker" into Google and this very image was very high up in the image results) And of course not only this one but many many more too.

Now I'm not saying scraping copyrighted material is morally right either, but I'd doubt they'd just feed an entire movie frame by frame (or randomly spaced screenshots from throughout a movie), especially because it would make generating good labels for each frame very difficult.

[–] [email protected] 24 points 11 months ago

I just googled "what does joker look like" and it was the fourth hit on image search.

Well, it was actually an article (unrelated to AI) that used the image.

But then I went simpler -- googling "joker" gives you the image (from the IMDb page) as the second hit.

[–] [email protected] 20 points 11 months ago (3 children)

WB and Disney would lose, at least without an amendment to copyright law. That in fact just happened in one court case. It was ruled that using a copyrighted work to train AI does not violate that works copyright.

load more comments (3 replies)
[–] [email protected] 10 points 11 months ago* (last edited 11 months ago)

The way it was done if I remember correctly is that someone found out v6 was trained partially with Stockbase images-caption pairs, so they went to Stockbase and found some images and used those exact tags in the prompts.

load more comments (8 replies)
[–] [email protected] 17 points 11 months ago (3 children)

Voyager just loaded a copyrighted image on my phone. Guess someone's gonna have to sue them too.

[–] [email protected] 16 points 11 months ago (3 children)

Yeah man, Voyager is making millions with the images on the app. It makes me so mad, they Voyager people make you think they are generating content on their own, but in reality is just feeding you unlicensed content from others.

load more comments (3 replies)
load more comments (2 replies)
[–] [email protected] 12 points 11 months ago (5 children)

When they asked for an Italian video game character it returned something with unmistakable resemblance to Mario with other Nintendo property like Luigi, Toad etc. ... so you don't even have to ask for a "screencapture" directly for it to use things that are clearly based on copyrighted characters.

[–] [email protected] 13 points 11 months ago* (last edited 11 months ago) (4 children)

you're still asking for a character from a video game, which implies copyrighted material. write the same thing in google and take a look at the images. you get what you ask for.

you can't, obviously, use any image of Mario for anything outside fair use, no matter if AI generated or you got it from the internet.

load more comments (4 replies)
load more comments (4 replies)
[–] [email protected] 69 points 11 months ago (6 children)

They literally asked it to give them a screenshot from the Joker movie. That was their fucking prompt. It's not like they just said "draw Joker" and it spit out a screenshot from the movie, they had to work really hard to get that exact image.

[–] [email protected] 69 points 11 months ago* (last edited 11 months ago) (74 children)

Because this proves that the "AI", at some level, is storing the data of the Joker movie screenshot somewhere inside of its training set.

Likely because the "AI" was trained upon this image at some point. This has repercussions with regards to copyright law. It means the training set contains copyrighted data and the use of said training set could be argued as piracy.

Legal discussions on how to talk about generative-AI are only happening now, now that people can experiment with the technology. But its not like our laws have changed, copyright infringement is copyright infringement. If the training data is obviously copyright infringement, then the data must be retrained in a more appropriate manner.

[–] [email protected] 29 points 11 months ago* (last edited 11 months ago) (16 children)

But where is the infringement?

This NYT article includes the same several copyrighted images and they surely haven't paid any license. It's obviously fair use in both cases and NYT's claim that "it might not be fair use" is just ridiculous.

Worse, the NYT also includes exact copies of the images, while the AI ones are just very close to the original. That's like the difference between uploading a video of yourself playing a Taylor Swift cover and actually uploading one of Taylor Swift's own music videos to YouTube.

Even worse the NYT intentionally distributed the copyrighted images, while Midjourney did so unintentionally and specifically states it's a breach of their terms of service. Your account might be banned if you're caught using these prompts.

[–] [email protected] 28 points 11 months ago (2 children)

You do realize that newspapers do typically pay the licensing for images, it's how things like Getty images exist.

On the flip side, OpenAI (and other companies) are charging someone access to their model, which is then returning copyrighted images without paying the original creator.

That's why situations like this keep getting talked about, you have a 3rd party charging people for copyrighted materials. We can argue that it's a tool, so you aren't really "selling" copyrighted data, but that's the issue that is generally be discussed in these kinds of articles/court cases.

load more comments (2 replies)
load more comments (15 replies)
[–] [email protected] 16 points 11 months ago (3 children)

Because this proves that the “AI”, at some level, is storing the data of the Joker movie

I don't think that's a justified conclusion.

If I watched a movie, and you asked me to reproduce a simple scene from it, then I could do that if I remembered the character design, angle, framing, etc. None of this would require storing the image, only remembering the visual meaning of it and how to represent that with the tools at my disposal.

If I reproduced it that closely (or even not-nearly-that-closely), then yes, my work would be considered a copyright violation. I would not be able to publish and profit off of it. But that's on me, not on whoever made the tools I used. The violation is in the result, not the tools.

The problem with these claims is that they are shifting the responsibility for copyright violation off of the people creating the art, and onto the people making the tools used to create the art. I could make the same image in Photoshop; are they going after Adobe, too? Of course not. You can make copyright-violating work in any medium, with any tools. Midjourney is a tool with enough flexibility to create almost any image you can imagine, just like Photoshop.

Does it really matter if it takes a few minutes instead of hours?

load more comments (3 replies)
[–] [email protected] 11 points 11 months ago (13 children)

I've had this discussion before, but that's not how copyright exceptions work.

Right or wrong (it hasn't been litigated yet), AI models are being claimed as fair use exceptions to the use of copyrighted material. Similar to other fair uses, the argument goes something like:

"The AI model is simply a digital representation of facts gleamed from the analysis of copyrighted works, and since factual data cannot be copyrighted (e.g. a description of the Mona Lisa vs the painting itself), the model itself is fair use"

I think it'll boil down to whether the models can be easily used as replacements to the works being claimed, and honestly I think that'll fail. That the models are quite good at reconstructing common expressions of copyrighted work is novel to the case law, though, and worthy of investigation.

But as someone who thinks ownership of expressions is bullshit anyway, I tend to think copyright is not the right way to go about penalizing or preventing the harm caused by the technology.

load more comments (13 replies)
load more comments (71 replies)
load more comments (5 replies)
[–] [email protected] 65 points 11 months ago (21 children)

Copyright issues aside, can we talk about how this implies accurate recall of an image from a never before achievable data compression ratio? If these models can actually recall the images they have been fed this could be a quantum leap in compression technology.

[–] [email protected] 34 points 11 months ago* (last edited 11 months ago)

It's not as accurate as you'd like it to be. Some issues are:

  • It's quite lossy.
  • It'll do better on images containing common objects vs rare or even novel objects.
  • You won't know how much the result deviates from the original if all you're given is the prompt/conditioning vector and what model to use it on.
  • You cannot easily "compress" new images, instead you would have to either finetune the model (at which point you'd also mess with everyone else's decompression) or do an adversarial attack onto the model with another model to find the prompt/conditioning vector most likely to create something as close as possible to the original image you have.
  • It's rather slow.

Also it's not all that novel. People have been doing this with (variational) autoencoders (another class of generative model). This also doesn't have the flaw that you have no easy way to compress new images since an autoencoder is a trained encoder/decoder pair. It's also quite a bit faster than diffusion models when it comes to decoding, but often with a greater decrease in quality.

Most widespread diffusion models even use an autoencoder adjacent architecture to "compress" the input. The actual diffusion model then works in that "compressed data space" called latent space. The generated images are then decompressed before shown to users. Last time I checked, iirc, that compression rate was at around 1/4 to 1/8, but it's been a while, so don't quote me on this number.

edit: fixed some ambiguous wordings.

[–] [email protected] 16 points 11 months ago (1 children)

You can hardly consider it compression when you need a compute expensive model with hundreds of gigabytes (if not bigger) to accurately rehydrate it

load more comments (1 replies)
[–] [email protected] 14 points 11 months ago

If you ignore the fact that the generated images are not accurate, maybe.

They are very similar so they are infringing but nobody would use this method for compression over an image codec

load more comments (18 replies)
[–] [email protected] 41 points 11 months ago (16 children)

God I fucking hate this braindesd AI boogeyman nonsense.

Yeah, no shit you ask the AI to create a picture of a specific actor from a specific movie, its going yo look like a still from that movie.

Or if you ask it to create "an animated sponge wearing pants" it's going to give you spongebob.

You should think of these AIs as if you asking an artist freind of yours to draw a picture for you. So if you say "draw an Italian video games chsracter" then obviously they're going to draw Mario.

And also I want to point out they interview some professor of English for some reason, but they never interview, say, a professor of computer science and AI, because they don't want people that actually know what they're talking about giving logical answers, they want random bloggers making dumb tests and """exposing""" AI and how it steals everything!!!!!1!!! Because that's what gets clicks.

[–] [email protected] 14 points 10 months ago

They interviewed her because she wrote about generative ai experiments she conducted with Gary Marcus, an AI researcher who they quote earlier in the piece, specifically about AI’s regurgitation issue. They link to it in the article.

[–] [email protected] 13 points 10 months ago (1 children)

All of this and also fuck copyright.

Why does everyone suddenly care about copyright so much. I feel like I'm taking crazy pills.

load more comments (1 replies)
[–] [email protected] 12 points 10 months ago* (last edited 10 months ago)

We asked this artist to draw the joker. The artist generated an copyrighted image. We ask the court to immediately confiscate his brain.

load more comments (13 replies)
[–] [email protected] 35 points 11 months ago (21 children)

I already know I'm going to be downvoted all to hell, but just putting it out there that neural networks aren't just copy pasting. If a talented artist replicates a picture of the joker almost perfectly, they are applauded. If an AI does it, that's bad? Why are humans allowed to be "inspired" by copyrighted material, but AIs aren't?

load more comments (21 replies)
[–] [email protected] 33 points 11 months ago* (last edited 11 months ago) (6 children)

This is a classic problem for machine learning systems, sometimes called over fitting or memorization. By analogy, it's the difference between knowing how to do multiplication vs just memorizing the times tables. With enough training data and large enough storage AI can feign higher "intelligence", and that is demonstrably what's going on here. It's a spectrum as well. In theory, nearly identical recall is undesirable, and there are known ways of shifting away from that end of the spectrum. Literal AI 101 content.

Edit: I don't mean to say that machine learning as a technique has problems, I mean that implementations of machine learning can run into these problems. And no, I wouldn't describe these as being intelligent any more than a chess algorithm is intelligent. They just have a much more broad problem space and the natural language processing leads us to anthropomorphize it.

load more comments (6 replies)
[–] [email protected] 27 points 11 months ago (6 children)

"Generate this copyrighted character"

"Look, it showed us a copyrighted character!"

Does everyone that writes for the NYTimes have a learning disability?

[–] [email protected] 37 points 11 months ago (71 children)

The point is to prove that copyrighted material has been used as training data. As a reference.

If a human being gets asked to draw the joker, gets a still from the film, then copies it to the best of their ability. They can't sell that image. Technically speaking they've broken the law already by making a copy. Lots of fan art is illegal, it's just not worth going after (unless you're Disney or Nintendo).

As a subscription service that's what AI is doing. Selling the output.

Held to the same standards as a human artist, this is illegal.

If AI is allowed to copy art under copyright, there's no reason a human shouldn't be allowed to do the same thing.

Proving the reference is all important.

If an AI or human only ever saw public domain artwork and was asked to draw the joker, they might come up with a similar character. But it would be their own creation. There are copyright cases that hinge on proving the reference material. (See Blurred Lines by Robin Thick)

The New York Times is proving that AI is referencing an image under copyright because it comes out precisely the same. There are no significant changes at all.

In fact even if you come up with a character with no references. If it's identical to a pre-existing character the first creator gets to hold copyright on it.

This is undefendable.

Even if that AI is a black box we can't see inside. That black box is definitely breaking the law. There's just a different way of proving it when the black box is a brain and when the black box is an AI.

load more comments (71 replies)
[–] [email protected] 23 points 11 months ago

The crux is that they went "draw me a cartoon mouse" and Midjourney went "here is Disney's Mickey Mouse™". A simple prompt should not be able to generate that specific of an image. If you want something specific, you should need to specific it, otherwise the AI failed to generalize or is somehow heavily biased towards existing images.

[–] [email protected] 20 points 11 months ago (1 children)

Or you do? The point is that these machines are just regurgitating the copyrighted data they are fed, and not actually doing all that transformative work their creators claim in order to legally defend feeding them work they dont have the rights to.

Its recreating the images it was fed. Not completing the prompt in unique and distinct ways. Just taking a thing it ate and plopping it into your hands.

It doesnt matter that you asked it to do that, because the whole point was that it "isnt supposed to" do that in order for them to have the legal protection of feeding it artwork they didnt pay the rights to.

load more comments (1 replies)
[–] [email protected] 17 points 11 months ago (1 children)

I'm pretty pro AI but I think their point was that the generated images were near identical to existing images. For example, they generate one from Dune that even has whisps of hair in the same place.

load more comments (1 replies)
[–] [email protected] 10 points 11 months ago (2 children)

It just proves that there is not actual intelligence going on with this AI. It's basically just a glorified search engine that claims the work of others as it's own. It wouldn't be as much of a problem if it attributed it's sources, but they can't do that because that opens them up to copyright infringement lawsuits. It's still copyright infringement, just combined with plagiarism. But it's claimed to be a creation of "AI" to muddy the waters enough to delay the inevitable avalanche of copyright lawsuits long enough to siphon as much investment dollars as possible before the whole thing comes crashing down.

load more comments (2 replies)
load more comments (1 replies)
[–] [email protected] 23 points 11 months ago (1 children)

I took a gun, pointed it at another person, pulled the trigger and it killed that person.

[–] [email protected] 15 points 11 months ago (3 children)

I opened the egg carton and found eggs in there.

load more comments (3 replies)
[–] [email protected] 20 points 11 months ago (7 children)

I suppose it's time to copyleft all the things on the internet

load more comments (7 replies)
[–] [email protected] 17 points 10 months ago (5 children)

I don't know why everybody pretends we need to come up with a bunch of new laws to protect artists and copyright against "AI". The problem isn't AI. The problem is data scraping.

An example: Apple's iOS allows you to record your own voice in order to make it a full speech synthesis, that you can use within the system. It's currently tooted as an accessibility feature (like, if you have a disability preventing you from speaking out loud all of the time, you can use your phone to speak on your behalf, with your own custom voice). In this case, you provide the data, and the AI processes it on-device over night. Simple. We could also think about an artist making a database of their own works in order to try and come up with new ideas with quick prompts, in their own style.

However, right now, a lot of companies are building huge databases by scraping data from everywhere without consent from the artists that, most of the time, don't even know their work was scraped. And they even dare to advise that publicly, pretend they have a right to do that, sell those services. That's stealing of intellectual property, always has been, always will be. You don't need new laws to get it right. You might need better courts in order to enforce it, depending on which country you live in.

There's legal use of AI, and unlawful use of AI. If you use what belongs to you and use the computer as a generative tool to make more things out of it: AI good. If you take from others what don't belong to you in order to generate stuff based on it: AI bad. Thanks for listening to my TED talk.

load more comments (5 replies)
[–] [email protected] 15 points 11 months ago

No it didn't.

[–] [email protected] 14 points 11 months ago (4 children)

For fun I asked an AI to create a Joker “in the style of Batman movies and comics”.

The Heath Ledger Joker is so prominent that a variation on that movie’s version is what I got back. It’s so close that without comparing a side-by-side to a real image it’s hard to know what the differences are.

load more comments (4 replies)
[–] [email protected] 14 points 10 months ago (1 children)

The fundamental philosophical question we need to answer here is whether Generative Art simply has the ability to infringe intellectual property, or if that ability makes Generative Art an infringement in and of itself.

I am personally in the former camp. AI models are just tools that have to be used correctly. There's also no reason that you shouldn't be allowed to generate existing IP with those models insofar as it isn't done for commercial purposes, just as anyone with a drawing tablet and Adobe can draw unlicensed fan art of whatever they want.

I don't really care if AI can draw a convincing Ironman. Wake me when someone uses AI in such a way that actually threatens Disney. It's still the responsibility of any publisher or commercial entity not to brazenly use another company's IP without permission, that the infringement was done with AI feel immaterial.

Also, the "memorization" issue seems like it would only be an issue for corporate IP that has the highest risk of overrepresentation in an image dataset, not independent artists who would actually see a real threat from an AI lifting their IP.

load more comments (1 replies)
[–] [email protected] 14 points 11 months ago (9 children)

Get rid of copyright law. It only benefits the biggest content owners and deprives the rest of us of our own culture.

It says so much that the person who created an image can be bared from making it.

load more comments (9 replies)
[–] [email protected] 13 points 10 months ago (8 children)

I think AI in this case is doing exactly what it's best at: Automating unbelievably boring chores on the basis of past "experiences". In this case the boring chore was "Draw me [insert character name] just how I know him/her".

Too many people mistakenly assume generative AI is originative or imaginative. It's not. It certainly can seem that way because it can transform human ideas and words into a picture that has ideally never before existed and that notion is very powerful. But we have to accept that, until now, human creativity is unique to us, the humans. As far as I can tell, the authors were not trying to prove generative AI is unimaginative, they were showing just how blatant copyright infringement in the context of generative AI is happening. No more, no less.

load more comments (8 replies)
[–] [email protected] 13 points 11 months ago* (last edited 11 months ago) (8 children)

Asks AI to generate copyrighted image; AI generates a copyrighted image.

Pikatchu.jpg

load more comments (8 replies)
[–] [email protected] 10 points 11 months ago (4 children)

I'm so sick of these examples with zero proof. Just two pictures side by side and your word that one of them was created (easily, it's implied) by AI. Cool. How? Explain to me how you did it, please.

load more comments (4 replies)
[–] [email protected] 10 points 11 months ago (20 children)

Copyright is a scam anyway so who cares?

load more comments (20 replies)
load more comments
view more: next ›