Hacker News Reader

Comment by muststopmyths | Hacker News Reader

If the artist's license was for non-commercial use, isn't using it for training an AI for commercial use a clear violation ?

I would love to know why not.

View on HN

Fair use. Licenses like that are enforced under copyright, if the use is not subject to copyright, the license terms are irrelevant.

We should note that whether AI training is actually fair use has yet to be tested in court. In particular, there's some wording in fair use about the effect on the market for the original and whether the derivative work can substitute for the original that I think a lot of programmers who blindly say this is fair use ignore. I am not a lawyer though.

Making money from the trained model isn't fair use anymore.

No, fair use does not preclude making money.

That's an over-simplication. Fair use does not necessarily preclude the user from making money, but it definitely can if that action diminishes the value of the copied work.

If we talk about the OpenAI magnitude of money it's definitely precluded, that's way beyond fair.

Lol what? That's not how any of this works.

>Commercial use does not automatically make the use copyright infringement, but it is an important deciding factor.

https://saltiellawgroup.com/2021/10/%E2%80%8B%E2%80%8Bcan-yo...

Isn’t this only the case if you give credit though?

That has to be legislated or written into law in the courts.

I'm not sure if it applies here. Midjourney's use intends to substitute the artists' work, and is commercial.

Unless it's a 1 for 1 replica I don't see how it's a substitute in the copyright sense.

We already have laws covering close derivatives - in this case it might not be clear who is the inspiration but there is no novel creativity... everything these networks produce is derived from artists - there is no added novelty.

Imagine how rich you would be if you were the first artist to make a red health bar in a video game.

I think that's exactly the sort of flawed analysis these copyright trolls do. They see something they never would have had a hand in and count it as "theirs", and then greed kicks in. It's like the old MPAA/RIAA calculations of lost revenue that assume someone would have bought a CD if they didn't download a song. Only worse because at least in the downloading case, they are a 1:1 substitute.

In almost all cases I'd guess that artists whose style is being copied by many genAI users are getting way more exposure than in some alternate universe where midjourney didn't exist. But nobody looks at it that way, they just see somebody else making profit and think they deserve some.

it's derived from the original and competes with the original

if it was parody or commentary or otherwise found a new market, it might be allowed under fair use, but the fourth factor of fair use concerns the effect on the market value of the original

That's the question that is currently working its way through the courts in many different cases. We don't yet know the answer.

No one is using Midjourney to replace artists. Artists are using Midjourney to speed up the most tedious parts of their work.

The term is used here in its legal meaning, therefore as competing in the same space as the original

Absolutely not true. Non-artists use it to replace the need to purchase art and art services.

That said it is unclear if the connection is strong enough between a given artist and their replacement.

That's just BS. If people weren't replacing artists, there would be no need to prompt with "in the style of ___". are you honestly saying that only artistX is using the prompt "in the style of artistX"?

Making a picture in the style of an artist is not replacing them. I don't know why anyone could get that idea. I can sing in the style Janice Joplin, pretty sure I'm not replacing her and that nobody thinks I am.

Nobody cares what you think you sound like in the shower. Nobody

If I can make a voice over that sounds like Neil DeGrasse Tyson, why would I pay him to make the voice over for me? If I can make artwork in the style of X why would I commission X to make that artwork instead? Nobody cares if it actually was made by them as long as people think it was made by them. So "them" gets screwed because nobody will use them. We're obviously talking in very strict ones and zeros here with "nobody". Except for the nobody caring what you think you sound like

Singing in the style of someone is a bit different than commanding a few promots to make up lyrics and AI sing them in a specific style. The former implies some effort, the latter is just effortless. That effort not being needed nor rewarded would eventually lead to much fewer taking the road to any type of artistry. Maybe that’s the next evolutional step culturally or maybe it a big loss.

Media companies don't care if you think it's replacing her or not. They care that (1) they made something that will sell and that (2) they made it cheaper than they did before.

All content will be behind a subscription and the algorithm will say "you like Janis Joplin? Let's autoplay some close-enough AI artists that don't exist that we won't have to pay any royalties to!" You'll get an offer to upgrade to Double Pro+ for a special introductory rate to skip those tracks.

I know you're really really sad that one person can't hold all similarities to their chosen likeness in copyrighted perpetuity, but it sounds like you have a problem that things are SIMILAR and the artist isn't getting paid for any adjacent likeness instead of content itself being directly stolen. Regarding that topic, you should be able to find my morally, ethically and legally backed response via midjourney by prompting it to create the world's smallest violin.

This discussion makes people immediately place themselves under the boot of technology and display active helplessness almost entirely because they can't handle the mentality that it's now a commodity to get what they consider a "unique artstyle". That you can be "unique" if you consistently draw a line in thicker shades around borders and use flashy colors in a way no one has decided to try yet, and that this style "belongs" to you being extrapolated to the point of your artistic hobby becoming the only reflection of your identity as a person. Sad for those who staked their lives around capitalistic ideals.

I don't think you replied to the right comment. Or you forgot to read and comprehend mine and just answered some phantom you were arguing with in the shower.

You're right. Instead of asking for art in the style of Van Gogh, I totally should have commissioned him to do it instead. I've replaced Van Gogh!

We're a stone's throw from publishers no longer needing illustrators. That job is going to be gone. It'll be the cheapest choice, so the market will make it the only choice.

My children's children are going to have a few classic books that were illustrated by humans, but anything new or reissued won't be.

Magazines, articles, and covers that might have featured human illustration will be gone. It'll be too easy to throw pennies for "draw something topical to article title in the style of artist that appeals to target demographic".

Direct duplicates are fair use as they are evaluated for criticism and the novelty of it being an actual duplicate at all

Everything else is a derivative work which falls outside of copyright claims

How would you like the law to change? That’s part of the equation too

Of course there are models trained only on licensed materials already, which reaches the same result of the material creators obsolescence

Fair use is a defence you can use when you infringe copyright [edit for clarity] or in other words the action you take would otherwise infringe.

It's not fair use because you want it to be, and it's not at all legally clear if this defence is valid in the case of AI training. But it's not clear it isn't, either.

This is basically what all the noise and PR money is about, currently, in hope that shaping the narrative will shape the legal decisions.

Fair use is not copyright infringement! It’s a limitation placed on copyright to balance the interests of copyright holders with the public interest.

Of course this all depends on jurisdiction, but e.g. in the US for example fair use does not show up [oops, see note] in the limitations and exceptions in the copyright definition (like e.g. government use does), but relies on case law, the "fair use doctrine". This means quite literally it is a framework for understanding the defence of a particular use based on precedent: the defendant(s) argued it.

It's a bit of a quibble to differentiate between allowable infringement or exception, here, so i've edited original.

[note: 0xcde4c3db points out correctly that it does actually show up since 1976, but in a way the defers much of the definition back to case law, so same effect]

I think it makes a difference, because it shifts the burden of proof. If fair use is a kind of infringement, then it’s up to the “infringer” to prove fair use. I think that’s ethically and legally backwards: the copyright holder should have to prove that the use was not fair.

Ethics aside, and again US specific, but the supreme court literally defines it as an affirmative defence.

Even this is a bit murky though, as a district court has argued it goes beyond that.

Maybe there's some nuance of your description that I'm missing, but fair use is codified as a limitation to copyright in US statute (although not until 1976) [1]. Application of it to any given situation is still heavily dependent on case law, though.

[1] https://www.law.cornell.edu/uscode/text/17/107

Fair enough - I had forgotten that quirk in the US situation, it was added to statute in a way that doesn't really define it, so you are back to case law.

Fair use is a legal defense that requires you to pass multiple tests. It's not a "thing" that you can just assert all AI training is and expect people to believe you.

and it seems unlikely to benefit society to let Microsoft and Google to hoover up 100% of creative works, then use that to compete with everyone in that industry to such an extent they can no longer make a living

The fair use defense in this situation is the equivalent of the YouTuber who puts "No copyright intended! I own nothing!" in the caption.

I don't think its fair use.

https://fairuse.stanford.edu/overview/fair-use/what-is-fair-...

There are complications, but google can use thumbnails because essentailly they are used to "review" the website.

Has google sampled and hosted the whole image on their own website and made more iamges in the style of say mickey mouse, they would have been taken to town by the owners.

This is why there are no commercial movies on youtube (without an explicit agreement) and why DCMA takedowns exist.

There's really no reason for anyone to be confident that this is or is not fair use. The law around fair use is a quagmire of conflicting ideas, cases, and opinions and most of the law is in the form of precedents since the letter of the law is so vague.

One only needs to read the Supreme Court decision in Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith. In this case, it was found to not be fair use, but with a forceful dissent from Elena Kagan. And this is just a straight-up use of a photograph / art-piece. Add technology to the mix and things get really messy quickly.

in that case, the source image was actually licensed:

"Years later, Goldsmith granted a limited license to Vanity Fair for use of one of her Prince photos as an “artist reference for an illustration.” The terms of the license included that the use would be for “one time” only. Vanity Fair hired Warhol to create the illustration, and Warhol used Goldsmith’s photo to create a purple silkscreen portrait of Prince, which appeared with an article about Prince in Vanity Fair’s November 1984 issue. "

So the somewhat necessary followup question: if Midjourney or ChatGPT includes a license in front of its model weights or access to its model that forbids using those weights or output from its model to create a competitor, is it fair use to ignore that license?

In both cases, we're relying on copyright. The companies are saying that you can access the models under a license. It's not too hard to circumvent that license or get access through a third party or separate service: but doing so would obviously be seen as deliberate circumvention. We can easily compare that to an artist putting up a click-through in front of a gallery that says that by clicking 'agree' you agree not to use their work for training. And in fact, it's literally the same restriction in both cases: OpenAI maintains that they can use copyright to enforce that people accessing their model avoid using their model to train AIs. Paradoxically, they also maintain that artists can not use copyright to block people accessing their images from using those images to train AIs.

Facebook has released model weights with licenses that restrict how those model weights are used. But I can get those models without clicking 'agree' on that license agreement. They're mirrored all over the place. We'll see what courts say, but Facebook's public argument is that their license doesn't stop applying if I download their software from a mirror. So I find it hard to believe that Facebook honestly believes in a fair use argument if they also claim that the same fair use argument doesn't apply to their own copyrighted material that they've released online (assuming that model weights can even be copyrighted in the first place).

This is one of my biggest issues with the fair use argument -- it's not that there's nothing convincing about it, in a vacuum I'd very likely agree with it. I don't think that licenses should allow completely arbitrary restrictions. But I also can't ignore that the vast majority of companies making this argument are making it extremely inconsistently. They know that their business models literally don't work if competitors figure out ways to use their services to rapidly produce competing models. If another company comes up with a training method to rapidly replicate model weights or to augment existing models using their output, none of these companies will be OK with that.

None of these companies believe that fair use invalidates license terms when it comes to their own IP.

I think companies like openAI, meta and midjourney (if they say you can't train on output) are hypocrites, and probably have no standing to enforce some of the restrictions they try and make. That doesn't change fair use. There are also many truly open source models that explicitly give the users freedom to use them as they want, and I'd prefer more providers license their's like that, ie Apache 2.0 or MIT.

Definitely it doesn't change the fair use argument, but it should change the way we talk about these companies and the way we contextualize what they're arguing and the way we talk about the surrounding debate. We should be up-front that there are plenty of examples where people are not granted these fair use rights and that fair use is often unequally applied to IP law.

And we should be pushing hard for the entire package (not to say that you're not, more just pointing out that this is a package deal) -- I am not a copyright maximalist, but I especially don't want a model where only artists and individuals are subject to copyright restrictions. I'm of the opinion that it is good for these companies to take a credibility hit for their hypocrisy and for us to make it clear that the fair use argument has to be something that applies to everybody. I hope that this hypocrisy is something that comes up in court if these lawsuits go through, and ideally I'd like it to become a sticking point where trying to have their cake and eat it too does hurt these companies' legal arguments.

I want limited copyright across the board, but a system with maximum fair use for companies like OpenAI and minimal fair use for consumers and individuals is worse than a system with expansive copyright for everyone. I think it's a good thing for companies like OpenAI to be unable to make selective arguments. We shouldn't treat their end goals as irrelevant.

It's also very true that there are open models that take a permissive approach, but those models are in my experience usually not the primary target of most criticism (outside of the "open models are dangerous" FUD). The linked comic explicitly calls out "profit-making companies that own the most-used channels of communication." I'm not sure the artist would phrase it this way, but I think reading between the lines the artist is very likely coupling together ownership with their complaint -- they are talking about the injustice of having their work appropriated into a form that is locked down and controlled and that exists for the benefit of the profit-making companies who want to own yet another channel of communication (in this case artistic communication itself). I suspect that part of that feeling of injustice does directly come from an understanding that as an individual anyone would be sued into oblivion if they tried to make the same arguments in the opposite direction.

Does that get rid of the fair use argument? No. But it is impossible to actually talk about the fair use debate in a useful way while ignoring the context of that debate. To just mention fair use and nothing else is leaving out a ton of context about why artists are mad, and about why this feels unjust to so many people, and about the regular inconsistencies in how these arguments are often applied. We don't have to get rid of fair use arguments, but I personally think it's more responsible to include that context. I worry that leaving that context out is playing directly into the hands of companies that are trying to create an outcome of selective enforcement for IP.

----

It's a bit of a sidenote, but on the subject of truly open models, I will say that it is great that those models are licensed the way that they're licensed, but I'd still rather have legal recognition that algorithmically generated weights are not subject to copyright at all. It's nice when companies are permissive, but it also subtly implies that they have a choice whether to be permissive. As someone who favors expanding fair use, I would prefer that they not have that choice. I disagree with the application of any license to model weights, permissive or not. I think at best those model weights could be subject to a database license, maybe, but I'm honestly a little skeptical of even that: I think that model weights are a factual and reproducible artifact of training, not a creative work.

Fair use is an argument to explain why you violated the license/copyright. It relies on the fact that you did make a copy without permission.

Fair use is not a copyright violation! It’s a limitation placed on copyright to balance the interests of copyright holders with the public interest.

It's seems just like a poor choice of words and semantic argument around the term "violate". Fair use is an affirmative defense — you concede you did this thing that would be illegal if not for this other extenuating information/circumstance. This is similar to claiming self-defense in a murder case — yes I killed that person, but it's not murder since I can show it was self-defense.

In this case it would be 'yes, I used this copyrighted work in a way that would be infringement were it not a fair use, but here's why it's fair-use.'

Thank you, yes - affirmative defense is exactly what I was trying to think of but couldn't remember.

Fair use, if it does indeed apply, doesn't care about license

I doubt it's still fair use if you become one of most valuable companies, hence the word "fair" in fair use.

It's a specific legal term - so the everyday meaning of the words isn't relevant as far as I understand it.

There is a reason why the word was chosen, in essence, it is about the meaning of the word

The fair in fair use has nothing to do with the size of a company. If Disney produces a satire of Penny Arcade, that's still fair use of Penny Arcade's IP.

Satire is always linked to the object of the satire, that's he fair part.

LLM text and images aren't.

The value of a company has nothing to do with whether copyright applies.

Unless the value of the company is based on one product with alleged copyright infringement.

OpenAI has one asset: GPT.

>Commercial use does not automatically make the use copyright infringement, but it is an important deciding factor.

https://saltiellawgroup.com/2021/10/%E2%80%8B%E2%80%8Bcan-yo...

That says nothing about the value of the company.

If it's an infringement when OpenAI does it, it would be an infringement when a company with 1/1000th the value did it.

And if so, should that non-commercial-use qualifier extend to works co/created with that AI?

Reply on news.ycombinator.com