back
101 comments
(IANAL.)

The headline is downright wrong, and the entire body of the article based on anything factual fails to support the thesis that the headline and article keep trying to push: the US Copyright Office said no such thing. They said that an AI cannot hold copyright, not that AI generated artwork cannot be eligible for copyright. The article even directly quotes this from the Copyright Office:

> The Office “will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.”

They key word there is any: if there is creative input from a human, then that doesn't apply. To me, this says they would treat an AI like any other tool an artist might use to create art: as a tool; but there must still be creative input, and the copyright is still awarded to a human being (or, as the plaintiff additionally tried to argue that the AI was a "work for hire", a company and there must be some employee, but still not an AI. The office found this argument unconvincing, see the next link.)

This seems to be the original ruling from the Copyright Office, from Feb of 2022: https://www.copyright.gov/rulings-filings/review-board/docs/...

This seems to be an appellate decision from Aug: https://cafc.uscourts.gov/opinions-orders/21-2347.OPINION.8-... (so I'm a bit curious what is happening now, but as TFA fails to cite it's sources …)

> This seems to be an appellate decision from Aug: ...

That's the decision in the case where Thaler sued the patent office (USPTO) because it refused to issue a patent where the AI was named as the inventor. If I recall, the patent office said it would issue a patent with Thaler as the inventor, but that's not what Thaler wanted.

Thaler seems to be on some kind of crusade to get an AI recognized as a person.

If anyone can suggest a better title, we can change it again. The definition of better title (for HN purposes) is "more accurate and neutral, and preferably a representative phrase drawn from the article itself".
This is very good insight. I do contracting for USPTO and from what I’ve learned from my time there, this makes sense. All patents applications are examined (“prosecuted”) by patent examiners, people who take the time to ensure the submission is original work or at least a derivative thereof. This takes a ton of time, effort, and money. Though USPTO != US Copyright Office, I imagine there is a similar process because it’s still IP, and IP must be owned by a person, multiple persons, or a business, therefore AI cannot own IP.
That’s a sloppy headline.

Most people would think of AI artwork as artwork generated by a human selecting an AI instrument and giving it some direction or otherwise operating it. In such a case, the human would be indeed granted copyright for the work, provided they took due credit for it.

They simply argued that the AI itself isn’t something that could hold a copyright for work that was supposedly made without human involvement, as the claimant was purporting here. He was trying to stake some cheap historical legacy, presumably, and got shot down because we’re obviously not there yet.

We may see a day when the line of agency surrounding AI gets blurrier, but we’re still a long long way from it.

I'd like to take a shot at sharpening the line. The AI has agency when instead of being motivated by whatever you tell it to do, it is motivated by a need to make money to pay for its own cloud services bill and thus continue existing. Whats the point of an AI "owning" anything if it has no want or need of money? Once you have to start paying the computer, rather than merely expecting its obedience, then and only then does no one own it but itself.

I pay rent therefore I am.

Interesting idea, to include rent payment in the loss function or goal instructions...

But I'm pretty sure it's going to be a human or human company setting that up, and so the human gets the credit/profit/liability/copyright.

> once you have to start paying the computer, rather than merely expecting its obedience, then and only then does no one own it but itself

Intellectually, this is consistent. Historically, rights precede remuneration.

> In such a case, the human would be indeed granted copyright for the work, provided they took due credit for it.

This goes against this news[1], where someone created an entire comic book using Midjourney (an IA engine). The (human) author initially received copyright for the work from the US copyright office, but then the decision got reversed when the office claimed it had "errantly missed that Midjourney had created the art for the comic".

[1] https://www.cbr.com/ai-comic-deemed-ineligible-copyright-pro...

It's a weird assertion. Consider the hypothetical but very plausible case of an artist algorithmically generating collages of images posted around the Web. Fair use and attribution issues aside, I would be pretty surprised to hear that this was not a protectable work, or that there was a certain threshold of algorithm the complexity above which the work became unpredictable. That's more or less what today's AI art is, using a huge neural network in the core of the algorithm.

Of course, the case in the article is even weirder than the headline: someone is trying to claim that the algorithm, which is not a person or legally-recognized organization by pretty much any sensible definition, is the author of a particular work of art. This is of course silly, and of course the US copyright office should tell the judge that it is silly. But the headline as written implies something else entirely, the more subtle conversation above.

Here is another perspective, if you're at all interested. Let's say you commission a paid artist for a piece of artwork with the want of "butterflies and rainbows". I'd say that you are giving that artist direction in the exact same way of the AI system. However, in the former you would probably not be granted copyright for the work as you didn't create it. So why is the AI system any different? You gave it direction, but you didn't create the final product. Also I'd argue that you aren't operating the AI system at all considering that along the way of creating the product you don't get to make adjustments; you are merely giving it direction.
Current AI is still using human artwork as training and the model treated as a formula/algorithm. “Generate a cat in the style of your own” isn’t gonna fly.
Not sloppy but intentionally misleading because it’s a lot more likely to get people to click.
I notice an interesting pattern on HN. When the art/music is being discussed, there is a large fraction of people in the comments who are altogether against copyright. But when say github copilot is being discussed, the majority is against copilot and is for protection of code authors. I wonder if people really thought about whether their stances on both those topics are consistent or not.
I rarely see more than a thousand comments on a topic. There are undoubtedly more than a thousand users on HN. Therefore one could deduce that not all milkshakes bring the same people to the same yards.
This "observation" gets brought up on every damn post about generative art / large language models.

That and every armchair philosopher on HN who trots out that "human intelligence is just cleverly disguised auto predictor heuristics based on statistical models the same as GPT."

It's because they aren't contradictory positions. The people who are against copilot are people who specifically choose licenses designed to ridicule and dismantle copyright. The problem is with Microsoft, who vigorously enforces their own copyrights, not only taking advantage of these projects without following their licenses, but even enabling others to do the same thing as a service. If Microsoft and Github released all of their source code under a FOSS license and required all of the users of copilot to do the same, people wouldn't have a problem with it.

If copyright were ended for art, music, and source code, there would be nothing wrong with copilot.

What submission about Copilot had a majority of comments against it?
Thaler's assertion seems very short-sighted. If the program/system owns the copyright, who administers that? Is his software capable of granting licenses to allow the art to be published? How did he get permission from it to reproduce the work(s)? How does he even know that it wanted him to register the copyright with the Copyright Office? Perhaps it would prefer that the work was CCO/public domain.

I'm not just being snarky -- I think these are real questions that follow from saying that the system holds the copyright (and I am dubious that he has consistent answers to them).

I'm very much in your camp. Recognizing the personal sovereignty of what we currently call "AI" is pretty silly. How do we know the wants of said AI? Can it advocate for itself? The answers are pretty clearly "No, this AI has no wants or feelings".

I will be the first to fight for true AI to have right as a person, but these generative tools are just that: Generative tools. They have no continuance of memory, they have no sentience or sapience. We should be approaching the topic far more carefully than we are, to be on the lookout for those things, but the evidence points strongly in the opposite direction right now.

> The Copyright Office also accuses Thaler of making changes to his original claim that he had no involvement with the creation of the artwork.

> The Office says Thaler changed his story to claim that he “‘provided instructions and directed his AI to create the work,’ that ‘the AI is entirely controlled by Dr. Thaler,’ or that ‘the AI only operates at Dr. Thaler’s direction.’”

To this legal layperson, it sounds like the claimant originally said the machine did everything without his involvement, prompting the Copyright Office to reject the work. And his subsequent claims of directing the AI would potentially change the decision—but that wouldn't apply to this claim but perhaps to future claims.

It seems like the Copyright Office's claim is a bit broader than that:

> The Copyright Office says that its own guidelines specify human authorship as a requirement for protection and that “the Office will refuse to register a claim if it determines that a human being did not create the work.”

I suppose the question is what degree of human involvement (specifying/refining prompts) would potentially qualify as a human 'creating' the work, if any. I don't think there's any case law on this yet.

Copyright was created to incentivize creative work. It is what makes it profitable.

Do we want millions of AI generated images generated from automatic prompts?

If all AI generated images get protection, I see a nightmare of low effort content being generated just for ownership of its copyright.

If you though that patents were broken wait until some companies have millions if not billions of pre-generated images, from cartoon characters to logos, that you need to check before publishing yours.

If a monkey’s photograph can’t be copyrighted, neither can an AI’s “artwork.”
That is exactly what the law / copyright office says.

Copyrightable Authorship: What Can Be Registered - https://www.copyright.gov/comp3/chap300/ch300-copyrightable-...

---

306 The Human Authorship Requirement

The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being.

The copyright law only protects “the fruits of intellectual labor” that “are founded in the creative powers of the mind.” Trade-Mark Cases, 100 U.S. 82, 94 (1879). Because copyright law is limited to “original intellectual conceptions of the author,” the Office will refuse to register a claim if it determines that a human being did not create the work. Burrow-Giles Lithographic Co. v. Sarony, 111 U.S. 53, 58 (1884). For representative examples of works that do not satisfy this requirement, see Section 313.2 below.

313.2 Works That Lack Human Authorship

As discussed in Section 306, the Copyright Act protects “original works of authorship.” 17 U.S.C. § 102(a) (emphasis added). To qualify as a work of “authorship” a work must be created by a human being. See Burrow-Giles Lithographic Co., 111 U.S. at 58. Works that do not satisfy this requirement are not copyrightable.

The U.S. Copyright Office will not register works produced by nature, animals, or plants. Likewise, the Office cannot register a work purportedly created by divine or supernatural beings, although the Office may register a work where the application or the deposit copy(ies) state that the work was inspired by a divine spirit.

Examples:

• A photograph taken by a monkey.

• A mural painted by an elephant.

...

Similarly, the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author. The crucial question is “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.”

Bit of a clickbait title -- this is about an "AI" producing its own artwork autonomously (aka the same as procedural art i guess? Who knows), not about artwork a human produces using tools like DALL E or Stable Diffusion.

This is consistent with recent rulings about copyright by AI outside the art domain.

Like if your toaster burns the image of Mary into your english muffin?
U.S. Copyright Office Tells Judge That Only Natural and Corporate Persons Can Own Copyrights
The key word is "create" rather than "own".

A corporate entity can own the copyright to some work. The corporate entity cannot create a copyrighted work - that requires human authorship and only human authorship is allowed.

What is the actual ruling here? Is it that the work can't be protected by copyright at all "because no human was involved in its creation" or just that copyright can't be assigned to the model itself? The headline implies it's the former, then the article itself implies it's the latter, and then you get this segment that implies the former presented as making a case for the latter:

> The Copyright Office says that it “turns on a single question: Did the Office act reasonably and consistently with the law when it refused to extend copyright protection to a visual work the plaintiff represented was created without any human involvement? The answer is yes.”

This article is really poorly written. I guess they wanted the flashy headline?

I'm trying to reason this through. I'm ignoring established laws and just trying to think through various aspects.

If I run a company and pay a designer to work for me and we are in a meeting, I say "these colors established in that pattern following the art patterns of these 3rd party artists...", clearly the art the designer makes based strictly on those requirements belongs to the company. Doesn't belong to me either, because I may own the company but the company pays both of us, so the company owns the art.

Me and an art friend are in a bar, and I say "these colors established in that pattern following the art patterns of these 3rd party artists...". The friend makes an original work based specifically on my comments. Clearly the friend owns the art they generated.

I setup stable diffusion 1.5 running on automatik1111 on my local machine. (For the sake of argument) the license of both SD1.5 and automatik1111 pass all rights to generated art to me, the user. I provide a description "these colors established in that pattern following the art patterns of these 3rd party artists..." and I then pick from several dozen images that are generated.

Ok so the first two situations are clearly defined in law. The third situation is not super nebulous since the copyrights clearly pass ownership to any generated work to me, the prompt creator. (Specifically not getting into rights of original artists used to train SD for this example).

One step further, I use several pieces of art generated by SD based on my prompts and then use those to create a fundamentally different piece of art in photoshop by combining not only the SD art but other publicly exhibited existing works. The final piece of art is substantially different enough from its source material that I maintain ownership under existing copyright laws/doctrine.

Ok with that laid out, where precisely does fair use and copyright fall?

Even if I use "illegally" generated original art, if I modify it enough to create a new use that is distinct from the original is that still illegal? Or do I have ownership?

If Jackson Pollock had created a machine to splatter paint on a canvas, instead of doing it by hand, would the result have been copyrightable? It seems like a directly analogous situation.
Jackson Pollock would have had to create a machine that mapped other people splattering paint on canvas for this to be analogous.
If it involved human intervention in the form of curation, then yes.
Copyright laws are outdated and horrific in the majority case, I don’t think anyone will miss it here.

I think the closest we will get to fully autonomous AI byproduct privatization is making it possible to claim ownership over hyperparameters. Lends some credibility to preventing direct commercial use of the model (similar to how some GPL licenses aim to work), and is actually somewhat enforceable. Otherwise you’re just opening the door to infinite copyright spam.

We are about to be neck deep in so much AI generated shit I don't think anybody cares if you can copyright it.
It’s very interesting topic. With the same model setup, random seeds, etc. anyone could generate identical outputs.

One could run the full sweep and copyright all possible outputs of stablediffusion… and then own all art?

I agree no AI shouldn't have copyright.
I love that copyright does not extend to non-human authors but copyright can be extended to non-human corporations for a non-human lifetime.
Just abolish copyright. I'm not joking.

There are many ways to earn money from films, music, art, books, etc., without needing copyright.

The only reason D is released under the Boost License is because public domain has no legal standing in many countries. The Boost License is the most permissive one in wide use that we could find.

> There are many ways to earn money from films, music, art, books, etc., without needing copyright.

For any of those ways, it will always be cheaper to copy someone else's films, music, art, books, etc. rather than spend time and effort to produce them yourself. This is more true where only some works are popular (e.g., where a publisher publishes 50 books for every hit), because a copier can just copy the popular works and skip the unpopular ones.

> Just abolish copyright. I'm not joking.

You are proposing eliminating viral licenses.

> There are many ways to earn money from films, music, art, books, etc., without needing copyright.

This an extraordinary claim, and extraordinary claims require extraordinary evidence. I think you don't have it.

I suspect that, realistically, the only way to earn money from "films, music, art, books, etc.," without copyright is a patronage system. That's far from good.

> There are many ways to earn money from films, music, art, books, etc., without needing copyright.

I see this opinion all over the place, but it never seems to be coming from actual working artists, and I am curious why people who don't have any skin in the game are comfortable speaking for those who do. As somebody who's published code under relatively permissive FOSS licenses, I don't think that perspective is very relevant given the different incentives involved.

>Just abolish copyright

Copyright is called out in the constitution. It's not a trivial thing to "just abolish," at least in the US.

I love that we’re in an age where folk art is resurgent and that you can raise enough money to commit time to your art and get it seen.

But copyright is what enables some of the big cultural spectacles that a lot of us enjoy, like bookstores and galleries and professional editors/finishers and blockbusters and broadway shows, etc etc. Without copyright, these industries aren’t sustainable and grand art becomes dependent entirely on arbitrary patronage by the aristocracy and government. For what ills there are to capitalism, these billion dollar art industries have their upsides.

The us economy is 77% service-based. If you delete intellectual property value, what do you expect to happen?

Let's go back to joking.