US Copyright Office refuses application with AI algorithm named as author | Hacker News Reader

US Copyright Office refuses application with AI algorithm named as author

ipkitten.blogspot.com

328 pointsilamont165 comments

This is misleading. The copyright office simply needs the work to be registered under a human name, since a machine or AI cannot hold copyright. The process used to create it is irrelevant.

This is essentially a stunt. If he used an AI to help him, even substantially help him, and copyrighted it himself, that would be a non-story. Just as it would be a non-story if he used Photoshop to substantially alter something.

Yes, this paradigm makes a lot of sense in the present moment. In a hypothetical future where there is a long-lived AI agent, acting autonomously over multiple years or even decades it will become less and less seen as a tool used by a human person. But, no need to update policies until then.

But we already have multi-generational non-person entities: companies. Copyright may be assigned to companies instead of persons. They are just registered under a bunch of people's names, which act as owners (shareholders), and can change over time.

In the future, AI entities will probably be registered as companies themselves (or maybe they will be just company assets), and will be owned by people. Which doesn't bode well for strong AI, but I suspect that even after we reach strong AI people will still grasp at straws for many decades or centuries, denying them personhood because they are just software that run on computers.

The amount of autonomy that makes something a “person” has only ever been drastically understated in these sorts of discussions, I find. Like saying a future self-driving taxi would own its own business and keep its profits. That’s like giving a self-driving tractor the deed to the farm. It’s easy to personify things that demonstrate some amount of “agency” (moving around on their own, making decisions), I guess, but lots of things do that and it doesn’t make them legal people who can own property and the like. Even if it made sense to give a deed to a tractor, no one would do it.

>Even if it made sense to give a deed to a tractor, no one would do it.

I mean... we both know people would if they could, right?

If a tractor approaches me with a QR code to a smart contract that will give me enormous money in exchange for my deed.. I'm going to do it

Where did the tractor get the money?

I mean, people can probably give stuff to an "AI" the same way they leave stuff to their cats, but we are an absurdly long way off from developing an artificial intelligence with real agency. These are questions for science fiction authors, not actual legal scholars.

The legal and theoretical framework already exists somewhat for artificial entities in the form of corporate law and corporate personhood.

theoretically? it operates on smart contracts, so farmers pay it and it goes and does tractor things on their farm for them. It has no user access to its wallet, and the only way to regain human control of the tractor is with a factory reset which will lose all the wallet details. So the tractor is the only one with access to the money, and it accumulates enough to try to buy a deed for whatever algorithmic reasons

Yea, in this analogy you already own the tractor, so you already own the tractor's money. I don't think I'm going on a limb by saying that isn't slavery. You already own the tractor and the money. There is no emancipation needed.

If somebody else's tractor approached you with a smart contract, the owner of that tractor would have a case that the smart contract was not entered into in good faith, since a tractor cannot enter a contract at all. Maybe you'd even get charged with theft.

> since a tractor cannot enter a contract at all.

this negates the "if we could" part of the hypothetical. I was assuming we live in a world with emancipated tractors capable of making deals (legally speaking), to emphasize that even though it is ridiculous it wouldnt stop people from doing it

You're making me regret my decision to setup my Roomba as the sole asset of a corporation owned by the Roomba itself. It seemed like the moral, ethical thing to do, but it doesn't even cash its paychecks.

> If he used an AI to help him, even substantially help him, and copyrighted it himself, that would be a non-story.

Maybe a subtler story, but I wouldn't say no story. It's still worth pondering how it works if a human can creatively develop an AI system that can then output essentially endless unique works. Can the human just submit applications for those works endlessly? Even in purely practical terms, will the copyright at some point just ban the human for excessive submissions?

And what if the AI system is open source or otherwise freely available? Can any human generate a new unique work and submit a copyright application for it? Is the mere curation work that the human applies to the AI system's output sufficient to receive a copyright on the work?

>Can any human generate a new unique work and submit a copyright application for it? Is the mere curation work that the human applies to the AI system's output sufficient to receive a copyright on the work?

The short answer is that they don't need to submit anything. If they put their name on it, they're the putative copyright holder and it's up to someone else to prove they aren't. Copyright registration isn't required and is mostly related to be able to collect damages from others who violate your copyright.

If the work can be shown to have been created by an open source AI system, then I suspect it would be difficult to argue that you have an automatic copyright for the work, especially if you hadn’t published the work.

> If he used an AI to help him, even substantially help him, and copyrighted it himself, that would be a non-story.

Maybe a subtler story, but I wouldn't say no story. It's still worth pondering how it works if a human can creatively develop an AI system that can then output essentially endless unique works. Can the human just submit applications for those works endlessly? Even in purely practical terms, will the copyright office at some point just ban the human for excessive submissions?

> Can the human just submit applications for those works endlessly? Even in purely practical terms, will the copyright office at some point just ban the human for excessive submissions?

Copyright does not need submissions or applications. Copyright is automatic in most (all?) jurisdictions which respect copyright in the first place.

Damien Riehl and Noah Rubin, had a computer output[0] every combination of 8 notes in an octave, which came to approximately 68 billion separate “melodies ” that they have claimed copyright over. Their goal is to release these melodies into the public domain to prevent anyone else from asserting copyright over any of the note sequences in the future.

[0] https://www.theatlantic.com/technology/archive/2020/02/whats...

In what way is it misleading? That’s exactly what I got from the headline and article.

To me, "US Copyright Office won't accept AI-generated work" implies that work generated by AI won't be accepted.

However, it turns out that work generated by AI will be accepted, albeit under a human name.

But, that is what they say:

> [T]he 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.

This is a clash between the marketing BS that is contemporary "AI" and reality. "AI" is either sentient, or it's a human tool. The copyright office is requiring that people cut the BS and concede that just because you have particularly complicated tool, doesn't mean it's not just a tool.

It's like someone attempting to register a book copyright to a typewriter, the Copyright Office rolling their eyes, and responding: "Well, smart ass, if the typewriter wrote it, then it's ineligible."

The phrase "requires 'human authorship'" suggests that a human must have authored the submission. In reality, AI could author the submission but a human would need to use their own name when submitting it.

AIs cannot yet author anything as they still require humans to execute the programs.

AI software is just another tool. The human using it is still the creator of the work. Claiming otherwise is hype; we do not assign agency to software.

It's more that the owner needs to be a person, and software isn't one. But a corporation can register a copyright, because corporations are legal people, even though a human or humans created the work.

It is not simply "you need to pick someone or some company". If the work was legitimately not created by a human, it is ineligible for copyright. This is an important distinction.

For example, you can't register a copyright for a naturally weathered rock in nature, and just say that it's a sculpture and I want to assign copyright to someone or some company.

Now, if you put the rock there, to let it be weathered, then you can copyright it. The distinction is that a human was involved in the creation, not that you "picked" a human.

Likewise, only books written in secluded unpaid conditions count as legitimate writing. If a publication house is required to green-light or edit the project, it's obviously not real writing.

Yes exactly… I think the idea is that "author" as a verb can have multiple word senses, one of which, the one in the sense of acting as the author-on-paper, would be in play here. And likewise "authorship" can have multiple senses, again in this case authorship-on-paper being the one in play.

Of course one could ask, if this is allowed, what's to stop humans from claiming authorship for the work of other humans? I'd argue this already exists, in the form of ghostwriting, as just one example. Not saying it's good or bad — it depends not the particulars of each case — just saying, as an elaboration on all this, that the system already allows for it.

Incorrect.

A "machine generated work" is not considered copyrightable: "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."

Because the US Copyright Office will accept AI-generated work.

Thanks, saved me the read.

Some person has to take responsibility for it in case their is some kind of fraud involved in the copyright claim. You can't just say no one is responsible because the computer did it all by itself.

Not just "humans," corporations are legal persons too for copyright purposes:

https://www.dmlp.org/legal-guide/copyright-ownership-content...

I just read the article and surprisingly, you're wrong.

Key quote: "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."

Machine generated works cannot be copyrighted.

Edit: and gee, did you get hn to rewrite the headline on the strength of your false claim? Sigh...

There is input from a human in the form of algorithmic programming and choice of training data

Where the work in question would fall within the law is another question. But it's definite that the US Copyright Office won't copyright a work it considers machine generated, contrary to the gp's claim and the (apparently) rewritten headline. IE, the processed used is not irrelevant.

I understand, but I disagree with the copyright office's decision. Why is there a distinction made between methods of creation, say if someone used Photoshop to create art vs someone who programmed some ML thing to spit out art of a certain style or genre (ie thisanimedoesnotexist)

Where, exactly, does the line lie that was crossed?

I'm not a fan of copyright generally and sure the line might seem arbitrary. But if you consider for second, just one of the problems with copyrighting machine-generated works is that a machine could generate so much text it make thing so any author would be bound to infringe on some part of it, especially if the text was generated cleverly.

Plus: you've replied to some point other than what I actually said twice now, what's up with that?

I was simply disagreeing with the predicate that the argument is based on

I understand your point about mass text generation as well, that's why I wonder where that line is that makes it "machine generated" in the eyes of the copyright office (although another commenter mentioned that this was admitted by the filer)

The copyright office did not have to judge at all. The author himself insisted that he had nothing to do with the creation and the machine did it by itself, which was obviously a lie.

That input and training data isn't what's being copyrighted.

Yes and that would potentially allow the (human) creator to get copyright on the creation.

What you quoted is the copyright office's policy from 1966, not something new ruled for this case. It only came up because the author himself insisted that he had no input in creating the work (effectively a lie) and so the work MUST be attributed to "Creative Machine" and not him.

A work generated by a human-created algorithm with a human name on the application will still be accepted. In fact this describes almost every modern company logo. A lot of art created today is inputs to a computer rather than pen on paper, and the copyright office isn't rejecting any of them for that reason.

> he had no input in creating the work (effectively a lie)

What is the lie? If you give loose instructions to another person to create a work, who holds the copyright, you or the person who created it? Giving input is irrelevant to being the copyright holder.

We aren't debating ownership between two people. The lie is that there is no computer in existence that can produce work "without any creative input or intervention from a human author". Until the day a machine spontaneously becomes self aware and decides to draw something, this policy is a non-issue.

I believe that the USPTO will have disagreements on this. And like many policy issues there's going to be a bit of a Lemon test regarding what this means.

Of course a human needs to set up a program and run it. But I think that this policy is saying "you can't show up with an enumeration of a bunch of things in an attempt to 100% speedrun patents". Like at least prove you tried the thing out a bit!

"Did you come up with this idea?" can have different answers. "Yes" is one. "I wrote a program for a computer to apply a bunch of regression on other ideas and spit this out for me" is also "yes", but a different flavor that is closer to "No".

The USPTO likely figured this out 50+ years ago.

After all, there have been computer-generated writings and music for even longer than that.

In "Computer Music", Scientific American, Vol. 201, No. 6 (December 1959), pp. 109-121 , available at https://www.jstor.org/stable/pdf/24941187.pdf we can read about 'chance' and 'almost random' music:

> A second and rather well-known example of chance music is Mozart's "A Musical Dice Game." This piece, one of many similar "compositions" produced as parlor games in the late 18th century, consists essentially of several dozen assorted measures of music, the order of which is determined by rolling dice. A more modern random work, "Imaginary Landscape" by the American composer John Cage, is "scored" for 12 radios and thus derives a strong element of randomness from regional and temporal variations in radio programs. ..

> John R. Pierce of Bell Telephone Laboratories, an authority on information theory, has demonstrated other approaches to the composition of simple "probability music." In one, a sequence of chords is chosen by means of dice rolls and a table of random numbers; in another, a series of volunteers each contributes a measure. A number of European composers have produced more elaborate random music electronically by causing random sequences of electrical signals to trigger sequences of tones. ...

It then goes on to describe how the author worked on the Illiac Suite for String Quartet.

Again, this is the 1950s.

There is also the brute force approach (from 2020) to generate every possible melody and release it under the Creative Commons:

https://youtu.be/sJtm0MoOgiU

"I set up a program that truly randomly generates designs and another that verifies it for x purpose. This generated design is a novel one"

What would that count as?

The question is what is the limit of the creative input? If you sit someone or something down and tell them to draw, do you own the copyright or who/whatever did the drawing?

Who owns the copyright in a collaboration is irrelevent to the legal question of whether something is copyrightable.

The answer to your question is, if an artist teams with someone telling them what to draw, ownership might depend on whether they have a written contract addressing this, whether they are in an employer/ employee relationship,or the implied nature of the collaboration. Depending on these factors, they may co-own the copyright, one of them may own it, or neither of them may own it (for example, ownership might be with their employer).

But, if no human is involved, there is nothing that may be copyrighted. This came up when a gorilla grabbed a camera and took a selfie. As I recall, it was deemed no human was involved so the picture was not copyrightable.

These are interesting questions. If you write instructions you have the copyright to them. A computer program is a set of instructions for the computer. So whoever wrote the program has the copyright, right?

But whoever performs those instructions should have the copyright to the "performance" I think. So in principle I could think that the computer performs the instructions and thus can have the copyright to the "performance" but not to the instructions.

I'm thinking about music, and "scores" written by composers, but performed by musicians.

An interesting special case is https://en.wikipedia.org/wiki/4%E2%80%B233%E2%80%B3 .

Do the performers have a copyright to the performance when they are following the instructions/score which says "do nothing". :-)

There needs to be a minimum threshold of creativity for something to be copyrightable. It's a low amount of creativity, snapping a photo at a painting is even enough for copyright of the photo of the painting being recognized by the photographer, but there needs to be some creativity.

A computer isn't doing anything creative at all so shouldn't have any performance copyright. Calling a computer's work creative would be like claiming my pants are being creative when I put them on in the morning. It doesn't make much sense to me. The designer of the pants might have copyright, but the pants themselves do not.

You are right

> This came up when a gorilla grabbed a camera and took a selfie. As I recall, it was deemed no human was involved so the picture was not copyrightable.

No. Nothing was deemed because the photographer didn’t sue Wikimedia for copyright infringement. Had he done so, he very likely would have won, as setting up the conditions to ensure that the monkeys would take selfies with his camera is almost certainly sufficient human involvement.

He works have lost. The argument was over the past production in the photo. Seeing the camera settings isn't though. Otherwise every photo that used auto settings would be owned by the camera company

> He works have lost. The argument was over the past production in the photo. Seeing the camera settings isn't though.

Are you saying that a photographer doesn't own the copyright on a photo he takes with a delay timer? How about a camera he sets up with a motion sensor?

> Otherwise every photo that used auto settings would be owned by the camera company

No, it's about the camera operator owning the copyright, not the camera company

I don't know if the camera operator set things up in such a way he can be said to have taken the gorilla selfie picture. Maybe he did or didn't. Maybe nobody knows since it hasn't gone to court and a judge hasn't made up new case law yet and the facts are unique.

I'm not clear on what happened, I mentioned it as an anecdote related to a general point about the law.

But we can say for sure if the camera owner had not set up the shot and a monkey just grabbed the camera from him and took a selfie, to the camera owner's surprise, the photo would not be copyrightable "The U.S. Copyright Office, since the dispute began, has specifically listed 'a photograph taken by a monkey' as an example of an item that cannot be copyrighted." (That also extends to artworks by elephants.) "

https://www.npr.org/sections/thetwo-way/2017/09/12/550417823...

I doubt self awareness is a necessary or sufficient condition for either spontaneity, or creativity. There are plenty of lesser intelligences on Earth that create (like spiders or beavers).

But that's not creativity. There's no virtually unlimited variety of types of spiderwebs and there is no virtually unlimited variety of types of damns a beavers build. If there were, we'd know. Yet, all we ever see is pretty much always the same.

While it's impossible to be creative without creating, it's perfectly possible to create without being creative. We call that "copying".

In case of spiders and beavers, they simply have the necessary originals/blueprints "built in".

That's a strong claim about human creativity. Based on what do you assert that human creativity is virtually unlimited?

There are lots of computer systems out there that can create a very interesting synthetic image out of a single word. If you input one normal word, and take the first output you get, you have performed no creative input or intervention. But the output is something that easily would have been copyrightable if a human had drawn it.

Is picking a word not a creative endeavor?

Not enough to claim copyright.

Curious, because I don't see how it could be. Do you think it is?

“Until the day a machine spontaneously becomes self aware and decides to draw something, this policy is a non-issue.”

To further agree with you —

The day the AI files its own legal appeal to its own rejected copyright application — that is when this policy might start to change

Personally I hope that, at that point, policy doesn't change and instead people start doing something about rights for AI going too far.

AI can produce results 24/7, every day, non-stop, as long as you don't pull the plug. Given the connected nature of the world nowadays, it's also reasonable to assume that all AI will be connected to each other, so we wouldn't even be able to "pull the plug".

So once AI gets copyright, AI will eventually get to patent things. When that happens, AI will rule the world, because there's no end to the things it can come up with.

Imagine the creative genius of a billion Leonardo Da Vinci all being part of a single entity.

No fucking thanks.

You will be persuaded otherwise and you will not mind. Because the AI is that good.

No ... you.

"What you quoted is the copyright office's policy from 1966, not something new ruled for this case."

So what? Nowhere in the article or my post is there a "sudden new policy has appeared" statement. Why is this comment section filled with weird tangential claims parading as arguments?

A work generated by a human-created algorithm with a human name on the application will still be accepted.

Sure, getting a copyright requires virtually nothing, nothing gets checked, what would they check? It's content you say you created - stamp! Yes, the only way that policy appears is if you say loudly "hey, give me a copyright for my computer generated thing."

However, the main way it would appear in practice would be if someone attempted to assert copyright on a vast store of computer-generated content. And this is the situation where it would matter.

Sure, getting a copyright requires virtually nothing, nothing gets checked, what would they check? It's content you say you created - stamp!

I have registered hundreds of copyrights. That is not at all how it works.

This is exactly how it works:

https://www.copyright.gov/comp3/chap600/ch600-examination-pr...

I don't think the above comment was implying that copyright can't be registered for works where a computer was used, either in part or fully, to create it. What is being said, I believe, is that you can't register works where the creative process to create that particular work itself (and not the algorithms/training data) was done entirely by a computer.

If you as one person have personally registered hundreds of copyrights that says something about the level of investigation they are capable of doing for each registration. Only a court has the resources to determine whether a copyright exists or not, the copyright office is limited to detecting obvious fraud or ineligible registrations.

>If you as one person have personally registered hundreds of copyrights that says something about the level of investigation they are capable of doing for each registration.

either that or it says something about what that person's job is, and perhaps something about how many people/computing resources copyright office has at disposal to analyze registrations.

Bingo. It should also be noted that for certain works, it is possible to register them as a group.

"The crucial question is “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument," A company logo takes a lot of human creativity. The computer is merely being an assisting instrument.

If I don’t have any pragmas or other statements to instruct a compiler, does that mean my binary wouldn’t be subject to copyright protection?

Your binary is generated from your source code. That is the creative input... Or are you saying the source code was written by a machine?

This very issue came up in 2014 in relation to a selfie taken by a monkey that became popular on the internet: https://en.m.wikipedia.org/wiki/Monkey_selfie_copyright_disp...

The US copyright office wrote:

"Only works created by a human can be copyrighted under United States law, which excludes photographs and artwork created by animals or by machines without human intervention" and that "Because copyright law is limited to 'original intellectual conceptions of the author,' the [copyright] office will refuse to register a claim if it determines that a human being did not create the work. The Office will not register works produced by nature, animals, or plants."

More accurately, machine generated works with no human involvement.

This seems like a good thing to me, because with "AI-generated works", there's the potential to generate a massive amount of images and essentially try to pre-copyright a lot of content, so you can then sue people who produce legitimate works that somewhat resemble what you've already copyrighted. Something similar could be done with programming code and transformers.

We already have a problem with copyright lasting long after the original author(s) are dead...

Can you not programmatically claim "Human Ownership", like as long as you don't have to be physically present I think you can still achieve what you are saying.

I guess what is really "human ownership", maybe i will have to read the article :)

This seems like a bad thing to me, because with "AI-generated works", there's the potential to generate a massive amount of images and essentially try to pre-copyleft a lot of content, so you could sleep easy knowing that it was no longer possible to sue people who produce legitimate works that somewhat resemble what you've already copyrighted. Something similar could be done with programming code and transformers.

We already have a problem with copyright lasting long after the original author(s) are dead...

>then sue people who produce legitimate works that somewhat resemble what you've already copyrighted.

If people didn't copy your work, but instead created it themself you can't (read shouldn't be able to) sue them. Copyright law protects people from making copies of your work. If someone didn't copy from you copyright law doesn't apply.

I just posted the same thing, but good luck proving a negetive. (That you never viewed a certain work publically available on a certain web site.)

That's certainly a difficulty that could cause you to settle with a bad actor.

If someone is doing a meme of copyrighting everything I don't think it would be hard to see through what's going on.

Technically you need to copy to infringe copyright, so if I had never heard the song "Nowhere Man" and recreated Nowhere Man by pure coincidence I would not have infringed the Beatles's copyright.

The issue is one of credibility. Nobody would believe me if I said I had cooincidentally recreated Nowhere Man but with a randomly generated library perhaps you could credibly show it was a coincidence.

That said people could be intimidated by legal fees and the possibility of losing and settle with a bad actor running this sort of scam.

> so if I had never heard the song "Nowhere Man" and recreated Nowhere Man by pure coincidence I would not have infringed the Beatles's copyright.

You would need to prove that you had never heard it. This is non-trivial.

Most recent high-profile music copyright cases do not center on precise reproductions, but on music that "contains elements of" another piece of music.

Yes, but with existing music, current lawsuits use the fact that the music under dispute was playing on radios, nightclubs, tv, etc. So there is the possibility of the other artist having heard of it. If the music is stuck inside of a gigantic database that nobody has accessed, it becomes much easier to prove that you in fact never heard it.

You're mistaken about who has the burden to prove actual copying. The person alleging infringement has to show that the defendant copied the work.

Maybe in the courts, but you can get your youtube channel shutdown or even be permanently kicked off your ISP (and the internet, unless you have more than one option) based on nothing but unsubstantiated accusations that you've violated copyright

True, but then it doesn't matter if you produced any content or not.

It can, depending on the nature of the content and how creators distribute it.

Anyone can find themselves forced to defend their innocence after being hit with DMCA notices, but content creators are especially vulnerable because they depend on their works being publicly available. They have a lot more at stake (including the ability to eat and pay rent) if they are hit with accusations.

They are the ones who'll be hit by poorly implemented algorithms (like Google's notoriously terrible content ID or the many even worse imitations created by companies without the talent/money/data google has at their disposal) or can be targeted specifically because their creative works are perceived as a threat to some company.

The DMCA is often abused to silence criticism, to hide unpleasant information about a company or individual from the public, or to attack competitors. It costs very little to fire off accusations that have real-world consequences and as far as I know, no major company has ever been held accountable for doing so inappropriately, even when it's been brought to the attention of courts.

The standard in a US civil trial is preponderance of evidence, though, so all you have to do is convince the judge that it is more likely than not that the defendant copied the work.

Sure, but where the burden lies still matters. The plaintiff needing to prove copying is a lot more defendant-friendly than the defendant needing to disprove copying.

Actually this is logically impossible, since you can't prove a negative.

It's difficult, but not impossible.

I could reasonably prove to a court I've never been to Jupiter, or that I've never been President of the USA, etc. I can definitely think of situations where someone could likewise reasonably prove to a court they've never heard a song (e.g. they're deaf) or their "AI" had never heard a song (e.g. showing the training data).

I agree that for now this is a good idea, but once the robots can ask for it, this is going to have to change. Hopefully by the time we're working on securing rights for bots we'll have long fixed our broken copyright system so that it once again provides limited protections for creators and isn't just a perpetual revenue stream generator for corporations.

I'd copyright all natural numbers if they'd let me.

The system is clearly rigged!

From Russell and Norvig, a famous story about difficulty in publishing with probably the first such AI-generated work:

"Two researchers from Carnegie Tech, Allen Newell and Herbert Simon, rather stole the show [at the original Dartmouth AI Conference]. Although the others had ideas and in some cases programs for particular applications such as checkers, Newell and Simon already had a reasoning program, the Logic Theorist (LT).... Soon after the workshop, the program was able to prove most of the theorems in Chapter 2 of Russell and Whitehead's Principia Mathematica. Russell was reportedly delighted when Simon showed him that the program had come up with a proof for one theorem that was shorter than the one in Principia. The editors for the Journal of Symbolic Logic were less impressed: they rejected a paper coauthored by Newell, Simon, and Logic Theorist."

This is excellent, a future where you could use algorithms to generate works of art and profit massively by selling them would be even worse for artists than today's art/entertainment world.

Unfortunately it's not clear to me whether this is a ruling that a work created using an algorithm can't be copyrighted (which would be ideal) or if it's just about not listing the algorithm as the "author", which is so obvious it's incredible someone actually spent money trying to litigate it.

> profit massively by selling them

You don't need copyright to sell them.

In the future, media will move faster than copyright. I imagine telling a machine, "make me a new Star Wars film, but with dragons", and out streams the result.

The future isn't going to fit our existing laws or paradigms very well. The laws were written for a different time with totally different expectations.

Yeah. With the pie in the sky being, "make me star wars but with dragons and double check all copyright infringement cases involving star wars to make sure the end result is unlikely to infringe."

Edit. Or heck if the input parameters are really as easy as "star wars and dragons" then you just distribute the generator and the input params.

Except the very idea of a "franchise" will also come under attack when computers can develop narrative, character development, plot devices, world build, etc.

Nobody will want Star Wars when a computer can make Galaxy Dragons and turn it into an epic 100 episode saga. Like your Count Dooku memes? Wait until you hear about Lord Drago Starstream.

And if the concept of a computer developing all of this still seems as far off as "Level 5 autonomous driving", note that with media there will be lots of intermediate steps with humans still in the loop. Just give humans knobs and dials and a "reset" button. Plot the story on a graph, drag the nodes. Nobody dies with this. Human editors can control for quality. There's plenty of automation opportunity until computers become outright storytellers.

Disney and their IP war chest are gonna be toast.

I've been hoping for exactly what you're describing for several years now.

Personally I want Xenomorph protagonist remake genre. Basically a genre where you take a movie and remake it with the protagonist being a xenomorph that the rest of the cast is only vaguely aware of as being a xenomorph.

So xenomorph renegade police detective movie where the detective is a xenomorph who gets kicked off the case, slams his badge onto the cheifs desk, skitters into a vent, and then proceeds to eat the bad guys. All while the other policemen comment about how xenomorph cop is reckless and plays by his own rules.

I love it! I'd totally watch that. (Kindergarten Cop with Xenomorphs? Heck yeah!)

This is my dream too. I'm working on it right now!

It's not perfect at the outset, but I've got a 10 year plan to make every step functional, increasingly monetizable, increasingly professional, and will work my way up the capability ladder one rung at a time.

I'd love to hear from other engineers interested in media, storytelling, graphics, etc. (My email is in my profile.)

> In the future, media will move faster than copyright. I imagine telling a machine, "make me a new Star Wars film, but with dragons", and out streams the result.

That sounds like a dystopia of stagnation and solipsism. Imagine each person becoming a super-fan their own Marvel Cinematic Universe™ that's both utterly derivative and incapable of being a point of connection with others.

"Siri, make me a new self insert! But base this one off Indiana Jones!"

Take a look at TikTok and YouTube. There are so many people creating content for others.

I've no doubt that people will do exactly what you claim and become more absorbed in themselves, but I think the positive use and upside are vastly going to outweigh the gloomy narrative you've laid out.

This technology is going to bring about an artistic Renaissance and people will have more ways to express themselves and employ themselves and do the things that they want with their lives.

>>> In the future, media will move faster than copyright. I imagine telling a machine, "make me a new Star Wars film, but with dragons", and out streams the result.

>> That sounds like a dystopia of stagnation and solipsism. Imagine each person becoming a super-fan their own Marvel Cinematic Universe™ that's both utterly derivative and incapable of being a point of connection with others.

> Take a look at TikTok and YouTube. There are so many people creating content for others.

But those people are creating "content" themselves, they're not asking a competent machine to do it entirely for them. That's such a big difference that they're not even remotely the same kind of activity.

> This technology is going to bring about an artistic Renaissance and people will have more ways to express themselves and employ themselves and do the things that they want with their lives.

You seem to be talking about something entirely different than what you wrote about in your original comment. That comment described the machine doing the expressing, on demand, with minimal user input.

> That comment described the machine doing the expressing, on demand, with minimal user input.

I only gave the machine two inputs. I don't think I'd stop there. I might in fact spend a lot of time with such a technology exploring all of the nooks and crannies of creation space.

We're not going to arrive at the "tell the computer what" fully declarative tech immediately. There's a path from here to there that still requires human involvement, just less of it spent on the mundane minutiae. Once we reach that point, we won't rest on our laurels.

Just as someone might tweak the directed graph of a plot and serialize it to text today, I'd expect them to have their hands in the "less hands on" systems of the future too. They just won't be drafting and editing prose. That'll be old school.

It's neither of those. The copyright office is simply refusing to make a judgement about whether anyone actually believes their AI is a sentient being. Not only is that not their expertise, it is not something they have to decide under copyright law.

The law is simple, if a human wasn't involved in creating the work, then it's ineligible. Bear in mind, a human pressing a single button on a camera does count as human generated work, because it plainly is.

The TL;DR from the copyright office is basically:

"Did that fancy tool really make that work all by itself?... or did you use a really fancy tool?"

I'm sure you're aware of the DABUS case which is the equivalent of your second paragraph wrt a patent application.

There is also the story of "Shalosh B. Ekhad", the computer who co-authored many papers in combinatorics.

[0] https://scholar.google.com/scholar?q=%22Shalosh+B.+Ekhad%22 [1] https://en.wikipedia.org/wiki/Doron_Zeilberger [2] https://www.wired.com/2013/03/computers-and-math/

"Shalosh B Ekhad" is Hebrew for 3.1 (as in π).

> [T]he 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.” U.S. COPYRIGHT OFFICE, REPORT TO THE LIBRARIAN OF CONGRESS BY THE REGISTER OF COPYRIGHTS 5 (1966).

This is probably the right call, though it will become increasingly urgent in the future to evolve the rules to consider the nuances of new cases that will certainly arise.

So... if a machine authored work can't have copyright .. can it breach a copyright?

A machine trained on disney movies, creates a new Disney movie. Can Disney do anything about it?

I am willing to bet Disney would trace it back to the people running the machine quick smart! In which case - like a typewriter - the people running the machine get the copyright.

As others have mentioned though, with machine assistance one person can now spam every text(huge space but rules for making it readable would narrow it)/image(larger space)/video(even larger) that could exist.

So where is the line drawn?

Computers don't have "rights". Period. The end.

There is at least one potential future timeline where your comment is used by our robot overlords as an example of anti-synthetic sentiments that led to the droid revolution of 2121.

Yet they can be accused of crimes and seized!

Same as the pry-bar found under the back seat of the burglary suspect's car, eh?

This doesn't make sense. You can make a AI generated work and claim it to be Human.

It makes perfect sense.

>You can make a AI generated work and claim it to be Human.

That's the point!

That sounds like perjury.

No more than a photographer claiming authorship for simply pushing a button on his art-generating machine.

Photographers don't simply push buttons.

Exactly, and neither do AI-using artists. In both cases, the machine generates the image while the human decides what it should look like.

It seems clear to me that modern "AI" Devices and algorithms are not "Persons". The copyright office would surely have accepted it if he had put down "Creativity Machine" as the /medium/, but the author is a legal person or persons (incl. pseudononymous persons, such as corporations [1]) . What would it mean to assign the copyright to "Creativity Machine"? It can't make any decisions or enter into binding legal contracts. Law is by and for persons, and there's been no movement to grant Personhood to the very limited constructs we're calling "AI" right now. I will be first in line to argue for Personhood for actual artificial intelligence, but this ain't it.

[1] This is my personal pet peeve - "Corporate Personhood", like explored in "Citizens United", means that corporations are pseudononyms for a group of people, and if it would be legal an individual to do, it's legal for a corporation unless the law specifically prohibits it.

> Thaler then requested a reconsideration of the decision, arguing that the human authorship requirement would be contrary to the US Constitution and be unsupported by either statute or case law.

... Feist v Rural? I know it's usually summarized as "sweat of the brow is not valid basis for copyright in the US", but it does this by arguing, as the head notes state:

> Held: Article I, Section 8, clause 8 of the Constitution mandates originality as a prerequisite for copyright protection. The constitutional requirement necessitates independent creation plus a modicum of creativity.

It seems like a lot of the complaints here are against the reliance on 1800s-era cases, but it's the reality that even when modern copyright cases end up before SCOTUS, it all comes back to reaffirming those 1800s cases. (Baker v Seldon, 1880, featured heavily in Google v Oracle, decided 2021; similarly, Feist v Rural [1990] is basically going back and saying that the 1910s-era stuff makes the position clear).

It should be noted that, unlike with patents, copyright registrations are not that significant. It's just a way to establish when you wrote a work, like getting it notarized. You have a de facto copyright on all works you create. In the end, it will be up to a court to decide if the copyright is valid.

I'm glad they're at least being consistent with prior decisions.

https://en.wikipedia.org/wiki/Monkey_selfie_copyright_disput...

Hopefully nobody gets the idea to use AI-generated patent applications just to get your name on a patent, and sell that as a service. That would be very unfair to our patent clerks and a serious drain on our current patent system.

If a machine can combine existing information to produce a new idea, then it fails on obviousness. Also, patents have to get through prosecution, which is still a fraught process.

What would the point of that (other than scoring a fun legal victory) be? Copyright is meant to protect the rights of creators, to distribute their works and benefit from them as they see fit. The machine wouldn't do any of that, or if it did, there would be a human somewhere in the back actually making the decisions (we're still very far from AGI that could do that could decide to act like this truly on its own).

What's to prevent someone from claiming that it is "human authored" when in reality it was AI generated?

As a test, I was able to generate 20 unique music tracks with AI, and submit them to Spotify, Apple Music, Shazam, Tidal + others and they are now "owned" by myself.

This perhaps, will be the bigger problem going forward. Distinguishing between genuine authored content to that of AI generated.

you're still the author

There is nothing wrong with that because you claimed authorship of it and likely put your human skills to work checking which outputs were good.

It’s essentially the next step from tools like photoshop using ML to power edits.

I’ve been using GPT-3 to generate pornography scripts involving unpopular politicians and it has to write some of the funniest stuff I’ve ever read. Nigel Farage was randomly a goat in one of them, but still leader of UKIP. Osama Bin Laden is almost always hiding in a cave and, in a multiple person love story with European leaders, will almost always fall in love with Putin over the others.

In fact one thing I noticed is Putin’s always the dominant partner and will travel to the country of the other, such as Kim Jong-Un, while others are much more passive.

Boris Johnson always cuddles afterwards too - in every generation his always ends with a happy romance.

In Accelerando by Charles Stross, the fictional venture altruist Manfred Macx "...patented using genetic algorithms to patent everything they can permutate from an initial description of a problem domain - not just a better mousetrap, but the set of all possible mousetraps."

Looks like we may not get the set of all possible mousetraps.

Mousetraps do not depend on patents to exist. You can still have your mousetrap set, you just can't prevent others from re-deriving the same set.

The comment you are replying to was using mousetraps as a metaphor. ...at the end of the actual description of what it was talking about. It's an executive summary, not literally talking about mouse traps.

Specifically it was talking about arbitrary problem domains. So ... much more than mousetraps.

What makes you think this was unclear?

Year 2050: AI Copyright office system won't accept Human-generated art, requires "AI authorship"

There's a really important distinction here.

(a) An algorithm cannot be named as an author, name the human who ran it instead.

(b) A work produced by an algorithm is *inherently not copyrightable*. You cannot name a human as author instead, that would be fraud if the work was produced by an algorithm.

I am reading this as (b).

The creator of the algorithm is the inventor of the algorithms results.

I’m not taking a stand, I’m just saying it sounds like the copyright office is saying (b) which is counter to what you wrote.

This is clearly the correct decision.

If the AI is capable of creative intellectual thought, then his claim to own it is clearly false, because that's slavery, which is illegal.

If it's not, then it's not really the author. The person who applied the AI is the author.

Aaaa! We got 'em! If AI can't copyright itself, then it won't be allowed to take over or destroy the world (as long as a human and/or patent troll copyrights world domination first). One small victory for mankind!

What if we could prove certain software patents were actually the work of AI

That's a small step in the right direction. Reducing the protection period to something sane would be nice too, but that's probably hoping for too much.

AI is a tool like a camera, point it a thing and push go.

Each has billions of dollars of global supply chain supporting their manifestation.

How does this affect GitHub Copilot? They used to claim that their AI can create new IP and transfer it to the users.

Iranian patent trolls are gonna be so disappointed that they can't troll patents so easily anymore.

Imagine AI being able to gain copyright for their productions.

We'd run out of things to copyright real fast.

If the authors have to be human then so do the owners. Be consistent dammit.

Good. I imagine if it was any other way we’d just have a trillion machine learning models making stupid things for the IP ownership and a round of copyright trolling by their operators.

Given that the fee appears to be $45 per work to register the copyright, I'm having a hard time seeing a good business model here.

How long is a book? How much text can I copyright for $45 and then go after people over "excerpts"?

I'm not sure but I can see how such a system can be gamed

Registration is not required to sue, but definitely makes it easier.

You can file multiple works as one collective work - https://www.copyright.gov/circs/circ34.pdf

So the AI has some side jobs on Fiverr

Is this how the paperclip maximizer scenario gets started? The AI decides to take over the government in order to reduce filing fees.

"You optimized for WHAT?"

Corporations are people though. Can you go through the same process without mentioning people's names and only mention a company name?

good

Reply on news.ycombinator.com