[Introduction: Francis Hamit, currently running a Kickstarter appeal for Starmen: A Novel, wrote this article as one of the Updates. He’s given File 770 permission to reprint it.]
By Francis Hamit: Not even God can get a Copyright. That’s one of the takeaways from Thaler vs. Perlmutter. Thaler wanted to register an image generated by a computer program he devised. Perlmutter, who runs the Copyright Office at the Library of Congress, said no. Only works made by humans can get copyright protection. Your pet monkey pounding away on a computer keyboard might produce something brilliant that everyone wants to buy but you won’t be able to protect it from infringement. Likewise anything from a celestial being also falls into the Public Domain. Only work that is the result of human creativity can be protected.
Copyright is a global law through various treaties so registration here protects your work in most but not all markets. It lasts a very long time and is about the money and controlling who gets to publish, display, distribute and adapt original work.
ChatGPT may seem like a brilliant innovation to some but all it really is is a very sophisticated computer program with a huge database. Even then it has to be trained by humans before it gets those amazing results. The “ghost in the machine” is us. (Pay no attention to the man behind the curtain.)
To be a real writer you have to be real. You have to do your own work. Artificial Intelligence is just that; artificial. And it shows in recent AI submissions submitted by quick buck artists to magazines in the science fiction field. There were so many that they closed down the submissions window. Easily detected and easily dismissed but still a drag on the process because they had to be dealt with.
I recommend that everyone register their copyrights. Forget about that mail it to yourself nonsense. Go online and pay the fees but understand that all you are protecting is the expression, the arrangement of the words, not the idea behnd the words. And all you have is the right to sue some person you think has infringed your work in a Federal District Court.
I’ve done this and don’t recommend it unless there is life-changing money at stake. You have to prove the case and the Federal Courts hate copyright cases with a passion. Why? There is the Law, which is simple. Black Letter as they say. And then there are the Facts and those are incredibly complicated.
Most cases settle before they get to trial. Mine did but it took four years and thousands of dollars and thousands of hours before I got paid. How much I’m not allowed to say. The Judgment is Sealed.
The Thaler vs. Perlmutter decision sets a precedent. Any taint of A.I. in your text can void your copyright registration. But the judge left open the use of A.I. as a tool similar to a spellchecker as long as that original spark of creativity is preserved. If you use A.I. to transform your short story into a screenplay is that simply a derivative work? Will the original copyright registration stretch to cover it? Ask yourself this: Which is easier? Learning how to write that screenplay yourself or defending your A.I. generated version in Federal Court?
Discover more from File 770
Subscribe to get the latest posts to your email.
(An AI-aided) Click
I’m not a lawyer but the decision implies the opposite. Human intent and creative expression makes the work copyrightable even if a mechanical process was a major part of the work.
eg in discussing photography:
“A camera may generate only a “mechanical reproduction” of a scene, but does so only after the photographer develops a “mental conception” of the photograph, which is given its final form by that photographer’s decisions like “posing the [subject] in front of the camera, selecting and arranging the costume, draperies, and other various accessories in said photograph, arranging the subject so as to present graceful outlines, arranging and disposing the light and shade, suggesting and evoking the desired expression, and from such disposition, arrangement, or representation” crafting the overall image. Id. at 59–60. Human involvement in, and ultimate creative control over, the work at issue was key to the conclusion that the new type of work fell within the bounds of copyright”
Life is a Cut-Up
Yeah, to follow on what Camestros said, you own copyright on your creative expression, even if that expression only consists of part of a work! The mistake is thinking that a work has a (singular) copyright. That’s merely the simple case, and counter-examples are probably more common than you think. For example, a musical recording can easily have three copyrights: lyrics, melody, and performance/arrangement. Bob Dylan may have written the words and melody to “All Along the Watchtower”, but Jimi Hendrix’s estate still owns the copyright to the guitar solo in his famous cover.
An extreme case can be found in open-source software, which is usually owned. piece-by-piece, by all the programmers who have contributed. Most commercial software falls under work-for-hire rules, and thus only has one owner, but works like the Linux kernel have more copyrights than anyone wants to count.
Bottom line, if you make a work based on AI output, you should still own the copyrights to any parts you wrote, but any parts written by the AI will effectively be public domain. Your copyright in the work will be extremely limited.
Yes, Burroughs is very relevant. Another interesting quirk of the case lies with what the plaintiff claimed INTIALLY rather than what they claimed in court. When they applied for copyright:
“Here, plaintiff informed the Register that the work was “[c]reated autonomously by machine,” and that his claim to the copyright was only based on the fact of his “[o]wnership of the machine.” Application at 2. The Register therefore made her decision based on the fact the application presented that plaintiff played no role in using the AI to generate the work, which plaintiff never attempted to correct. ”
But when they challenged the denial of copyright in court they advanced other arguments including that there is human involvement (eg prompt writing etc). That didn’t wash because that’s not what they had stated to the official originally:
“Plaintiff’s effort to update and modify the facts for judicial review on an APA claim is too late. On the record designed by plaintiff from the outset of his application for copyright registration, this case presents only the question of whether a work generated autonomously by a computer system is eligible for copyright. In the absence of any human involvement in the creation of the work, the clear and straightforward answer is the one given by the Register: No.”
Again, I’m not a lawyer but I think this narrows what the court was saying. Somebody says to the copyright office: a machine made this without human intervention, can I have the copyright? The office says no. That decision is challenged and the court finds the office was correct to make its decision based on the information given.
So what happens if somebody goes to the copyright office and says: I made this with the help of a machine – even though it is the exact same process? I’m not sure.
I’m reminded of trap streets on maps. Since on a map the raw and true information can’t be copyrighted, it’s very hard to prove someone ripped off your map to publish their own. They could just say we went out and gathered the same true information as you. So what map makers do is put fake place names onto their maps in places that won’t hurt anyone. If someone just copies your map, they also copy your fake places. This proves they copied you and you can sue them.
Maybe a AI generated work could gain similar protection by adding a number of human written (and there for copyrightable) landmines throughout the text? Anyone looking to exploit the non-copyright status of the main work would still trip over copyright when they copied the human elements. There’s be no easy way to tell them apart after all.
If they included your human written elements you could then still sue them.
I’d also warn people that the original post was very US centric, which can be important to non-US authors and if you publish internationally. The UK, for example, provides 50 years of protection to machine created works.
There’s precedent of a sort in the monkey selfie case. If an animal can’t own a copyright, neither can a machine.
@Graeme: Yes, copying the human-written parts would be a copyright violation, I’d expect. But you might have to have some way of proving that it was indeed you and not the AI that wrote the bits in question! If you’re trying to sue someone, the burden of proof is generally on you.
For something like Pride and Prejudice and Zombies, where part of the work is public domain, the proof would be easy, because the public domain parts are well-known and well documented. For something where the public domain parts are AI generated, though, it might be a lot more difficult to prove who wrote which bits.
The “trap streets” trick wouldn’t really work because there’s no real streets to compare to. You’d need something like that, though, but better. I honestly don’t know what. Nor am I going to lose sleep worrying about it, since I have no intention of collaborating with an AI. 🙂
Burroughs threw out far more than 99 percent of his cut-ups, keeping the lines that he found creatively useful. (I guess he also did prompting by choosing which texts to cut up and how long the bits were.)
He saw the cut up technique as a lens through which to view reality in new ways.
A mechanical process, with a ghost in the machine.
Something similar is happening now. I experimented with AI-assisted writing, as an experiment, not for publication. It was great. It thought up all kinds of angles that never occured to me. I couldn’t use very much of what it produced verbatim, of course, and even when I did I had to give it prompt after prompt after prompt to massage that material in the direction I wanted. The end result was something I would have had a hard time producing alone, and certainly not so efficiently. It was writing with a (very weird, nonhuman) collaborator I could boss around.
Even that guy who did the Bob the Wizard cover used the tool in an innovative way. As in, the AI gave him the pulpy noir man in the lame trademark “made of smoke” style, and he went “huh, OK… now do Tinkerbell” and placed the two in an extremely eye-catching juxtaposition. It was a good cover made by a good artist.
ChatGPT sucked at writing pixel scroll titles though.
The only one I really liked was “Hello, Dalek!”
Brian Z: But ChatGPT should be really good at writing Scrolls — the Washington Post looked into a dataset used to train programs of that kind. Their stats showed File 770 ranked 3,445th with 2.5 million “tokens”. Even a blind pig, never mind 10,000 monkeys, ought to be able to pick a fine Scroll from all that compost.
Well, it can’t pun to save its life. I had it give me a hundred titles and even after providing it with very clear advice on how it’s done 99 of them sucked.
How intelligent does the AI have to be in order to create an issue? What if my brilliant writing is inspected by spiel Chequers (Chequers is the English Prime Minister’s country cottage), and the have the unmitigated gall to make corrections?