Judge snuffs man's quest to have AI-created art protected by copyright Copyright issues have dogged AI since chatbot tech gained mass appeal, whether it's accusations of entire novels being scraped to train ChatGPT or allegations that Microsoft and GitHub's Copilot is pilfering code. But one thing is for sure after a ruling [PDF] by the United States District Court for the District of Columbia – …
With regard to the "must be human created", I was curious and looked up the situation of animal-created art and, of course, the (in)famous Macaque selfie.
Turns out that yes, this holds absolutely true: only a legal person can hold copyright, which an animal is not, and the copyright may also not be claimed by the human owner/trainer/wild-life photographer by proxy.
https://en.wikipedia.org/wiki/Animal-made_art
There's precedence, especially as I would argue your average elephant, cat, monkey/ape has more sentience/sapience than over-hyped "rapid look-up engines cum neural network" called "AI". Naturally, therefore this hold true for a computer.
At least right now. Our future machine overlords may revisit this and add an inflammatory reply some time long after I'm gone and give a flying f*ck.
The current labour strife between Hollywood producers and writers and actors is in-part due to the threat of AI technology reducing or eliminating the use of humans in productions.
While movie production is a very complicated process, a significant part is already run by computers (special effects, audio processing, audience opinion evaluation, even projection. And dozens more topics.)
Add a little bit of "AI" to the mix and you have the risk that copyright in a movie or tv production could be challenged on the same basis: that humans didn't do the whole thing, only a part.
I can foresee a whole lotta political posturing and scheming and loitering and lobbying by the likes of Disney (again), Sony (again), and others to "fix" this little issue.
As usual, it will be the lawyers and politicians that gain the most.
"Add a little bit of "AI" to the mix and you have the risk that copyright in a movie or tv production could be challenged on the same basis: that humans didn't do the whole thing, only a part."
No as the studio would be claiming that the operator of the computer was the creator.
This case was about a guy trying to claim that the AI itself should be credited as the creator and should hold the copyright, the court told him that that wasn't how the law worked and that he was the creator and should register the copyright in his name.
"Add a little bit of "AI" to the mix and you have the risk that copyright in a movie or tv production could be challenged on the same basis: that humans didn't do the whole thing, only a part."
That won't work that way for three independent reasons:
1. Generally, if humans did some part, that part will be covered. If the film as a whole is in question, they can point to the human-made parts which will be copyrighted and use that to prevent anyone copying the film legally. The protection they're looking for from copyright is usually either to prevent people from making copies available for free or to prevent someone from using their characters and setup in an unauthorized sequel. The former only requires that enough of the final film is copyrighted, and the latter requires that the starting kernel of the script is, both of which are easy even if part of the dialogue is not.
2. Studios will not be trying to claim that an AI program itself is the copyright holder. For each part that is automatically generated, some studio employee or contractor spent time running different prompts, judging the result, and editing the final thing. Those actions are creative, so even if the output from the computer isn't copyrightable, they'll have a claim for why they did put human creativity back into the process and should get the protection anyway.
3. They are copyrighting a lot of things, and they're not going to tell someone how each one was generated. Copyright already applies to things generated with computers, such as animation models and the video created from them. It will not be easy to identify those parts that were specifically created by a program without sufficient human input. In order to challenge their copyright, you'd have to be able to say that this part doesn't count and provide reasoning, but they won't provide you the information necessary to prove it.
The position in the US is one thing - the position elsewhere is quite another. For example, section 9(3) of the UK's Copyright, Designs and Patents Act 1988 provides:
"In the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken."
Thaler tried to have his Creativity Machine registered as the copyright author, rather than himself, as I undertand it. So the UK would also reject his application under the section you quote. If he tried to register himself as the copyright owner, then he would probably succeed. But that doesn't appear to be his motivation. As noted in the article, he has form for trying to get software registered as the author and owner of intellectual property, rather than people.
I really wonder how big Professor Ryan Abbott's fee was for expressing his very useful opinion (useful to the plaintiff, that is). He's quite a character.
I'm looking forward to the different copyright disputes.
I have played around with an image generating AI
.. I had some free "credits" to play with so no expense incurred from hammering the AI.
I wanted to see what "variety" of results I got, so made repeated calls of "prompt A" and then repeated calls of "prompt B"
Both prompts very similar (just minor variations on a theme to see what limits hit trying to fine tune output)
I noticed that with repeated calls of the same prompt that got fairly similar images back, in some cases strikingly similar - which is to be expected (I'm onviously not sure how the algorithms work, I'm expecting some "randomness" to the generation but it will be limited, especially with a detailed prompt)
What was more interesting was that in some cases, a "Prompt B" image was very similar to a "Prompt A" image (though, TBF, in some cases I was often trying complex prompts to see when additional prompt text made little difference ).
So, based on that, I could easily see 2 people in a copyright spat over AI generated images that are very similar.
If they were both derived from the same prompt then will we be reaching a point where people try and copyright a certain prompt text used against a certain AI (which is a problem as there is not just one result possible for any given prompt so essentially prohibiting other images)
.. and gets even more fun if the 2 people in a copyright spat used different prompts to give effectively the same result.
To complicate things more (and I would guess especially if you guess right on which AI used originally) you can often get reasonably close (with enough attempts, especially with a bit of refining after inspecting first few tens of results) to replicating an AI generated image by essentially passing in a good description of the image you want to (approximately) reproduce .
I did not try to reproduce an AI image in my tests but was able to get some reasonable results in getting AI to approximate an actual image (human generated AFAIK) I was describing (main limitation on this was beyond a certain point adding more detail to your prompt was not necessarily effective )
Latest versions of Photoshop, used in creation of much digital art whose creators claim copyright, have AI tools within them such as Generative Fill.
So at what level does the mechanical process in those tools take precedence over the artistic choice to use the tool?
How about a bucket on a string with a hole punched in it, spilling paint on a canvas, running entirely on the mechanical process of gravity and hydrodynamics?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
