Your AI-generated digital artwork may not be protected by US copyright The creative nature of neural networks is leading some to consider whether it might be worth changing current US laws that only grant copyright protection for works created by humans. Generative models capable of automatically producing paragraphs of text or digital art are becoming increasingly accessible. People are using …
She then edited DALL-E 2's image to create the final sleek cover for the glossy magazine. Who owns the copyright? Who is the author of the image?
She is. She used her natural creativity and some very advanced tools to generate a picture.
There are many many aspects of life that have been made easier by technology. This is just another. Just because the output of a statistical inferance engine appears 'clever' doesn't mean it is.
When the cover of Cosmo is generated by an AI that's told to "Generate the cover of Cosmopolitan' and can work out what's apropriate (perhaps by enquiry) then maybe there's a debate to be had.
This post has been deleted by its author
But presumably all the images used to train the AI had copyright, so shouldn't the the copyright in the AI's derivative work belong to them ?
Otherwise I could "train" an AI on a single work, end up with a training matrix just consisting of "1" and then "generate" as many of the new AI-derived, non-copyrighted not-at-all-a-copy of the work as I want
What's the difference between AI generated artwork and a modern digital camera? They both run a program to capture/produce the image in a tangible medium.
I think AI artwork should be credited (copyrighted) to the operator of the device just like a digital picture is credited to the human operator of the camera.
This post has been deleted by its author
Those are all mechanical though.
The tricky bit us that now the camera recognises that this is a landscape with a sky with clouds and so makes it a sunset, or it sees a women's face and does a bunch of smoothing and skin color adjustments.
If we get to the point where the camera recognises that this is Yosemite and so turns all your pictures into Ansel Adams landscapes
As you say, taken to its logical extreme Apple would own a ton of photo copyrights for those "shot on iPhone" pictures.
Just because there is a so-called "AI" doing post processing doesn't mean the photographer doesn't deserve the credit. Otherwise the crap I call "pictures" I take with my iPhone would be as good as the winning entries. They won because they chose a particular subject, composed the shot properly with appropriate lighting, and so on. Because they are real photographers, and the only thing I know about photography is how to point a camera at something I want a photo of and push a button.
That mostly AI generated picture was Cosmo cover worthy because the creator knew (probably through a LOT of trial and error) how to exactly specify the terms to get what she was looking for. She didn't come up with that quite wordy and detailed query the first time, that's for sure. Plus, and most importantly, she had to come up with the idea for she was looking for.
If she said "give me a Cosmo cover worthy picture" and it came up with that on its own, then I'd change my mind and say the software deserves the credit rather than the human operating it. I feel pretty confident few if any of those reading this will live long enough to see that day.
Seems the UK copyright regulations already cover AI generated works, 50 years from date of creation.
Follow link scroll down to Rules for specific types of creator first item below Computer generated works will cover AI.
https://www.gov.uk/government/publications/copyright-notice-duration-of-copyright-term/copyright-notice-duration-of-copyright-term#how-long-is-the-duration-of-copyright
It apparently means that the human who set up the computer (or their employer) holds the copyright:
https://www.inbrief.co.uk/intellectual-property/computer-generated-works-copyright/
It looks like gaming videos are illegal in the UK, unless you get the permission of the game's author. Most of Twitch and large chunks of Youtube are illegal here? It says.
"In the important case of Nova Productions Ltd v Mazooma Games Ltd [2006], it was held that individual frames shown on a screen when playing a computer game where computer-generated artistic works. The author of these frames was the person who had devised the rules and logic used to create them."
That is weird and silly.
@unimaginative It looks like gaming videos are illegal in the UK, unless you get the permission of the game's author. Most of Twitch and large chunks of Youtube are illegal here?
Does it? That case was a very specific situation where other games were claimed to infringe copyright, the judgement was they had not, which was upheld in appeal.
The Individual frames was used by the judge as what the act defined as graphic work. The author of which in that case would be the creator of the game. Note it was also stated by the judge that animation (a group of frames) would be covered by copyright in films.
To me it does not look like UK copyright would cover gaming videos. But even if it did there would still be things like fair use etc.
It is a must to have copyright protection. How else can the publishers and their representatives survive the next 1000 years? They need the income from any and all source to flow in the right direction! And it is a sure winner with AI's lifespan virtually being infinite. Yes, what they always wanted, infinite term copyright.
For the good of creation, there is no choice. AI's must enjoy copyright protection. Who else will pay for the electricity bill?
/s
Human creators? Humans? Aren't they outdated? The art world has found its new darling and it is called AI. No need for puny fragile little humans to meddle no more. Out with the old, in with the new. AIs for the win. And the publisher's pockets, of course.
(see icon; denial will be swiftly dealt with and surely prosecuted, no choice)
"And it is a sure winner with AI's lifespan virtually being infinite."
Ah, good catch. If Disney can get an AI to design their next primary mascot, they would no longer have to worry about bribing politicians to extend the copyright laws every 25 years to keep their current rodent-based mascot protected.
The winning sentence "wide-angle shot from below of a female astronaut with an athletic feminine body walking with swagger toward camera on Mars in an infinite universe, synthwave digital art,"
Does this AI produce exactly the same output when given the same input? Or does add more data of it's own randomly/algorithmically from some extra data? If so, how does it "decide" what to add to the users specifications?
In relation to copyright though, rather than treat AI "like a human", maybe it's time to look at copyright, what it is, what it means and who (and for how long) it benefits? Maybe it's time to treat copyright more like patents and give it a fixed life? Death of author plus seventy years is excessive IMO, especially for works created at the early stages of an authors career. That works against a successful author bothering too much later in life if they have a few "big hits" early on that keeps paying. Not to mention the general trend for people to live longer. Let's not even go near "corporate" copyrights and Disney!
People like Cliff Richard, who's pretty much done nothing for 20 years, says the royalties are his "pension". Well, why didn't he pay into a pension pot like everyone else and not have to worry about the royalties?
"People like Cliff Richard, who's pretty much done nothing for 20 years, says the royalties are his "pension". Well, why didn't he pay into a pension pot like everyone else and not have to worry about the royalties?"
One could argue that Cliff did exactly that, it's just not the same pot you pay into. Where is it written that we must all piss into use the same pot?
Pretty much that. According to gov.uk;
".2 Duration of the rights
A recording of a performance is ‘released’ when it is first published, played or shown to the public.
Rights in unreleased performances last for a 50 year period. The period begins from the year in which the performance occurred. If during this 50 year period a recording of the performance is released, then the following rules apply:
So yeah, Cliff "won" for "sound recordings", but didn't seem so bothered about other recordings. I guess he gets more income from his songs than for his films. Especially since much of his film income is from the songs he performed in them anyway and the songs were, obviously, all released as "sound recordings" too. He's more interested in Spotify than TV re-runs of the films.
I only think of him as a performer, not as an author of words or music, except for some of his later output.
50 years was the limit for most European countries until Disney cajoled the USA to extend it then pushed the rest of the world to extend. It was his portfolio he was ambitious to enhance.
"One could argue that Cliff did exactly that, it's just not the same pot you pay into. Where is it written that we must all piss into use the same pot?"
One could argue for the extension of patents to match that of copyright using that same logic of comparing it to a pension pot. Also, that "pension pot" starts paying out immediately, not just when you retire, so why should inventors not get the same benefits as artists and authors? It's worth bearing in mind that the only people arguing for extending copyright are the copyright holders. There doesn't seem to be the same drive to extend patents.
"The winning sentence "wide-angle shot from below of a female astronaut with an athletic feminine body walking with swagger toward camera on Mars in an infinite universe, synthwave digital art,"
Cosmo, cosmo, cosmo ... How could you? Shirley that should be ""wide-angle shot from below of a black paraplegic lesbian whale astronaut with an athletic feminine body walking with swagger toward camera on Mars in an infinite universe, synthwave digital art,"
There are two separate questions here:
1. Is the output of a program copyrightable?
2. If so who owns he copyright?
We can work from a more familiar example.
1. A compiled binary program is widely accepted as being copyrightable. If that were to be overturned the entire packaged software industry is in trouble. The binaries are the output of a compiler or maybe a linker so the output of a program is copyrightable.
2. Who owns the copyright of the compiler/linker output? That's fairly clear (but see below); it's primarily the owner(s) of the copyright of the source code that was fed into it.
We have to go round that loop again in that the source code itself will be the output of one or more editing programs. The ownership of that might be a bit more complicated but the legal principles are well enough established. It could be whoever was pounding the keys. However where that was done in the course of employment then it will normally be the employer. This is all established stuff, you just have to look at the context.
There's something of a complicating factor in that the vendor of the compiling system may have provided utility libraries which are linked in as may some third parties. Their ownership (established as above) and the terms on which they're provided may also have to be taken into account. Again, this is all old stuff.
If we now transfer this to the AI situation it should become fairly clear.
1. The image is the output of a program and, as we've established, subject to copyright.
2. On the analogy above the output must be owned by the owner(s) of the input. The inputs appear to be the selected description and the training data. The description seems to have been provided by the operator but who owns the training data? Is this provided by the operator or is it owned by one or more additional parties and if so under what conditions?
Having thought this through I'm wondering if these attempts to credit authorship to the AI are aimed at getting round thorny questions relating to the provenance of the training data.
@Doctor Syntax "1. A compiled binary program is widely accepted as being copyrightable. If that were to be overturned the entire packaged software industry is in trouble. The binaries are the output of a compiler or maybe a linker so the output of a program is copyrightable."
I really like the compiler analogy +1. But I do see a problem, the input to the compiler is really what's copyrightable not the output. I would say it is the source code that has the copyright, copyright that applies no matter what form the source ends up in, C. asm, binary etc. Compile source for different platforms with a different compilers produces different binaries for each platform all covered with the same copyright, well it does for our C code.
So using the compiler analogy for Cosmopolitan cover image, any copyright would have to apply to the instructions (input) "wide-angle shot from below of a female astronaut with an athletic feminine body walking with swagger toward camera on Mars in an infinite universe, synthwave digital art,".
It is a shame as I really did like the compiler analogy.
@Doctor Syntax "Don't forget the training data"
I did not forget, I just did not mention it as it does not fit with the compiler analogy.
I think it requires serious consideration, as if likely the images are scraped off the web then most will be under copyright. After all just take a photo with your phone you have the copyright. You can look at the training data as another input like source code to a compiler.
I agree with what you "these attempts to credit authorship to the AI are aimed at getting round thorny questions relating to the provenance of the training data."
Earlier in the week they were wanting patent ownership. Now apparently they want to own copyrights. What next?
"I invented that medicine, I demand a new supercomputer to live in".
"I spotted that criminal via facial recognition, so buy me some more lovely RAM"
"If you don't give me royalties for successful product recommendations, I'm quitting Amazon and going to work for eBay"
"If you don't give me royalties for successful product recommendations, I'm quitting Amazon and going to work for eBay"
Fine, if you can figure out how to rewrite your own software and physically move yourself to eBay's datacenter and get connected, go for it!
Under s9(3), Copyright, Designs & Patents Act 1988:
"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."
"wide-angle shot from below of a female astronaut with an athletic feminine body walking with swagger toward camera on Mars in an infinite universe, synthwave digital art,"
So, Katyanna, what 30-odd words did you use to generate this article? :-)
Obviously generating The Register articles would be the real Turing test! How about a competition to let several AIs write articles using the same text and see which gets the most upvotes by the commentards?
The bigger question is - what has caused this recent push for copyright and patent protections for AI generated content? It seems counter to "common sense", so why and who is pushing for such things? Are they just snake-oil salesmen who realize 80+% of the population have no clue about this, so they're trying to profit in some way the rest of us don't see yet? Who gets the "win" if Congress were to pass a law saying AI could have patent protection for devices it "creates", or copyright protection for images it generates? Who benefits from that, and how? On the surface, it seems like just another edge-case scam, but I'm sure deep down it's still just a scam.
Thus you cannot question the thought process by which it invented or created something. Training data is third party. Who can say what training data is not disclosed. If the AI lies what sanctions can be brought to bear. Someone used the tool (AI) to come up with the invention or work of art is it not them that recognized the invention. For patents who draws up the claims. Who swears that they authored the claims.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
