Those claiming AI training on copyrighted works is “theft” misunderstand key aspects of copyright law and AI technology. Copyright protects specific expressions of ideas, not the ideas themselves. When AI systems ingest copyrighted works, they’re extracting general patterns and concepts - the “Bob Dylan-ness” or “Hemingway-ness” - not copying specific text or images.
This process is akin to how humans learn by reading widely and absorbing styles and techniques, rather than memorizing and reproducing exact passages. The AI discards the original text, keeping only abstract representations in “vector space”. When generating new content, the AI isn’t recreating copyrighted works, but producing new expressions inspired by the concepts it’s learned.
This is fundamentally different from copying a book or song. It’s more like the long-standing artistic tradition of being influenced by others’ work. The law has always recognized that ideas themselves can’t be owned - only particular expressions of them.
Moreover, there’s precedent for this kind of use being considered “transformative” and thus fair use. The Google Books project, which scanned millions of books to create a searchable index, was ruled legal despite protests from authors and publishers. AI training is arguably even more transformative.
While it’s understandable that creators feel uneasy about this new technology, labeling it “theft” is both legally and technically inaccurate. We may need new ways to support and compensate creators in the AI age, but that doesn’t make the current use of copyrighted works for AI training illegal or unethical.
For those interested, this argument is nicely laid out by Damien Riehl in FLOSS Weekly episode 744. https://twit.tv/shows/floss-weekly/episodes/744
This process is akin to how humans learn by reading widely and absorbing styles and techniques, rather than memorizing and reproducing exact passages.
Like fuck it is. An LLM “learns” by memorization and by breaking down training data into their component tokens, then calculating the weight between these tokens. This allows it to produce an output that resembles (but may or may not perfectly replicate) its training dataset, but produces no actual understanding or meaning–in other words, there’s no actual intelligence, just really, really fancy fuzzy math.
Meanwhile, a human learns by memorizing training data, but also by parsing the underlying meaning and breaking it down into the underlying concepts, and then by applying and testing those concepts, and mastering them through practice and repetition. Where an LLM would learn “2+2 = 4” by ingesting tens or hundreds of thousands of instances of the string “2+2 = 4” and calculating a strong relationship between the tokens “2+2,” “=,” and “4,” a human child would learn 2+2 = 4 by being given two apple slices, putting them down to another pair of apple slices, and counting the total number of apple slices to see that they now have 4 slices. (And then being given a treat of delicious apple slices.)
Similarly, a human learns to draw by starting with basic shapes, then moving on to anatomy, studying light and shadow, shading, and color theory, all the while applying each new concept to their work, and developing muscle memory to allow them to more easily draw the lines and shapes that they combine to form a whole picture. A human may learn off other peoples’ drawings during the process, but at most they may process a few thousand images. Meanwhile, an LLM learns to “draw” by ingesting millions of images–without obtaining the permission of the person or organization that created those images–and then breaking those images down to their component tokens, and calculating weights between those tokens. There’s about as much similarity between how an LLM “learns” compared to human learning as there is between my cat and my refrigerator.
And YET FUCKING AGAIN, here’s the fucking Google Books argument. To repeat: Google Books used a minimal portion of the copyrighted works, and was not building a service to compete with book publishers. Generative AI is using the ENTIRE COPYRIGHTED WORK for its training set, and is building a service TO DIRECTLY COMPETE WITH THE ORGANIZATIONS WHOSE WORKS THEY ARE USING. They have zero fucking relevance to one another as far as claims of fair use. I am sick and fucking tired of hearing about Google Books.
EDIT: I want to make another point: I’ve commissioned artists for work multiple times, featuring characters that I designed myself. And pretty much every time I have, the art they make for me comes with multiple restrictions: for example, they grant me a license to post it on my own art gallery, and they grant me permission to use portions of the art for non-commercial uses (e.g. cropping a portion out to use as a profile pic or avatar). But they all explicitly forbid me from using the work I commissioned for commercial purposes–in other words, I cannot slap the art I commissioned on a T-shirt and sell it at a convention, or make a mug out of it. If I did so, that artist would be well within their rights to sue the crap out of me, and artists charge several times as much to grant a license for commercial use.
In other words, there is already well-established precedent that even if something is publicly available on the Internet and free to download, there are acceptable and unacceptable use cases, and it’s broadly accepted that using other peoples’ work for commercial use without compensating them is not permitted, even if I directly paid someone to create that work myself.
This process is akin to how humans learn by reading widely and absorbing styles and techniques, rather than memorizing and reproducing exact passages. The AI discards the original text, keeping only abstract representations in “vector space”.
Citation needed. I’m pretty sure LLMs have exactly reproduced copyrighted passages. And considering it can created detailed summaries of copyrighted texts, it obviously has to save more than “abstract representations.”
There is an easy answer to this, but it’s not being pursued by AI companies because it’ll make them less money, albeit totally ethically.
Make all LLM models free to use, regardless of sophistication, and be collaborative with sharing the algorithms. They don’t have to be open to everyone, but they can look at requests and grant them on merit without charging for it.
So how do they make money? How goes Google search make money? Advertisements. If you have a good, free product, advertisement space will follow. If it’s impossible to make an AI product while also properly compensating people for training material, then don’t make it a sold product. Use copyright training material freely to offer a free product with no premiums.
Generative AI does not work like this. They’re not like humans at all, it will regurgitate whatever input it receives, like how Google can’t stop Gemini from telling people to put glue in their pizza. If it really worked like that, there wouldn’t be these broad and extensive policies within tech companies about using it with company sensitive data like protection compliances. The day that a health insurance company manager says, “sure, you can feed Chat-GPT medical data” is the day I trust genAI.
“This process is akin to how humans learn… The AI discards the original text, keeping only abstract representations…”
Now I sail the high seas myself, but I don’t think Paramount Studios would buy anyone’s defence they were only pirating their movies so they can learn the general content so they can produce their own knockoff.
Yes artists learn and inspire each other, but more often than not I’d imagine they consumed that art in an ethical way.
I’ll train my AI on just the bee movie. Then I’m going to ask it “can you make me a movie about bees”? When it spits the whole movie, I can just watch it or sell it or whatever, it was a creation of my AI, which learned just like any human would! Of course I didn’t even pay for the original copy to train my AI, it’s for learning purposes, and learning should be a basic human right!
I am thrilled to see the output you get!
learning should be a basic human right!
Education is a basic human right (except maybe in Usa, then it should be one there)
Yeah. A human right.
In the meantime I’ll introduce myself into the servers of large corporations and read their emails, codebase, teams and strategic analysis, it’s just learning!
That would be like you writing out the bee movie yourself after memorizing the whole movie and claiming it is your own idea or using it as proof that humans memorizing a movie is violating copyright. Just because an AI is violating copyright by outputting the whole bee movie, it doesn’t mean training the AI on copyright stuff is violating copyright.
Let’s just punish the AI companies for outputting copyright stuff instead of for training with them. Maybe that way they would actually go out of their way to make their LLM intelligent enough to not spit out copyrighted content.
Or, we can just make it so that any output made by an AI that is trained on copyrighted stuff cannot be copyrighted.
If the solution is making the output non-copyrighted it fixes nothing. You can sell the pirating machine on a subscription. And it’s not like Netflix where the content ends when the subscription ends, you have already downloaded all the not-copyrighted content you wanted, and the internet would be full of non-copyrighted AI output.
Instead of selling the bee movie, you sell a bee movie maker, and a spiderman maker, and a titanic maker.
Sure, file a copyright infringement each time you manage to make an AI output copyrighted content. Just run it on a loop and it’s a money making machine. That’s fine by me.
Yeah, because running the AI also have some cost, so you are selling the subscription to run the AI on their server, not it’s output.
I’m not sure what is the legality of selling a bee movie maker, so you’d have to research that one yourself.
It’s not really a money making machine if you lose more money running the AI on your server farm, but whatever floats your boat. Also, there are already lawsuits based on outputs created from chatgpt, so it is exactly what is already happening.
Yeah, making sandwiches also costs money! I have to pay my sandwich making employees to keep the business profitable! How do they expect me to pay for the cheese?
EDIT: also, you completely missed my point. The money making machine is the AI because the copyright owners could just use them every time it produces copyright-protected material if we decided to take that route, which is what the parent comment suggested.
They should pay for the cheese, I’m not arguing against that, but they should be paying it the same amount as a normal human would if they want access to that cheese. No extra fees for access to copyrighted material if you want to use it to train AI vs wanting to consume it yourself.
And I didn’t miss your point. My point was that the reality is already occurring since people are already suing OpenAI for ChatGPT outputs that the people suing are generating themselves, so it’s no longer just a hypothetical. We’ll see if it is a money making machine for them or will they just waste their resources from doing that.
Media is not exactly like cheese though. With cheese, you buy it and it’s yours. Media, however, is protected by copyright. When you watch a movie, you are given a license to watch the movie.
When an AI watches a movie, it’s not really watching it, it’s doing a different action. If the license of the movie says “you can’t use this license to train AI, use the other (more expensive) license for such purposes”, then AIs have extra fees to access the content that humans don’t have to pay.
Both humans and AI consume the content, even if they do not do so in the exact same way. I don’t see the need to differentiate that. It’s not like we have any idea of the mechanism by which humans consume a content to make the differentiation in the first place.
I don’t think that’s a feasible dream in our current system. They’ll just lobby for it, some senators will say something akin to “art should have been always a hobby, not a profession”, then make adjustments for the current copyright laws so that they can be copyrighted.
There is actually already a website where people just recreated the bee movie by hand so idk it might actually work as a legal argument.
Copyright laws protects the ability of copyright holder to make money. The laws were created before AI and now obviously have to be adapted to new technology (like you didn’t really need copyright before the invention of printing). How exactly AI will be regulated is in the end up to society to decide, which most likely will come down who has the better lobby.
The Times’ lawyers must be chuffed reading this.
tweet is good, your body argument is completely wrong
I rather think the point is being missed here. Copyright is already causing huge issues, such as the troubles faced by the internet archive, and the fact academics get nothing from their work.
Surely the argument here is that copyright law needs to change, as it acts as a barrier to education and human expression. Not, however, just for AI, but as a whole.
Copyright law needs to move with the times, as all laws do.
I don’t think LLMs should be taken down, it would be impossible for that to happen. I do, however think it should be forced into open source.
Kids pay for books, openAI should also pay for the material access used for training.
OpenAI like other AI companies keep their data sources confidential. But there are services and commercial databases for books that people understand are commonly used in the AI industry.
OpenAI like other AI companies keep their data sources confidential.
“We trained on absolutely everything, but we won’t tell them that because it will get us in a lot of trouble”
That would be true if they used material that was paywalled. But the vast majority of the training information used is publicly available. There’s plenty of freely available books and information that you only require an internet connection for to access, and learn from.
This process is akin to how humans learn by reading widely and absorbing styles and techniques, rather than memorizing and reproducing exact passages.
Machine learning algorithms are not people and are not ingesting these works the same way a person does. This argument is brought up all the time and just doesn’t ring true. You’re defending the unethical use of copyrighted works by a giant corporation with a metaphor that doesn’t have any bearing on reality; in an age where artists are already shamefully undervalued. Creating art is a human process with the express intent of it being enjoyed by other humans. Having an algorithm do it is removing the most important part of art; the humanity.
Look… All I have to say is… Support the Internet Archive!
(please)
