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
Studied AI at uni. I’m also a cyber security professional. AI can be hacked or tricked into exposing training data. Therefore your claim about it disposing of the training material is totally wrong.
Ask your search engine of choice what happened when Gippity was asked to print the word “book” indefinitely. Answer: it printed training material after printing the word book a couple hundred times.
Also my main tutor in uni was a neuroscientist. Dude straight up told us that the current AI was only capable of accurately modelling something as complex as a dragon fly. For larger organisms it is nowhere near an accurate recreation of a brain. There are complexities in our brain chemistry that simply aren’t accounted for in a statistical inference model and definitely not in the current gpt models.
That knowledge is out of date and out of touch. While it’s possible to expose small bits of training data, that’s akin to someone being able to recall a portion of the memory of the scene they saw. However, those exercises essentially took what sometimes equates to weeks or months of interrogation method knowledge gained over time employed by people looking to target specific types of responses. Think of it like a skilled police interrogator tricking a toddler out of one of their toys by threatening them or offering them something until it worked. Nowadays, that’s getting far more difficult to do and they’re spending a lot more time and expertise to do it.
Also, consider how complex a dragonfly is and how young this technology is. Very little in tech has ever progressed that fast. Give it five more years and come back to laugh at how naive your comment will seem.
Dammit, so my comment to the other person was a mix of a reply to this one and the last one… not having a good day for language processing, ironically.
Specifically on the dragonfly thing, I don’t think I’ll believe myself naive for writing that post or this one. Dragonflies arent very complex and only really have a few behaviours and inputs. We can accurately predict how they will fly. I brought up the dragonfly to mention the limitations of the current tech and concepts. Given the worlds computing power and research investment, the best we can do is a dragonfly for intelligence.
To be fair, Scientists don’t entirely understand neurons and ML designed neuron-data structures behave similarly to very early ideas of what brains do but its based on concepts from the 1950s. There are different segments of the brain which process different things and we sort of think we know what they all do but most of the studies AI are based on is honestly outdated neuroscience. OpenAI seem to think if they stuff enough data into this language processor it will become sentient and want an exemption from copyright law so they can be profitable rather than actually improving the tech concepts and designs.
Newer neuroscience research suggest neurons perform differently based on the brain chemicals present, they don’t all always fire at every (or even most) input and they usually present a train of thought, I.e. thoughts literally move around in the brains areas. This is all very different to current ML implementations and is frankly a good enough reason to suggest the tech has a lot of room to develop. I like the field of research and its interesting to watch it develop but they can honestly fuck off telling people they need free access to the world’s content.
TL;DR dragonflies aren’t that complex and the tech has way more room to grow. However, they have to generate revenue to keep going so they’re selling a large inference machine that relies on all of humanities content to generate the wrong answer to 2+2.
Your first point is misguided and incorrect. If you’ve ever learned something by ‘cramming’, a.k.a. just repeating ingesting material until you remember it completely. You don’t need the book in front of you anymore to write the material down verbatim in a test. You still discarded your training material despite you knowing the exact contents. If this was all the AI could do it would indeed be an infringement machine. But you said it yourself, you need to trick the AI to do this. It’s not made to do this, but certain sentences are indeed almost certain to show up with the right conditioning. Which is indeed something anyone using an AI should be aware of, and avoid that kind of conditioning. (Which in practice often just means, don’t ask the AI to make something infringing)
I think you’re anthropomorphising the tech tbh. It’s not a person or an animal, it’s a machine and cramming doesn’t work in the idea of neural networks. They’re a mathematical calculation over a vast multidimensional matrix, effectively solving a polynomial of an unimaginable order. So “cramming” as you put it doesn’t work because by definition an LLM cannot forget information because once it’s applied the calculations, it is in there forever. That information is supposed to be blended together. Overfitting is the closest thing to what you’re describing, which would be inputting similar information (training data) and performing the similar calculations throughout the network, and it would therefore exhibit poor performance should it be asked do anything different to the training.
What I’m arguing over here is language rather than a system so let’s do that and note the flaws. If we’re being intellectually honest we can agree that a flaw like reproducing large portions of a work doesn’t represent true learning and shows a reliance on the training data, i.e. it cant learn unless it has seen similar data before and certain inputs provide a chance it just parrots back the training data.
In the example (repeat book over and over), it has statistically inferred that those are all the correct words to repeat in that order based on the prompt. This isn’t akin to anything human, people can’t repeat pages of text verbatim like this and no toddler can be tricked into repeating a random page from a random book as you say. The data is there, it’s encoded and referenced when the probability is high enough. As another commenter said, language itself is a powerful tool of rules and stipulations that provide guidelines for the machine, but it isn’t crafting its own sentences, it’s using everyone else’s.
Also, calling it “tricking the AI” isn’t really intellectually honest either, as in “it was tricked into exposing it still has the data encoded”. We can state it isn’t preferred or intended behaviour (an exploit of the system) but the system, under certain conditions, exhibits reuse of the training data and the ability to replicate it almost exactly (plagiarism). Therefore it is factually wrong to state that it doesn’t keep the training data in a usable format - which was my original point. This isn’t “cramming”, this is encoding and reusing data that was not created by the machine or the programmer, this is other people’s work that it is reproducing as it’s own. It does this constantly, from reusing StackOverflow code and comments to copying tutorials on how to do things. I was showing a case where it won’t even modify the wording, but it reproduces articles and programs in their structure and their format. This isn’t originality, creativity or anything that it is marketed as. It is storing, encoding and copying information to reproduce in a slightly different format.
EDITS: Sorry for all the edits. I mildly changed what I said and added some extra points so it was a little more intelligible and didn’t make the reader go “WTF is this guy on about”. Not doing well in the written department today so this was largely gobbledegook before but hopefully it is a little clearer what I am saying.