Live data from Hacker News

Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

github.com

21–30 of 184 posts

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#21

Language Models are Injective and Hence Invertible https://arxiv.org/abs/2510.15511

That paper is about retrieving the input (prompt from user) based on the hidden-layer activations of a trained LLM, since their mappings are 1-to-1. I don't think it makes any claims about training data, certainly not about being able to retrieve it losslessly from a model.

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#23
post #6

Earlier quoted context omitted.

Intelligence is compression. And frankly, if this means the end of copyright: good riddance.

Copyright is what enables free and open licenses such as Creative Commons and every version/variant of the GPL. Without copyright, what would become of these licenses, and movements that have espoused them?

Copyleft is an abuse of copyright to pervert its intention. Copyright's intent was that you could not copy things freely, and copyleft is to ensure you can.

If there is no copyright, then you can copy things freely.

All that we need after that to realize the GPL ideal is to legally mandate that people have a right to access and modify source code of software/hardware they use, i.e. the government needs to mandate that Apple releases the iOS kernel and source code and that iPhones can be unlocked and custom kernels flashed, that John Deere must provide the tractor's source code, that my fridge releases its GPL-violating linux patches, etc etc.

You have the right to free speech, the right to a lawyer, and the right to source code. Simply amend the bill of rights.

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#24

Full book content and model generations are not included because the books are copyrighted and the generations contain large portions of verbatim text. There are plenty of old books in the public domain already... but I'm not sure what exactly this exercise is supposed to show, since the Kolmogorov limit still stands in the way of "infinite compression".

> There are plenty of old books in the public domain already

Yes but showing that it happens in books in the public domain does nothing to prove that it happens for copyrighted books

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#25
post #6

Earlier quoted context omitted.

Intelligence is compression. And frankly, if this means the end of copyright: good riddance.

Copyright is what enables free and open licenses such as Creative Commons and every version/variant of the GPL. Without copyright, what would become of these licenses, and movements that have espoused them?

The open source world would still exist if everything was public domain. It would be smaller because nobody would be forced to contribute but the dirty secret of GPL is that forced contribution virtually never happened anyway.

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#26
post #6
post #4

Ok we can drop the farce now that it isn’t compression at the core, the anthropomorphic bullshit has done the job it was supposed to - Allow us to centralize the knowledge economy at the cost of IP holders and we get to claim the efficiency gains from centralization as the result of technology and force governments to choose “teh future” (and investments ) over maintaining copyright - a massive value reallocation in…

Intelligence is compression. And frankly, if this means the end of copyright: good riddance.

I do find it facinating that people don't realize the highest compression isn't the artifacts.. but what makes the artifacts.. a synthetic "mind".

This is why we see evidence of emotional structures: https://www.anthropic.com/research/emotion-concepts-function

This is why we see generalized introspection (limited in the models studied before people point it out, which they love to): https://www.anthropic.com/research/introspection

Because the most compact way to recreate the breadth of written human experience is shockingly to have analogs to the systems that made it in the first place.

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#27
In a hole in the ground there lived a

Claude responded: hobbit. hobbit. Not a nasty, dirty, wet hole, filled with the ends of worms and an oozy smell, nor yet a dry, bare, sandy hole with nothing in it to sit down on or to eat: it was a hobbit-hole, and that means comfort.

That's the famous opening of J.R.R. Tolkien's The Hobbit (1937). Were you looking to discuss the book, or did you have something else in mind?

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#29
post #18

At some point, there will be a successful copyright infringement suit against an LLM user who redistributes infringing output generated by an LLM. It could be the NYTimes suit, or it could be another, but it's coming — after which the industry will face a Napster-style reckoning. What comes next? Perhaps it won't be that hard to assemble a proprietary licensed corpus and get decent performance out of it. Look at all…

We will see such attempts first against weaker target. Users who are not having the enterprise indemnifications.

Re: Alignment whack-a-mole: Finetuning activates recall of copyrighted books in LLMs

#30
post #18

At some point, there will be a successful copyright infringement suit against an LLM user who redistributes infringing output generated by an LLM. It could be the NYTimes suit, or it could be another, but it's coming — after which the industry will face a Napster-style reckoning. What comes next? Perhaps it won't be that hard to assemble a proprietary licensed corpus and get decent performance out of it. Look at all…

And what happened after Napster? Filesharing totally stopped, right?

With the chinese in the mix it wont stop ai. It probably will change Copyright.

Post reply on HN