[flagged]
Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
321–330 of 623 posts
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#322Wow, it feels like this argument rewired my brain. When I first read about the chardet situation, I was conflicted but largely sided on the legal permissibility side of things. Uncomfortably I couldn't really fault the vibers; I guess I'm just liberal at heart. The argument from the commons has really invoked my belief in the inherent morality of a public good. Something being "impermissible" sounds bad until you rea…
FWIW I like to explain it to folks like this: ignore all of your moral baggage around licensing and just focus on the fact that licensing is a legal tool of art that pretty much only becomes relevant in the context of threatening lawsuits.
BSD-type stuff is very simple because it says "here is this stuff. you can use it as long as you promise not to sue me. I promise not to sue you too."
Very simple.
GPL-type stuff is intrinsically more complex because it's trying to use the threatening power of lawsuits, to reduce overall IP lawsuits. So it has to say "Here is this stuff. You can use it as long as you promise not to sue me. I am only going to sue you, if you start pretending like you have the right to sue other folks over this stuff or anything you derive from it. You don't have the right to sue others for it, I made it, so please stop pretending and let's stop suing each other over this sort of thing."
Getting the entire legal nuance around that sort of counterfactual "I will only sue you if you try to pretend that you can sue others" is why they're more complex. And the simplest copyleft licenses like the Mozilla Public License have a very rigid notion of what "the software" is, so like for MPL it's "this file is gonna never be used in a lawsuit, you can edit it ONLY as long as you agree that this file must never be used by you to sue someone else, if you try to mutate it in a way that lets you sue someone else then that's against our agreement and we reserve the right to sue you."
Whereas for GPL it's actually kind of nebulous what "the software" is -- everything that feeds into the eventual compiled binary, basically -- and so the license itself needs to be a little bit airy-fairy, "let's first talk about what conveying the software means...", in various ways.
The interesting thing here is that as far as the courts are initially ruling, these from-scratch reimplementations are not human works and therefore are not copyrightable, which makes them all kind of public domain. Slapping the MIT license on it was an overstep. If that's how things go then Free Software has actually won its greatest sweep with LLM ubiquity.
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#323This take, which I've seen in a few different places now, seems 100% bonkers. A world where anyone can cheaply reimplement anyone else's software and use it on hardware of their own choosing in their own designs and for their own purposes is a free software utopia.
This isn't a problem, this is the goal. GNU was born when RMS couldn't use a printer the way he wanted because of an unmodifiable proprietary driver. That kind of thing just won't happen in the vibe coded future.
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#324Would software be more or less free in a world without copyright?
I argue more free. EULAs and restrictions on how+for what software can be used, like DRM, typically use copyright as their legal backing. GPL licenses turn that on it's head but that doesn't redeem the original, flawed, law.
This seems to follow the letter but not the spirit of the license. If this does pass legal muster, we can do the same to whatever proprietary software we wish, which makes a dramatically different but IMO better ecosystem in the end.
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#325Earlier quoted context omitted.
You might wish that were true, but there are very strong arguments it's not. Training on copyleft licensed code is not a license violation. Any more than a person reading it is. In copyright terms, it's such an extreme transformative use that copyright no longer applies. It's fair use. But agreed that we're waiting for a court case to confirm that. Although really, the main questions for any court cases are not going…
Transformative is not the only component of determining fair use, there’s also the economic displacement aspect. If you’re doing a book report and include portions of the original (or provide an interface for viewing portions à la Google Books) you aren’t a threat to the original authors ability to make a living. If you’ve used copyrighted books and turned them into a free write-a-book machine, you are suddenly using…
Sure. But it seems very difficult to argue that LLM's are harming that ability to make a living in a direct way.
This is for the same reason that search results or search snippets aren't deemed to harm creators according to copyright. Yes there might be some percentage lost of sales. And truly, people may be buying less JavaScript tutorial books now that LLM's can teach you JavaScript or write it for you. But the relation is so indirect, there's very little chance a court would accept the argument.
Because what the LLM is doing is reading tons of JavaScript and JavaScript tutorials and resources online, and producing its own transformed JavaScript. And the effect of any single JavaScript tutorial book in its training set is so marginal to the final result, there's no direct effect.
And the reason this makes sense is that it's no different from a teacher reading 20 books on JavaScript and then writing their own that turns out to be a best-seller. Yes, it takes away from the previous best-sellers. But that's fine, because they're not copying any of the previous works directly. They're transforming the facts they learned into a new synthesis.
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#326Earlier quoted context omitted.
The U.S. Supreme Court just declined to hear a case, thus upholding a lower court precedent that LLM output are not copyrightable: https://www.reuters.com/legal/government/us-supreme-court-de... What this means in practice is that (currently), all output of an LLM is legally considered to not be copyrightable (to the extent that it's an original work). If it happens to regurgitate an existing copyrighted work, though…
There’s several large settlements that say Anthropomorphic/OAI didn’t want to have legal precedent. In general if it’s not outright regurgitated it would be derivative.
The out of court settlements that avoid precedent don't mean anything in a broader legal context. Legally speaking, right now in the USA, output of LLMs is not copyrighted and cannot be copyrighted (without substantial transformation by a human).
I don't think this means the same thing as whether or not LLM output can infringe on someone else's copyright though (that does pose an interesting question -- can something non-copyrightable in general infringe on something copyrighted?).
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#327Earlier quoted context omitted.
You're really missing the point in multiple ways. First, precedents on copyright law are irrelevant to patent law. Second, AI generated works generally can be copyrighted under the human creator's name.
No, I think you are quite incorrect, at least on the latter point: "Lower courts upheld a U.S. Copyright Office decision that the AI-crafted visual art at issue in the case was ineligible for copyright protection because it did not have a human creator." Not eligible for copyright protection does not mean it can be copyrighted "under the human creator's name". It means there is no creative work at all. No copyright.
No, you're still missing the point. Did you even read the court's opinion?
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#328[dead]
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#329Someone should put this to the test. Take the recently leaked Minecraft source code and have Copilot build an exact replica in another programming language and then publish it as open source. See if Microsoft believes AI is copyright infringement or not.
The big question is: if copyrighted material was used in the training material, is the LLM's output copyright infringement when it resembles the training material? In your example, you are taking the copyrighted material and giving it to the LLM as input and instructing the LLM to process it. Regardless of where the legal cards fall, this is a much less ambiguous scenario.
There's a couple of different issues here that all get mangled together. If you're producing effectively the same expression that's infringement. You draw Captain America from memory, it's still Captain America, and therefore infringement. If you draw Captain Canada by tracing around Captain America that's also infringement but of a different type.
When it comes to software, again it's the expression that matters -- literally the actual source code. Software that does the same thing but uses entirely different code to do it is not the same expression. Like with the tracing example above, if you read the original source code then it's harder to claim that it isn't the same expression. This is why clean room implementations are necessary.
Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft
#330Earlier quoted context omitted.
That's why he is saying it's not equivalent. For it to be the same, the LLM would have to train on/transform Minecraft's source code into its weights, then you prompt the LLM to make a game using the specifications of Minecraft solely through prompts. Of course it's copyright infringement if you just give a tool Minecraft's source code and tell it to copy it, just like it would be copyright infringement if you used a…
What if Copilot was already trained with Minecraft code in the dataset? Should be possible to test by telling the model to continue a snippet from the leaked code, the same way a news website proved their articles were used for training.
I feel as though the fact that you are asking a valid question shows how transformative it is; clearly, while the LLM gets a general ability to code from its training corpus, the data gets so transformed that it's difficult to tell what exactly it was trained on except a large body of code.