Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

writings.hongminhee.org

Post reply on HN
Print view View original

261–270 of 623 posts

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#261

Post by nradov »

zmmmmm wrote: post #249
nradov wrote: post #182

Earlier quoted context omitted.

Nothing changes for drug patents regardless of whether an LLM was used in the discovery process.

Even if all I have to do is tell my agent, "here is a patent for a drug, analyse the patent and determine an equivalent but non-infringing drug" and it chugs away for a couple of hours and spits out a drug along with all the specifications to manufacture it? I guess the state of play will be that for new drugs the original manufacturer will already have done that and ensured that literally anything that could be foun…

Yes, even so. Human drug researchers have been doing the same thing for decades. As soon as one pharmaceutical company launches a successful small-molecule drug everyone else jumps to find a minor tweak that will hit the same target (ideally with fewer side effects) while evading the patent. There is already specialized software to help with this process so I'm skeptical that LLM agents would be very helpful for this use case.

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#262

Post by throw-qqqqq »

VorpalWay wrote: post #235
amelius wrote: post #203

Earlier quoted context omitted.

You will probably run into design patents.

Software patents is not a thing in EU.

Very much is. "Software programs, as such" are exempt in the EPC article 52. However if the software program interacts with the world - if it has a "further technical effect" - it is patentable.

https://en.wikipedia.org/wiki/Software_patents_under_the_Eur...

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#264

Post by nradov »

reverius42 wrote: post #228
nradov wrote: post #182

Earlier quoted context omitted.

Nothing changes for drug patents regardless of whether an LLM was used in the discovery process.

Not sure why this should be true; the US Supreme Court recently chose to let precedent stand that AI creations are not copyrightable. https://www.reuters.com/legal/government/us-supreme-court-de... That also seems relevant for this whole discussion, actually -- if a work can't be copyrighted it certainly can't have a changed license, or any license at all. (I guess it's effectively public domain to the extent that it…

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.

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#265

Post by paxys »

zmmmmm wrote: post #162

The really interesting question to me is if this transcends copyright and unravels the whole concept of intellectual property. Because all of it is premised on an assumption that creativity is "hard". But LLMs are not just writing software, they are rapidly being engineered to operate completely generally as knowledge creation engines: solving math proofs, designing drugs, etc. So: once it's not "hard" any more, does…

"Hard" or "easy" has never been part of the premise.

A company spends a decade and billions of dollars to develop a groundbreaking drug and patents it.

I think of a cool new character called "Mr Poop" and publish a short story about him with an hour of work.

Both of us get the exact same protection under the law (yes yes I know copyright vs patent etc., but ultimately they are all about IP protection).

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#266

Post by munk-a »

crazygringo wrote: post #252
madeofpalk wrote: post #246

Earlier quoted context omitted.

A human reading a unit of work is not a “copy”. I’m pretty sure our legal systems agree that thought or sight is not copying something. Training an LLM inherently requires making a copy of the work. Even the initial act of loading it from the internet and copying it into memory to then train the LLM is a copy that can be governed by its license and copyright law

> Training an LLM inherently requires making a copy of the work. But that's not relevant here. Because the copyleft license does not prohibit that (and it's not even clear that any license can prohibit it, as courts may confirm it's fair use, as most people are currently assuming). That's why I noted under (1) that it's not applicable here.

It's absolutely prohibited to copy and redistribute for commercial purposes materials that you're unlicensed to do so with. This isn't an issue when it comes to the copy-left scenario (though it may potentially enforce transitive licensing requirements on the copier that LLM runners don't want to follow) but it is a huge issue that has come up with LLM training.

LLM training involves ingesting works (in a potentially transformative process) and partially reproduce them - that's a generally restricted action when it comes to licensing.

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#267

Post by crazygringo »

foresto wrote: post #260
crazygringo wrote: post #241

Earlier quoted context omitted.

Sorry, I misspoke. Transformation is what makes the LLM itself legal -- its training data is sufficiently transformed into weights. And so, a work being sufficiently transformative is one way in which copyright no longer applies, but that's not the case here specifically. The specific case here is essentially just a clean-room reimplementation (though technically less "clean", but still presumably the same legally).…

That's interesting, but it misses my point: The library's test suite and interfaces were apparently used directly, not transformed. If either of those are considered part of the library's source code, as the license's wording seems to suggest, then I think output from their use could be considered a work based on the library as defined in the license.

Legally that's been established as acceptable.

Google LLC v Oracle America assumed (though didn't establish) that API's are copyrightable... BUT that developing against them falls under fair use, as long as the function implementations are independent.

Test suites are again generally considered copyrightable... but the behavior being tested is not.

So no, it's not considered to be a work based on the library. This seems pretty clear-cut in US law by now.

Also, the LGPL text doesn't say "work based on the library". It says "If you modify a copy of the Library", and this is not a "combined work" either. And the whole point is that this is not a modified copy -- it's a reimplementation.

In theory, a license could be written to prevent running its tests from being run against software not derived from the original, i.e. clean-room reimplementations. In practice, it remains dubious whether any court would uphold that. And it would also be trivial to then get around it, by taking advantage of fair use to re-implement the tests in e.g. plain English (or any specification language), and then re-implementing those back into new test code. Because again, test behaviors are not copyrightable.

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#268

Post by SatvikBeri »

paxys wrote: post #259
NewsaHackO wrote: post #225

Earlier 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…

Is there a legal distinction between training, post-training, fine tuning and filling up a context window? In all of these cases an AI model is taking a copyrighted source, reading it, jumbling the bytes and storing it in its memory as vectors. Later a query reads these vectors and outputs them in a form which may or may not be similar to the original.

Judges have previously ruled that training counts as sufficiently transformative to qualify for fair use: https://www.whitecase.com/insight-alert/two-california-distr...

I don't know of any rulings on the context window, but it's certainly possible judges would rule that would not qualify as transformative.

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#269

Post by makerofthings »

If an AI can license-wash open source software like this then the licenses become meaningless. Which is fascinating. Commercial software cloning that is simple enough for an average person to drive is next and the ultimate form of piracy, see an app for $10? Don’t fancy paying? Just ask ChatGPT for a clone. Future is going to be wild.

Re: Is legal the same as legitimate: AI reimplementation and the erosion of copyleft

#270

Post by GardenLetter27 »

foresto wrote: post #260
crazygringo wrote: post #241

Earlier quoted context omitted.

Sorry, I misspoke. Transformation is what makes the LLM itself legal -- its training data is sufficiently transformed into weights. And so, a work being sufficiently transformative is one way in which copyright no longer applies, but that's not the case here specifically. The specific case here is essentially just a clean-room reimplementation (though technically less "clean", but still presumably the same legally).…

That's interesting, but it misses my point: The library's test suite and interfaces were apparently used directly, not transformed. If either of those are considered part of the library's source code, as the license's wording seems to suggest, then I think output from their use could be considered a work based on the library as defined in the license.

But the tests were transformed to the new language, they are not copied as-is.

Software patents would work as you describe, but not copyright.

Post reply on HN