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

221–230 of 623 posts

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

#221

Post by jbergqvist »

palmotea wrote: post #168
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…

> So: once it's not "hard" any more, does IP even make sense at all? Why grant monopoly rights to something that required little to no investment in the first place? Even with vestigial IP law - let's say, patents: it just becomes and input parameter that the AI needs to work around the patents like any other constraints. I think it still does: IIRC, the current legal situation is AI-output does not qualify for IP pr…

Does this matter in practice though? By modifying some of the generated code and not taking a solution produced by an LLM end-to-end but borrowing heavily from it, can't a human claim full ownership of the IP even though in reality the LLM did most of the relevant work?

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

#222

Post by kemitchell »

kelseyfrog wrote: post #103

In the corporate world, we've started using reimplementation as a way to access tooling that security won't authorize. Sec has a deny by default policy. Eng has a use-more-AI policy. Any code written in-house is accepted by default. You can see where this is going. We've been using AI to reimplement tooling that security won't approve. The incentives conspired in the worst outcome, yet here we are. If you want a diff…

Not Invented Here's long, slow mutagenic march toward full antibiotic resistance continues apace.

There is a fundamental corpo-cognitive dissonance, to boot. If "AI" is cheap enough and good enough to implement security-relevant software from `git init` repeatedly, why isn't it also cheap enough and good enough to assess and approve the security of third-party software at pace with internal adoption? Is there some basis to believe LLMs' leverage on production differs from its leverage on analysis of existing code?

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

#223

Post by winstonwinston »

> Blanchard's account is that he never looked at the existing source code directly.

That’s a weird statement while releasing the new version of the same project. Maybe just release it as a new project, chardet-ai v1.0 or whatever.

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

#224

Post by reverius42 »

Gigachad wrote: post #207
robmccoll wrote: post #206

Earlier quoted context omitted.

As described, this would not be the same thing. If the AI is looking at the source and effectively porting it, that is likely infringement. The idea instead should be "implement Minecraft from scratch" but with behavior, graphics, etc. identical. Note that you'll need to have an AI generate assets or something since you can't just reuse textures and models.

AI models have already looked at the source of GPL software and contain it in their dataset. Adding the minecraft source to the mix wouldn't seem much different. Of course art assets and trade marks would have to be replaced. But an AI "clean room" implementation has yet to be legally tested.

For copyright purposes I think there is an important legal distinction between training data (fed in once, ahead of time, and can in theory no longer be recovered as-is) and context window data (stored exactly for the duration of the model call).

I'm not sure there should be, but I think there is.

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

#225

Post by NewsaHackO »

Gigachad wrote: post #207
robmccoll wrote: post #206

Earlier quoted context omitted.

As described, this would not be the same thing. If the AI is looking at the source and effectively porting it, that is likely infringement. The idea instead should be "implement Minecraft from scratch" but with behavior, graphics, etc. identical. Note that you'll need to have an AI generate assets or something since you can't just reuse textures and models.

AI models have already looked at the source of GPL software and contain it in their dataset. Adding the minecraft source to the mix wouldn't seem much different. Of course art assets and trade marks would have to be replaced. But an AI "clean room" implementation has yet to be legally tested.

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 copier to copy Minecraft's source code into a new document and say you recreated Minecraft.

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

#226

Post by hyperman1 »

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…

I've always thought the opposite: IP law was created to make sure creativity stays hard, and hence controllable by the elites.

Patents came along when farmers started making city goods, threatening guilds secrets. Copyright came when the printing press made copying and translating the bible easy and accessible to all. (Trademark admittedly does not fit this view, but doesn't seem all that damaging either)

To Protect The Arts, and To Time Limit Trade Secrets were just the Protect The Children of old times, a way to confuse people who didn't look too hard at actual consequences.

This means that the future of IP depends on what lets the powers that be pull up the ladder behind them. Long term I'd expect e.g. copyright expansion and harder enforcement, just because cloning by AI gets easy enough to threaten the status quo.

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

#227

Post by NiloCK »

robmccoll wrote: post #206
Gigachad wrote: post #197

Someone 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.

As described, this would not be the same thing. If the AI is looking at the source and effectively porting it, that is likely infringement. The idea instead should be "implement Minecraft from scratch" but with behavior, graphics, etc. identical. Note that you'll need to have an AI generate assets or something since you can't just reuse textures and models.

A room "as clean" as the one under dispute (chardet) is very easy to replicate.

AI 1: - (reads the source), creates a spec + acceptance criteria

AI 2: - implements from spec

AI 1 is in the position of the maintainer who facilitated the license swap.

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

#228

Post by reverius42 »

nradov wrote: post #182
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…

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's public at all?)

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

#229

Post by integralid »

lukev wrote: post #219

I agree with the thrust of this article, that norms and what we perceive as good or desirable extend considerably beyond the minimum established by law. But a point that was not made strongly, which highlights this even more, is that this goes in every direction . If this kind of reimplementation is legal, then I can take any permissive OSS and rebuild it as proprietary. I can take any proprietary software and rebuil…

It goes in one direction only.

Companies can take open-source software and make a proprietary reimplementation. You can't take a proprietary software and make an open source GPL version.

I am absolutely certain that if you tried you would be sued to oblivion. But big company screwing up open source is not even news anymore. In fact I (still) believe that the fact that even though LLMs were trained on tons of GPL and AGPL or even unlicensed software it's considered ok to use LLM code in proprietary projects is example of just that.

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

#230

Post by azakai »

bmcahren wrote: post #99
sigseg1v wrote: post #60

Earlier quoted context omitted.

What does derivative mean here? Because IMO it means that the existing work was used as input. So if you used a LLM and it was trained on the existing work, that's a derivative work. If you rot13 encode something as input, so you can't personally read it, and then a device decides to rot13 on it again and output it, that's a derivative work.

LLMs do not encode nor encrypt their training data. The fact they can recite training data is a defect not a default. You can understand this more simply by calculating the model size as an inverse of a fantasy compression algorithm that is 50% better than SOTA. You'll find you'd still be missing 80-90% of the training data even if it were as much of a stochastic parrot as you may be implying. The outputs of AI are n…

> LLMs do not encode nor encrypt their training data. The fact they can recite training data is a defect not a default.

About this specific point, it is unclear how much of a defect memorization actually is - there are also reasons to see it as necessary for effective learning. This link explains it well:

https://infinitefaculty.substack.com/p/memorization-vs-gener...

Post reply on HN