A new twist in the "AI license laundering of chardet" story https://github.com/chardet/chardet/issues/327

[vonubelgarten@mastodon.sdf.org]

@cwebber even funnier with *closed source* proprietary Java or C# apps (and Android, perhaps?!) as these can be decompiled to a very ugly IR code that can be somewhat usable to guide a LLM!

[sprocketclown@mastodon.social]

@cwebber What constitutes laundering of copyleft codebases?

[gumnos@mastodon.bsd.cafe]

@SprocketClown

The way I read it in this context is that an existing codebase has license (whether GPL, LGPL, or proprietary or whatever), and that by "laundering" the codebase through an LLM, the output no longer retains the retains the license terms. In the US at least, the Supreme Court has ruled that LLM output is uncopyrightable.

So as @cwebber highlights, either the licensewashing works, in which case LLMs can scrub licenses off proprietary codebases giving a leg up on "reproducing" proprietary codebases into the public domain; or it doesn't work, in which case LLM-produced code becomes subject to the licensing of the original code.

[feld@friedcheese.us]

@cwebber

> Their claim that it is a "complete rewrite" is irrelevant, since they had ample exposure to the originally licensed code (i.e. this is not a "clean room" implementation). Adding a fancy code generator into the mix does not somehow grant them any additional rights.

The human didn't write the code, the LLM did. "They" which had "ample exposure to the originally licensed code" does not exist; "they" are ephemeral.

1. Start a fresh session / clean context, make it meticulously document the architecture, APIs, etc

2. keep those documents, throw away the code, start a new session with an LLM that has clean context and tell it to build off those documents.

That's clean room. If the original code was not in the LLM's context, it's not violating the license.

This is how you can do this. Proving beyond a reasonable doubt he didn't do it this way is going to require a lot of evidence nobody will have.

[feld@friedcheese.us]

@cwebber how is than an "obvious slop response"? I don't see anything odd other than the "core claim" statement but I would probably have phrased it similarly

[cwebber@social.coop]

@feld The headings, the emdashes, the framing of sentences, all classic AI "speech patterns" especially in markdown documents

[cwebber@social.coop]

@feld the author clearly at least was *assisted* in writing this response

[ralph_social@dresden.network]

Krass, dass sich AI-Firmen einfach Open Source Code schnappen und die Lizenzen "waschen" wollen.

Das ist genau das Problem mit dem aktuellen AI-Hype: Die großen Player denken, sie können einfach alles verwenden was im Netz steht. Und wenn's rechtlich eng wird, wird halt schnell die Lizenz geändert...

Respekt an Mark Pilgrim dass er sich dagegen wehrt! Open Source lebt von Vertrauen und klaren Regeln - nicht von solchen Manövern.

#OpenSource #AIEthics #Licensing

[gerardthornley@hachyderm.io]

@cwebber Reading through all the comments there left me wondering if anyone has (yet) hooked up an LLM to be a project maintainer. Interactions via issues and just let it loose. People would be utterly mad to ever include it in their supply chain, and yet people do do mad things.

[avirr@sfba.social]

@cwebber Isn’t this what forks are for?

[vv@solarpunk.moe]

@feld @cwebber the AI is still trained on the code beforehand

[vv@solarpunk.moe]

@feld @cwebber a "clean context" doesn't mean that there's no training data, it's still trained on a bunch of source code which likely includes the original

[feld@friedcheese.us]

@vv @cwebber proving the original was trained by the model or is in the model is quite difficult to do and is questionable whether or not it really matters anyway.

Chris Lattner was "trained on" GCC when he wrote LLVM. He studied it a lot. GCC compiles code C/C++ successfully, LLVM compiles C/C++ code successfully.

Both produce completely working bytecode and generally you don't *need* one compiler over the other to get an end result that is acceptable.

Should LLVM be allowed to have an Apache license because of this?

These are tough questions.

[pikesley@mastodon.me.uk]

@cwebber I felt my brain getting smoother as I read that

[svines@gts.svines.rodeo]

@cwebber Microslop committed to picking up the legal bill for anyone concerned about copyright issues with AI outputs from copilot so one could hypothetically use their tools to "clean room" implement Photoshop and then have Satya fight Adobe for your right to do so. Sounds fun to me!

https://blogs.microsoft.com/on-the-issues/2023/09/07/copilot-copyright-commitment-ai-legal-concerns/

[valpackett@social.treehouse.systems]

@svines @cwebber call that Project Photoslop

[borisbarbour@mastodon.social]

@svines @cwebber

This is topical:

"To protect against this, customers ... must not attempt to generate infringing materials, including not providing input to a Copilot service that the customer does not have appropriate rights to use."

[lukeharby@infosec.exchange]

@cstanhope @cwebber

Claude after being explicitly instructed not to base anything on LGPL/GPL-licensed code