135
AI’s Unpaid Debt: How LLM Scrapers Destroy the Social Contract of Open Source (www.quippd.com)
submitted 10 months ago by to c/programming@programming.dev
37 comments fedilink hide all child comments
 

TL;DR: The big tech AI company LLMs have gobbled up all of our data, but the damage they have done to open source and free culture communities are particularly insidious. By taking advantage of those who share freely, they destroy the bargain that made free software spread like wildfire.

all 37 comments

sorted by: hot top controversial new old
[–] 27 points 10 months ago (6 children)

So the issue is that AI strips the provenence of the open source contributors and then the output it spits out based on the data it consumed is not subject to the same open source licensing that applies to the open source projects, and these AI companies make profit from this but the open source contributors don't see a dime. We'll that's kinda always been the case though, so many amazing open source projects get coopted by tech giants like Microsoft and then repackaged as proprietary software for a profit, embrace, extend, extinguish, but back then they needed a team of developers to do that, now it's more automated I guess with AI

  • permalink
  • fedilink
  • source
  • hideshow 6 child comments
    •
  • [–] [S] 19 points 10 months ago

    Copyleft software isn't supposed to just be repackaged as proprietary, though. Permissive licenses, sure - but people know what they were signing up for (presumably) there.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 10 months ago (4 children)

    I believe the discourse that the FSF has managed to spread is greatly harming for the developers and communities. They are copyleft absolutists who believe no restrictions should be imposed in the use of our code, not even to megacorps that massively profit from it with oftentimes nothing in return.

    I am in the process of making a revised version of the copyfarleft Cooperative Software License with a lawyer and once its done I will switch most of my development to it, with a clear warning for any company that uses my code to fuck off (or pay me I guess).

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 8 points 10 months ago* (1 child)

    If the code used to train LLM was released with copyleft license then there’s only way to interpret how the output should be licensed. There’s nothing absolutist about it, just how GPL and such were intended to work. If you don’t like it, don’t use it to train models.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 6 points 10 months ago

    I think you misinterpreted my comment? I mean the Free Software Foundation is copyleft absolutist, as in, they will defend that model of licensing no matter what. I agree, of course, an LLM can be trained legally with GPL code, as you say that's how it is.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 7 points 10 months ago (1 child)

    You don't want copyleft. What you're looking for is called "copyright"

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 3 points 10 months ago* (last edited 10 months ago)

    No, its copyfarleft. Both it and copyleft USE copyright. I recommend you the Telecommunist Manifest on this topic, and you can find the stupid take of the FSF on this in here. I don't want copyleft anymore, I don't think it is enough. The FSF's justification is hipocritical and coward as they state that "...embedding that desire (ethical behavior) in software license requirements will backfire, by legitimizing fundamentally unjust power over others" while using the power of copyright themselves, and in a world where we already see bad actors profiting from collective work.

    Edit: Adding to this, the first word of the GNU GENERAL PUBLIC LICENSE is Copyright lmao

  • permalink
  • fedilink
  • source
  • parent
    • [–] -5 points 10 months ago

    no. you're trash.

  • permalink
  • fedilink
  • source
    •
    [+] -9 points 10 months ago* (2 children)

    These guys: AI bad! It takes jobs!
    Also these guys: Check out this thumbnail tho!

  • permalink
  • fedilink
  • source
  • hideshow 2 child comments
    •
    • [+] -22 points 10 months ago (25 children)

    destroy the bargain that made free software spread like wildfire

    If you didn't want your code to be used by others then don't make it open source.

  • permalink
  • fedilink
  • source
  • hideshow 25 child comments
    •
  • [–] [S] 26 points 10 months ago (15 children)

    Do you understand how free software works? Did you read the post? I'd love to clarify, but I'm not going to rewrite the article.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 15 child comments
    •
  • [–] 0 points 10 months ago (4 children)

    Also - this conclusion is ridiculous:

    By incorporating copyleft data into their models, the LLMs do share the work - but not alike. Instead, the AI strips the work of its provenance and transforms it to be copyright free.

    That is absolutely not true. It doesn't remove the copyright from the original work and no court has ruled as such.

    If I wrote a "random code generator" that just happened to create the source code for Microsoft Windows in entirety it wouldn't strip Microsoft of its copyright.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] [S] 0 points 9 months ago (3 children)

    That is absolutely not true. It doesn't remove the copyright from the original work and no court has ruled as such.

    Sorry, I just got around to this message. That is the idea of the provenance -- clearly, the canonical work is copyright. It is the version that has been stripped of its provenance via the LLM that no longer retains its copyright (because as I pointed out, LLM outputs cannot be copyright).

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 1 point 9 months ago (2 children)

    That doesn't make it "no longer copy-written" though. The original copyright holder retains their copyright on it. I can't see any court ruling otherwise.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] [S] 0 points 9 months ago (1 child)

    The output of the LLM can be incorporated into copyrighted material and is copyright free. I never claimed that the copyright on the original work was lost.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 1 point 9 months ago

    I highly doubt the law is settled on this topic and you're assuming it is. I can't see the courts accepting that your duplicate version of my work created through "magic" is not going to be a violation of my copyright. Especially if my work was included as input to the "magic box" that created the output.

  • permalink
  • fedilink
  • source
  • parent
  • [–] -1 points 10 months ago (9 children)

    Yes. And this is kinda hand-wavy bullshit.

    By incorporating copyleft data into their models, the LLMs do share the work - but not alike. Instead, the AI strips the work of its provenance and transforms it to be copyright free.

    That's not how it works. Your code is not "incorporated" into the model in any recognizable form. It trains a model of vectors. There isn't a file with your for loop in there though.

    I can read your code, learn from it, and create my own code with the knowledge gained from your code without violating an OSS license. So can an LLM.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 9 child comments
    •
  • [–] 3 points 10 months ago (3 children)

    No you can't. In the same way you can't watch a Mickey mouse movie and then draw your own Mickey mouse from what you recall from the movie.

    Copying can be done manually by memory, it doesn't need to be a 1:1 match. Otherwise you could take a GPL licensed file, change the name of 1 variable, and make it proprietary code.

    LLMs are just fancy lossy compression algorithms you can interact with. If I save a Netflix series in my hard drive, then re encode it, it is still protected by copyright, even if the bytes don't match.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 1 point 10 months ago (2 children)

    No you can't. In the same way you can't watch a Mickey mouse movie and then draw your own Mickey mouse from what you recall from the movie

    Yes, I can. I can create a legally distinct mouse-bases cartoon.

    You're right that if an llm gives you copyrighted code that it would be a potential problem. But the article saying that it somehow "strips the code of any copyright" is ridiculous.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 3 points 10 months ago (1 child)

    Is there anything in the LLMs code preventing it from emitting copyrighted code? Nobody outside LLM companies know, but I'm willing to bet there isn't.

    Therefore, LLMs DO emit copyrighted code. Due to them being trained on copyrighted code and the statistical nature of LLMs.

    Does the LLM tell its users that the code it outputted has copyright? I'm not aware of any instance of that happening. In fact, LLMs are probably programmed to not put a copyright header at the start of files, even if the code it "learnt" from had them. So in the literal sense, it is stripping the code of copyright notices.

    Does the justice system prosecute LLMs for outputting copyrighted code? No it doesn't.

    I don't know what definition you use for "strip X of copyright" but I'd say if you can copy something openly and nobody does anything against it, you are stripping it's copyright.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 0 points 10 months ago

    I don’t know what definition you use for “strip X of copyright” but I’d say if you can copy something openly and nobody does anything against it, you are stripping it’s copyright.

    Just what was stated in the fucking article

    By incorporating copyleft data into their models, the LLMs do share the work - but not alike. Instead, the AI strips the work of its provenance and transforms it to be copyright free.

    That's bullshit.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 2 points 10 months ago

    I can read your code, learn from it, and create my own code with the knowledge gained from your code without violating an OSS license. So can an LLM.

    Not even just an OSS license. No license backed by law is any stronger than copyright. And you are allowed to learn from or statistically analyze even fully copyrighted work.

    Copyright is just a lot more permissive than I think many people realize. And there's a lot of good that comes from that. It's enabled things like API emulation and reverse engineering and being able to leave our programming job to go work somewhere else without getting sued.

  • permalink
  • fedilink
  • source
  • parent
  • [–] [S] 2 points 10 months ago (3 children)

    I can read your code, learn from it, and create my own code with the knowledge gained from your code without violating an OSS license.

    Why is Clean-room design a thing then?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] -2 points 10 months ago (2 children)

    create my own code with the knowledge gained from your code

    Not copy your code. Use it to learn what algorithms it uses and ideas on how to implement it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 15 points 10 months ago* (7 children)

    "If you didn't want people in your house then don't have doors" buddy... That's not how anything works.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 7 child comments
    •
  • [–] -5 points 10 months ago (6 children)

    If you put a fucking sign on your door saying "come on in!" then don't be angry when people do?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 4 points 10 months ago* (last edited 10 months ago) (5 children)

    We do hang signs on the doors but they say something sightly different

    https://en.wikipedia.org/wiki/Open-source_license

    Public domain licenses are truly as you describe, but copyleft licenses are far from that. There are also many "source available" licenses which aren't open at all. Just because you can read a book doesn't mean you can print and sell it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] -1 points 10 months ago (4 children)

    Who is wholesale copying OSS code and releasing it under a non-compliant license with an LLM?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 3 points 10 months ago (3 children)

    Uh... Lots of people? That's kinda the problem. Maybe use a search engine. There are plenty of cases of LLMs 'laundering' copyleft code into (often) proprietary codebases. And that's just the most blatant and brain-dead obvious example; the use of GPL code to train commercial models is a bit more subtle and nuanced but no less nefarious, and the laws are currently unequipped to handle that part at all.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 0 points 10 months ago (2 children)

    You don't need an LLM to find and copy GPL code. The LLM isn't adding anything new here.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •