105
Copyrightability of LLM-generated code: Can we license “vibe code” into Free Software? (fsfe.org)
submitted 1 day ago by to c/opensource@lemmy.ml
24 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 4 points 1 day ago* (8 children)

work wholly generated by AI

Wholly

The push for protection of works created by AI has been spearheaded by Thaler, chief executive of neural network firm Imagination Engines. In 2018, he listed an AI system, the Creativity Machine, as the sole creator of an artwork called A Recent Entrance to Paradise

Further down:

So, it seems likely to me that this is where we will wind up: work wholly written by an LLM (the spirits or macaque in the above cases) will not be eligible for copyright; instead, the applicant will have to show sufficient creativity in the prompts (the questions in the Urantia case) and in editing and compiling the output.

It's like you googled it and copy pasted the link without reading it.

You can copyright something the moment you modify it or use it in something else. The law is more about how AI can't hold onto copyrights. It's to protect us from OpenAI saying they own everything you generate with their model. It stop copyright mills from running gens 24/7 and copyrighting every possible way to write a simple function.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 8 child comments
    •
  • [–] 4 points 20 hours ago (2 children)

    It is a bit more complicated than that, and the legal motivation from the Copyright Office is different than what you describe.

    One of the most informative cases, in my opinion, is the Copyright Office's back and forth on Zarya of the Dawn. In their view, even thousands of refining prompts cannot be sufficient to convey copyright protection on an AI generated image since it lacks the necessary control and connection to the person. Suryast is another helpful case. I think they're still very incomplete when you start to consider works like those from Jackson Pollock or, my personal favorite hypo, an artist with a prosthetic that uses statistical machine learning to control the end effector and thus the paintbrush. But those are just hypotheticals for now.

    I help chair a conference of IP attorneys and others who specialize in software licensing and AI so we've been talking about this for a while now. We actually just had the conference this week so I can probably share some resources or presentations if people really want to get nerdy about it.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 2 points 18 hours ago* (last edited 18 hours ago)

    I'd actually love some documentation because it is indeed very complicated.

    I'm also curious about your opinion on which way they will swing, since I see the courts kind of sticking to the middle more than anything.

    The Zarya case seems pretty clear. From an article though:

    The agency wrote that, unlike a photographer, users of Midjourney do not have the same degree of control over the final images. Photographers can control many parts of their photographs — the framing, the lighting, the subject, the presentation of the subject, the exposure time, the depth of field, etc — even though the camera ultimately captures the image.

    By and large, Sota models do give that kind of control nowadays, so I don't know if that would change things.

    I feel like it's going to end up as a case by case basis, where it depends on the feeling of effort put into it more than anything else.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 1 point 19 hours ago

    I suppose the general understanding thus far is similar to how much copyright a boss is entitled to the intellectual work of his underlings? I mean, that hypothetical artist is actually controlling the tool, vibe coders aren't controlling anything.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 3 points 1 day ago (4 children)

    If it's vibe coded, the person who wrote the prompt had minimal impact on the output. They can't, because they don't understand how it works. Their contribution is primarily limited to 1. writing a spec, which is often laughably vague, 2. testing the output to see if it works, and 3. if it doesn't work right, telling the AI to fix the problems.

    This was also in response to the 'trust me bro' snide comment. I'm not going to go into detail for something that is practically sealioning.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 3 points 1 day ago (1 child)

    What you say applies to slop, not anyone using AI seriously. Which is a small minority, but still.

    If you put in a lot of work, you might not be the author of the realization but you are author of the concept (assuming it is not trivial). Why should you not be able to protect your ideas?

    What is the status of a movie director compared to the actors?

    Who is usually mentioned when discussing whether the movie was well crafted or not?

    The best actor cannot save a shitty plot.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 day ago

    You're thinking of patents, not copyright. You can't copyright an idea, only the implementation of an idea. I could spend years perfecting a play, but when a troupe performs that play, I don't own their performance.

    And you underestimate the idiocy of self-proclaimed 'vibe coders'. Go into practically any FOSS community and see what they have to say about vibe-coded pull requests. The people making them have no clue what the code is doing, they're just chasing clout with minimal effort.

  • permalink
  • fedilink
  • source
  • parent
  • [–] 3 points 1 day ago (1 child)

    That's really going to depend on how strictly you define "vibe coded". If that means that someone put in a single prompt and the LLM spit out an entire program, then yes, I doubt that would be copyrightable.

    However, if the person "vibe coding" does understand how to program, and they use multiple prompts and organize the code to make a working product, then the code could be copyrightable by compilation:

    https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 day ago

    That's a real big 'if'. Someone who understands and shapes the code isn't really vibe coding. I might still not agree with what they're doing, but it's not vibe coding if they can take responsibility for every line of code produced.

  • permalink
  • fedilink
  • source
  • parent