docs(contributing): add policy on generative AI/LLMs - #1586
Conversation
shiftinv
added
t: meta
| This policy **applies to all parts of the project**, including, but not limited to, code contributions, documentation contributions, and conversations in issues/pull requests. | ||
| An **exception** applies specifically for translating texts to interact with issues, pull requests, or discussions in the repository. | ||
|
|
||
| Not only is AI-generated content typically a detriment to the overall quality of the issue tracker and codebase, resulting in additional work for maintainers and often wasting valuable time; it also raises concerns about copyright and ownership of intellectual property, as LLMs have been shown to occasionally reproduce copyrighted material verbatim, and it remains unclear whether the output of LLMs is copyrightable in the first place. |
There was a problem hiding this comment.
I don't see the importance of mentioning ai's copyright issues here. I think this section could focus more on the problems of cognitive offloading that tends to be done with increased usage of llms.
There was a problem hiding this comment.
I'm not a lawyer (probably for the best), but I do think licensing is a concern in this context for open-source projects; e.g. under EU law, some AI-generated content seemingly may not be copyrightable at all, depending on the extent to which a human influenced the process (which can be questionable?).
I've shortened that section a little in 669b53a, in any case.
Mentioning cognitive offloading is a good call, but I'm a little torn on whether to include it, I already considered mentioning the potential environmental impact but scrapped the idea for now. I didn't want to turn this policy into a lecture on ethics and sustainability, but they're certainly valid points, so... not sure?
2 participants
Summary
Perhaps somewhat overdue at this point, this adds a policy to the
CONTRIBUTING.mdguidelines on the use of generative AI in contributions to this project/repository.I tried being as clear as possible wrt. the rationale for this particular stance on AI contributions; for the most part, it can simply become a licensing liability, while also eating up the reviewers'/maintainers' (currently rather limited) time.
I'll keep this as a draft for the time being to allow for more discussion, suggestions/reviews and other ideas are welcome <3
Checklist
uv run nox -s lintuv run nox -s pyright