Post At35NwLPNuLXpdF3Am by froztbyte@mastodon.social
(DIR) More posts by froztbyte@mastodon.social
(DIR) Post #AstfpKmGIwpUCJZ55U by whitequark@mastodon.social
2025-04-08T18:48:03Z
1 likes, 0 repeats
i've been sitting on this update for the last two weeks, but, here we gothe Amaranth project's policy on the use of LLMs / "generative AI" in one word: nohttps://github.com/amaranth-lang/amaranth/pull/1573
(DIR) Post #AstfpQDk4CmX4zSb7A by whitequark@mastodon.social
2025-04-08T19:17:41Z
1 likes, 0 repeats
the choice of the policy is directly related to and inspired by Servo's; both the current "no" policy, and the proposed "yes if de minimis" policy. I was aware of the proposal for quite a while before it became widespread knowledge recently and was intently waiting to see how it works out for themas best as I can tell, the only strong view in favor of "yes" is by an outsider without contributions, Sid Askary, who claims it might hurt funding (!)that's no way to run an open source project.
(DIR) Post #AstfpVn1MSyMLr0LIW by whitequark@mastodon.social
2025-04-08T19:18:35Z
0 likes, 0 repeats
sources:- https://github.com/servo/servo/discussions/36379#discussioncomment-12752742- https://github.com/servo/project/blob/main/governance/tsc/tsc-2025-02-24.md#ai-policy-review
(DIR) Post #AstqqhVuM9fMUwlmO9 by whitequark@mastodon.social
2025-04-08T19:22:13Z
0 likes, 0 repeats
@thejpster I've thought quite a lot about what's the right thing to do here, yeah. I think one of the main counterarguments people have is that "it's unenforceable", which is technically true. Originally I thought about going for a similar "de minimis" policy in the interest of keeping it more enforceable in practical terms, but that clearly sends the wrong message when used as the language of a policy, so I concluded that "no" is the appropriate choice
(DIR) Post #AstqqiWearrBdY7tBo by dramforever@mastodon.social
2025-04-09T01:03:01Z
0 likes, 0 repeats
@whitequark @thejpster If anything, for enforcability the rule should be *expanded* to any contribution of which, for example, the process of creation cannot be otherwise explained.
(DIR) Post #Astqqj52X1MXMCDLmq by whitequark@mastodon.social
2025-04-09T01:06:03Z
0 likes, 0 repeats
@dramforever @thejpster that's not what made me pause; rather, it is the fact that if you normalize the acceptance of LLM-encumbered contributions which do not readily appear as such (e.g. because they're too trivial) by way of not making it possible to mark them as such but doing nothing when it looks good enough, then the rejections will inevitably feel capricious and arbitrary
(DIR) Post #AstqqjwZK0BG2DG6Do by whitequark@mastodon.social
2025-04-09T01:06:30Z
0 likes, 0 repeats
@dramforever @thejpster "why does he get to use LLMs but not me?" when the former is an open secret among some parts of the community but not among the maintainers is a pretty awful question to find yourself answering
(DIR) Post #AstqqkWj9Z6VqMAya8 by dramforever@mastodon.social
2025-04-09T01:12:14Z
0 likes, 0 repeats
@whitequark @thejpster yes i would agree that enforcability is a separate issue on disallowing this practice. "cannot be always proven" should not be a barrier to making it a rule that it's not allowed
(DIR) Post #Astqql8esXRfjzvGhk by shironeko@fedi.tesaguri.club
2025-04-09T01:57:24.962683Z
0 likes, 0 repeats
@dramforever @whitequark @thejpster Yeah, the unenforceable argument is a distraction. There's no way for free software projects to tell if a submitter owns the copyright of the code they are submitting to begin with, so if we accept the unenforceable argument then we can throw out code licensing all together.
(DIR) Post #AsujnlptKBZeVujSim by whitequark@mastodon.social
2025-04-09T12:04:55Z
0 likes, 0 repeats
@shironeko @thejpster @dramforever I mean I don't really believe in the utility of copyright, so to me this isn't an interesting counterargument. Throw it out for all I know.
(DIR) Post #AsujnmvFHlS1soFFho by shironeko@fedi.tesaguri.club
2025-04-09T12:13:15.221329Z
0 likes, 0 repeats
@whitequark @thejpster @dramforever that's fine, don't put a license on the code then.
(DIR) Post #AsukVVFqKqA5pSiy3M by shironeko@fedi.tesaguri.club
2025-04-09T12:21:10.455217Z
0 likes, 0 repeats
@whitequark @dramforever @thejpster All I'm saying is if the unenforceable argument is accepted, then the project could not be a free software project in the first place because of the exact same concern. Therefore, any project that adopts a free software license should not accept the unenforceable argument.
(DIR) Post #AsulPHpBSF78ORr1EG by shironeko@fedi.tesaguri.club
2025-04-09T12:31:15.218260Z
0 likes, 0 repeats
@dramforever @thejpster @whitequark In other words, "a software project that accept code from the general public can have an defined set of copyright licenses" is logically incompatible with "a software project cannot assert where a given code contribution originates". The first is the basis of a free software project, the second is the unenforceable argument, they cannot both be accepted as true.
(DIR) Post #At35NwLPNuLXpdF3Am by froztbyte@mastodon.social
2025-04-13T12:37:25Z
0 likes, 0 repeats
@shironeko @thejpster @dramforever @whitequark eh, that's a bunch of bikeshedding. it misses the core part of what catherine already pointed out (the ethics and action of shifting the onus onto reviewers), and further fails to interrogate the motivation and dynamics of a person that would be doing something like this (dumping llm code on people without telling them about it). a stance of "no" is clear and direct: anyone found to violate the established norm can then be easily removed
(DIR) Post #At35NxAoInSmP3I6IC by shironeko@fedi.tesaguri.club
2025-04-13T12:52:45.565811Z
0 likes, 0 repeats
@froztbyte ? I fail to see how bike shedding applies here.