Cross posted from https://lemmy.ca/post/70076035
Am I gonna have to switch to debian... cause I will if I have to. Pop OS is a branch off of ubuntu... so... I haven't seen much of a stance on this either way and I think this is by design. What are your opinions?
5 Comments
rumschlumpel@feddit.org · 5 pts · 16d
Debian is probably not a good option for anti-LLM users, either. https://feddit.org/post/34713404
hoserhobbes@lemmy.ca · 2 pts · 16d
I mean Linus is okay with LLM coding tools so hardcore anti-LLM people really shouldn't use the Linux kernel at this point...
ProdigalFrog@slrpnk.net · 5 pts · 16d
System76 accept AI contributions for their Cosmic DE, so they appear to be Pro AI.
AntiX (anti fascist/capitalist distro) and Gentoo have made clear stances against AI use, and I think a few others have as well, but unfortunately in the minority.
mmstick@lemmy.world · 1 pts · 12d
Accepting some contributions that used LLM assistance and being pro LLM are two entirely different things. As per the AI policy[0], use of LLMs is heavily discouraged. Contributions that fail to check every item on that list will not be reviewed. Those that correctly document their use of LLMs in each commit description will be heavily scrutinized with an unfavorable level of trust as the baseline. This filters out most LLM contributions so that only the highest quality contributions are reviewed. Lower quality contributions will likely be rewritten by hand in a completely different way if it solves a problem.
[0] https://github.com/pop-os/pop/blob/master/.github/PULL_REQUEST_TEMPLATE.md
ProdigalFrog@slrpnk.net · 1 pts · 7d
That is a better stance than most AI-accepting projects, and while I think with that method Cosmic will avoid unskilled vibe coded pull requests, it is ultimately permissive or tolerant of some level of Corporate LLM contribution, even if small.
In effect, that means potential contributors have little incentive from the project itself not to use corporate AI, with all the environmental and social problems they bring. It also still opens up S76 to potential legal issues down the road if any of those higher quality LLM contributions that don't need to be re-written happen to include copyrighted code from the AI's training data (which, as a profitable business, I would've thought S76 would be particularly wary of).
Personally, I'd love to see a policy that doesn't allow code from any corporate AI's, and instead only allows code contributions from self-hosted AI's that were carefully trained on truly public domain code (which does not currently exist, as even the ethical Swiss AI attempt is pretty flawed, but perhaps someday an effort like that is successful). Gentoo has such a policy now.