@vpinon@pouet.chapril.org
Good blog post, braindead on ai. Sorry, but that is just a terrible take. The genie absolutely is going back in the bottle because these models are not sustainable and are so expensive to run there's no point. Also, to again quote the fdroid ai policy issue:
"The only thing that should matter in deciding policy is this codebase. And the arguments seem to narrow down to the following:
Pro-AI:
Everyone's doing it, we can't stop them, so we just need to ask them to disclose they're doing it.
Anti-AI:
Once we let AI into the codebase, we won't be able to remove it if we later decide we don't want it here without a lot of hard work done by the people who didn't actually introduce it in the first place. It's effectively the code equivalent of asbestos.
Even if people disclose their AI use and are held responsible for it, there's realistically no way to verify if all code that their model was trained on is compatible with our license because the training data on pretty much all of these models is proprietary, which is legalese for "you can't see it, so off".
There's a pattern of people effectively doing drive by PR with AI "help" where they show up, drop whatever their model returned from their prompt and not understand the code they're submitting, followed by feeding as prompts any feedback they get on the pr when reviewed. This is both disrespectful of the time of the project and code reviewers, but also effectively makes it impossible to hold those people accountable if a problem is found with their PR later (especially the legality of AI code relating to the licenses of the code they were trained on).
Cognitive surrender leading to both lower quality contributions in the short term, and a reduction in skilled coders in the future who are able to take on more senior roles like code review.
It may be just me, but the pros of allowing AI with an interim policy are literally just surrendering to peer pressure, and the cons are a lot of good reasons that require slowing down and actually considering if those issues can be mitigated to allow AI use in the future."
If you want my thoughts on a good policy, the Gentoo one is what kde should go with, even as an interim. Kde makes really good software, I would know, I am using it right now and have used it since I switched to arch. Once you allow the genie out of the bottle(in this case allowing ai code) then it becomes very very hard to put it back in and would take a lot of contributor time that could be spent adding to kde. I would also rather not learn another gui framework cause kirigami is really good
@tim@social.firestrike.de 
mastodon 4.6.8