The discussion which was opened 2 days ago has become pretty intense. The issue has been locked for now.
These fuckers are trying to open the floodgates for LLM slop in KDE.
So frustrating. Anybody who thinks these policies are ‘pragmatic’ or that ‘LLMs are bad but we can’t fight the trend’ needs to reassess. Just because something exists doesn’t mean it has to be used, especially with the dubious “benefits” on offer by these tools. Personally I will do everything I reasonably can to not run LLM code, and it’s exhausting that such major projects are relenting on this and brushing aside all of the valid criticism these things get. The madness of another hype cycle but it actively poisons existing projects this time
Also lol at the pro-LLM svin guy getting exposed as a literal nazi in that discussion
Even if the policy says that maintainers can reject slop PRs that won’t protect them from heat for rejecting PRs. And given how the discussion is going I don’t see the leadership actually standing behind maintainers that reject slop PRs out of principle even if the policy as written allows for that.
Establishing that a PR is bad enough that it should be rejected in a community project is more work than merging a passable PR.
deleted by creator
I’m getting tired of people acting like AI code isn’t insidiously hard to understand and dysfunctional. It’s super useful for random tiny short-lived projects. It’s not useful for ANYTHING that wants long-term maintenance, but people just love the here-and-now return. I say this as someone who’s job is swiftly turning into “make our AI spaghetti code we pooped out functional again because AI can’t handle it anymore”.
AI is literally good at ONE THING - making an output that LOOKS GOOD. Not an output that is good. Not an output that makes sense. An output that, at first glance, a human says “Looks okay”. That’s such a problem.
I’m not just trying to fight the national policy issues of datacentres, I am ALSO trying to fight the club bylaw issues of AI slop ruining valuable software.
deleted by creator
The problem is time wasted. LLMs allow for unnaturally high volume of low quality code, that would need to be inspected by a person. So even though yes, we can read LLM code, it would simply be unfeasable for the vast majority of projects and their maintainers.
deleted by creator
there are many documented cases on how much ai generated code leads to human burn out (reviewing increasing code quantity, low quality and hidden hard to identify problems) that leads to contributors leaving the project. also, one of the common side effects is the lack of learning and knowledge sharing that happens naturally on collaborative projects. both will affect the long term stability of the project, something that kde has put a good amount of effort until now. while on the short term it might increase the features released and more code, the bigger risk is loosing contributors, the possibility of training new ones and not depending on LLMs to develop software.
fight national policy issues with club bylaws
i doubt there are many people, outside of a couple of idealists, that think that project policies will have a big impact into something outside the project itself. all of what you say it’s true but it’s a different discussion to have
deleted by creator
waste of energy
i think you are viewing this from the wrong angle. it won’t probably have effect on the political environment but it will have effect on the health of the project itself. the most probable long term scenario after accepting llm generated contributions is a dependencies on them with no human contributors available.
i agree that the problem is bigger. there are more important fights to have and the effects of individual efforts are minimal. i also think than just throwing projects under the bus and just letting llms taking over them is not a good idea.
deleted by creator
this is a fatalist scenario, sure, but i think it’s a quite possible one. AI is a volatile topic and many mantainers can leave the project just by these type of policies (already happened for many other projects, gnome being the most nototious i remember). this will also end up burning other maintainers (reason on parent post), empower maintainers and developers that prefer llm generated code and tilt the balance in that way.
that it happens or not will depend on how things go in many fronts and how much the community works to avoid it (first step is already the petition to ban all llm contributions). none of this will happen in the short term and the end result is still far away but the fatilist scenario is the one most of the community wants to avoid and is the reason why there are many voices against even giving an inch on the topic.
/me shrugs and starts setting fluxbox + ROX up again.
Fuck 'em!
deleted by creator
