There’s nothing unique about that; we’re battling AI slop merge requests everywhere.
I don’t have the time to pay attention to Gwenview anymore though, sorry.
Indeed. That request isn’t important. I was solely asking whether, and, if so, how, one might notify the relevant project’s maintainer.
Though, in retrospect, perhaps, a comment is best. Apologies, if so… 
I hesitated to talk here for a long time. As I’m not a KDE dev. Let’s say first I’m anti-IA.
Still, I want to emphases than it’s seems there is two “kind of use” for AI for some people here:
– short, one-liner (kind of autocomplete)
– piece of software (whole complex function, state, etc…)
I think we can prove the first one is not efficient. We have LSP that use little energy, and AIg… Choosing to consume more for little convenience is a political choice. (and I could go on about the fact that it proves the abuse of boilerplate)
For the second one, it’s something new. I could argue this is a good use.
AI is great for classification, it is great for data analysis. It’s made for this use-case. Generating data with this statistical tool is highly inefficient. Like really. We are talking about compressing x second into y Joule… Is something we really want ?
One thing to ask ourself is what we want and how can we get what we want. Yes, not eating meat, no taking the plane or things like that don’t change the world. But if you structure yourself with organizations, you fight, you strike to promote the end of plane, you ask you government to tax plane seat, etc… You structured “group” of singular action start to structure the society as much as the society structure it. Here, we have a project, community-owned, and people trying to organize themselves to have an impact. Can this project be part of that ? Yes. Do developers want it ? Maybe not.
At the end, it’s a matter of how the decision-makers here see this society issue. I think this thread can change mind of people. I hope, at least.
I don’t think talking about copyright is really good here, as I think we should promote the end of “intellectual property” (and even “property”, but it’s maybe too much ^^').
(so yes, we are not perfect, still I will lobby for this as I think this is a good time, a good opportunity, a good environment to take a stance on this singular subject)
I don’t have much to add to the points made above but complete abstainment from genAI slop in all workflows is the only decent way forward. You can talk about some ethical compromises being nessesary, but this is not one of those!
There is nothing that would suggest KDE could be “left behind” by a vibecoded “alternative”; such software is way less solvent than the boosters would like you to believe, and developers reliant on mr. claude to do most of the work for them cannot be expected to be able to solve new complex problems or provide a stable, refined, cohesive product.
Adopting “AI” is a footgunning from a pure productivity standpoint too, as it’s been repeatedly shown not to result in any proven acceleration of work if the user cares whatsoever about the quality of their work. (If they don’t, that’s another story, but I assume you do!)
So while you should know better than “ethics aside”, even then it rots everything it touches. There is no ethical nor real practical usecase for LLMs. Plus the costs of the few that can at all pretend to be useful are ballooning already as the bubble approaches its breaking point.
I think it’s the opposite, at some time in the future, as more powerful and efficient hardware will become available for personal computers, there will be less need to pay for the cloud AI. I mean, it still may be better in some aspects, but local AI will become comparable to what we currently have in the cloud, so it will be enough for many use cases.
Maybe, but that will be decades out, assuming it even happens at all after the bubble bursts and people start focusing on the things AI is actually useful for and stops trying to shove it into literally everything.
Definitely not decades. Local are already getting quite good. And with each generation of consumer hardware, it gets more accessible for a wider range of enthusiasts, also things like the Nvidia ARM laptops may help. And of course it may get improved in some ways not requiring more power, such as better harness and other tools, may be some more specialized models, more efficient algorithms to execute the models, etc.
Relatively good garbage is still garbage. LLMs may help produce a higher quantity of content but with quality reversely proportional, and in a software project meant for real people to be using that’s the opposite of productive.
Much of the ethical issues and deskilling potential is unchanged in the case of local models. And decades or not, the bubble isn’t looking like it has even one year left in it.
To your point, we can already do a lot of useful things locally, such as creating automatic subtitles, speech to text, auto correct, object detection, and many other little useful things. However if we want to use it to do all the things major tech companies claim it can, then that will be decades out if we actually want to be able to do those things properly locally. So like maybe one day, but not any time soon.