Some weeks or months ago I have returned after a break to read these forums more regularly. I have noticed a dramatic change compared to a couple years ago - a big part of posts, especially blueprints and custom components, seem to be written with LLM, usually without any kind of disclosure.
Now, I don’t think sharing work created with the use of those tools should be forbidden. I know how useful such tools can be, I use them as well.
But I’m rather annoyed when I’m reading yet another post here that was clearly written not by a human, pretending it was written by a human.
Is it only me bothered by this?
I’d very much like it if there was an enforced rule asking people to disclose if they used AI to write the post itself and if and how they used AI to create the project they are sharing.
I think it’s a different rule though, it’s specifically about helping others. I haven’t seen that rule being broken, at least on such a scale as when showing off new projects.
Oh, don’t forget it has to have sections and lists, ideally lists of lists. Nobody can read a paragraph of text, a list is so much better.
Last I saw the new rules stipulate that plagiarism (using LLM output as your own work) will not be allowed. You must use citations and preferably not copy the output word for word.
It’s a work in progress. This could change before release.
There is the annoying formatting of posts, but if you think about that it’s mind numbing, and all you can do is respond with TA;DR and ignore it…
I’m worried about the code that people are giving us and there is no way they understand it or would be able to support it if / when it breaks.
What I suggest is to respond to AI written code with a link to this thread… (nothing else)
I wouldn’t be that harsh. I see a lot of integrations and components that people enjoy using, that were clearly written with help of AI (or at least the description was written with one). So, well, I think it’s fine if they would be clear on that.
OK, I’ll bite. Does “TA” mean “Too AI”? I might start using that one!
I feel for the mods trying to figure out where to draw the line here. On the one hand, if I want AI-generated output, I’ll ask AI. I come here for the experience of real humans.
On the other hand, I do ask AI for help sometimes. There are lots of languages, commands, file formats and processes I’ve forgotten over the years, or maybe had less exposure to in the first place. With help from AI, I can dust off those forgotten or weak skills and do things more quickly than if I had to go back and hit the books. Are the results as efficient as I would like? Probably not. Are they sometimes flat-out wrong and I have to go back and prompt the stupid thing to get real? Yup. But in the end I’ve learned something and pushed through my project a bit quicker. And if someone else needs the product of our (me and AI) collaboration, I’m happy to share.
Like in maths exams, 50% for showing your working out calculations.
Don’t plagiarise. Mandate showing the conversation you used with your AI engine to arrive at the end result as part of the source code, so it can be reliably verified, replicated, and expanded on.
Should I ask for this the very next time I see obvious AI code posted here? Would it be too forward? Would people object? Should it be part of the push to tame this unfettered monster that is scrambling our lives with hallucinations?
Otherwise you just have AI slop. Incorrect, unreliable, and anonymous garbage, polluting our forums and leading everybody, including other AI bots, astray.
Example: I see your post may have traces of AI assistance. For the benefit of readers, could you advise what AI assistance you used, what sources were cited, and the conversation you used to arrive at your post.