I want to ask this openly because I’d rather follow a clear norm than guess.
When I answer questions here, I sometimes use an LLM to help structure or draft the reply. The technical content is still mine to stand behind. I test what I can, I say when I haven’t tested something, and I stay in the thread to follow up if the fix doesn’t land.
Recently one of my answers got flagged as AI, and it made me realise I don’t actually know where the line is. So:
Is AI-assisted drafting acceptable if the poster verifies the solution and owns the follow-up?
Should we be disclosing it? A tag, a line at the end, something standard?
Where does it clearly cross the line? Unverified answers, wall-of-text dumps, answering things you don’t understand?
I’m not raising this about any one person @ashokkarale or any one thread. I just think a lot of us are quietly in the same position and a stated norm would help everyone answer with more confidence.
@loginerror tagging you since this is a moderation question more than a technical one. Would be useful to have a stated position so it’s applied the same way for everyone, rather than case by case. Happy to follow whatever the norm turns out to be.
I want to start with saying thatI am much more of a “lurker” then someone who asks and answers questions here. It’s mostly because we haven’t really got to a stage where I develop robot processes frequently and I just like to keep myself up to date.
Personally I am not a huge fan of LLM sourced answers but I can see some cases where they could be of use, especially if it’s used to format the answer rather then giving the answer.
If I ask a question to a community like this one I would like to have a “real” answer rather then something I could ask copilot myself.
With that said, I think the guidelines are quite clear on this:
To help set the correct expectations for our users who opted to share their questions with our Community and receive advice from our Community experts, we would appreciate a shout-out to the source LLM used in case it assisted in providing an answer.
We embrace AI development and are aware of the strengths of the LLMs, and this is why we would like to acknowledge the amazing work done by their authors."
So if you use an LLM you should let everyone know this and personally I think you should also state if the whole answer is from an LLM or if you just translated something from your native language, asked it to structure an answer you came up with, etc.
It’s the accurate part that I mostly find that these AI-generated answers lack, and this bugs me.
Instead of actually reproducing the error and using an LLM to help solve the issue people prompt it with the topic at hand and copy paste the answer without making a sanity check/testing it.
I agree that most of the AI answers on here are just randoms trying to farm ‘solution’ points so they can get a gift card at the end of the month. With little concern for quality.
This discussion came multiple times..In this current state with llm and forum bots as well AI answers are acceptable, but verifying them before posting would be good.Also writing a note if you used AI would help users understand.at the end who asked the question can decide if they need to go through or not. As many times people already would have check with LLM and then land here so to consider their time as well its good to declare.and when someone does not declare and when you feel its LLM generated you can react with , so it tags it and again that feature is added to help the users itself.
This is kind of how we see it too. If a private conversation with an LLM solves the issue, it doesn’t need to be public.
We even make it easy for anyone via the latest changes to the Autopilot and via our extensieve uip cli toolkit and all the UiPath skills.
It is unfortunate how some people will try to take advantage to try to game our monthly forum rewards.
Serious question for a community feedback/brainstorm: what should we do? Should we scratch them/rework them?
I think scratching them would be the wrong route, but I think identifying answers that are AI or answers that are unproductive (or just not answering the question) would be a good start
, so it tags it and again that feature is added to help the users itself.