8
$\begingroup$

Should questions like this be off-topic?

I think this is not really a suitable type of a question for Stack Exchange.

It seems to me that the idea of Stack Exchange is to share knowledge, but in a case like this it's more like a job or a task of fixing output on an AI model. The person asking is not really interested in the knowledge, instead they want to get a product - a working script. Also it seems to me that questions like this are always going to be about very specific scripts that are only useful in very narrow context so it doesn't seem to me like they are likely to be useful to others in the future.

Should fixing AI scripts be considered officially off-topic on BSE?

$\endgroup$

5 Answers 5

Reset to default
5
$\begingroup$

I have already left my personal opinion about this in the comments of the question.

Personally, I do not see a suitable place for questions that arise from insufficient knowledge and the use of ChatGPT.

As you correctly say, in most cases such questions are problems that arise from the fact that the questioner has no idea about the matter, asks for a solution at ChatGPT (which also has no idea about the matter), but the solution is faulty, and the dear people here are then challenged to look for a very individual solution to a very individual problem.

And since the questioner actually only wants others to solve his personal problem quickly, without searching intensively for a solution himself, I think the effort is too high and the competence and time of the answerers would be better spent elsewhere.

This has little use for this platform from my point of view, and the questions should be closed as "not suitable for this site" or "too localized in its current form" in my opinion.

$\endgroup$
5
$\begingroup$

If someone posted random garbage and hoped the community would turn it into code for them, that would be out.

I guess, for me, the question is: 'How close is this post to doing that?'

OP has made some effort.. but it looks like guesswork.

The comment I would be tempted to make would be : 'Go back a few steps, follow some tutorials, code some simple things by hand, get an idea of the landscape you're working in. Then you'll be in a better position to see and fix the flaws in the slightly screwy maps GPT has given you to it, and the questions that remain will genuinely be your own, not GPT's.'

But I don't know how that translates to policy.

$\endgroup$
4
$\begingroup$

Personally I don't have a particular problem with ChatGPT related questions per-se, as long as they abide by our rules and code of conduct here.

Whether we want to accept them or even answer them here as a community is a very valid question of its own, that probably warrants a dedicated discussion.

The issue I see in this particular case is that, as you correctly described, it is essentially a "job post". Also more broadly seems to me like a general programming question that probably belongs in stack overflow instead.

It basically reads as a "fix it for me" post, with no apparent broader goal of learning, just an attempt to get some working solution with minimal effort.

$\endgroup$
3
$\begingroup$

Until we can find a counter-example, I propose to generalize all questions with ChatGPT code as "low effort".

It's not off-topic: it's about using Blender, and an error related to a Python script. Perhaps in this case it could be considered off-topic, as mathutils is not a Blender-specific library - though whenever I google python vectors, I only see numpy vectors and Blender related docs.

It's not too localized: while trying to import bpy.mathutils hasn't been a problem historically, I predict it might become more popular as chatGPT repeats its mistakes.

The ultimate problem of ChatGPT is that solving a single problem like this will help no one; after solving one bug, there's probably 10 more bugs waiting in line. Also agreeing to solve ChatGPT's mistakes is agreeing to eventually solve every possible syntax mistake.

Maybe it belongs to Code Review [Apparently Code Review accepts only working code, thanks Andrew!]

If someone writes a question with a good title, explains what he wants, then pastes ChatGPT code and describes what this code does that he needs, and where the code fails, the code snippet will serve as a clarification of the problem, rather than too localized definition of the problem, then such a question I would find to be valid, with the chatGPT part being a valuable addition.

$\endgroup$
4
  • $\begingroup$ I think if someone asks specific single question that is meaningful but related or illustrated by AI generated code, that's awesome. But I also think we are going to see people just pasting generated code and asking "why doesn't it work" or "I would greatly appreciate any help regarding the issue." My concern is that it is no longer a question then. I think Crantisz made a brilliant point. I like the concept of minimal level of knowledge to understand the answer a lot. If it's not there, is it still a question, or "fix it for me" request?.. If it's not a question, I think it is off-topic. $\endgroup$
    – Martynas Žiemys
    May 11 at 17:52
  • $\begingroup$ @MartynasŽiemys by definition of the "too localized" flag, your question should be useful to other (future?) users of BSE. Therefore I'm not so sure the answer has to be useful to the OP. It's probably a useful heuristic, however. $\endgroup$
    – Markus von Broady
    May 11 at 18:04
  • $\begingroup$ Sure. It does make sense that in addition to that, usefulness should also be judged in wider context considering more than just the person asking. $\endgroup$
    – Martynas Žiemys
    May 11 at 18:40
  • 1
    $\begingroup$ Just want to clarify that Code Review doesn't accept non-working code, so "fix my code" is off-topic there. The usual site for debugging questions is Stack Overflow, but if it's due to a trivial issue/typo (missing import statement), then it will be closed anyway. $\endgroup$
    – Andrew T.
    May 13 at 11:40
2
$\begingroup$

Although the answers given are good enough, I feel the need to add a few thoughts.

Of course, as a good question involves independent research and study of the topic, that is not the main point. The main thing is that the questioner must have at least minimal knowledge in the area in which he asks.

By minimal, I mean the level of knowledge that allows you to understand the answer, to assess its correctness and importance. You will never respond to a request to model something (within this site) but you will give answers to questions that a beginner may have in the process of performing this task.

Correcting code is asking you to do a task, but not finding an answer to a question. So that it is off-topic.

$\endgroup$
1
  • 1
    $\begingroup$ I agree with you but I think it also depends on the framing of the question. For example I think it can be beneficial to help someone who used a script that doesn't work in one of the recent versions because of an API breakage : It can be helpful to others having the same issue in a separate script using the same API functions. In the case of GPT it's almot always a very specific usecase that the asker didn't frame correctly within the constraints of GPT and so the script is terrible and is barely salvageable : It will only help 1 person to fix it, and only in this very specific scenario. $\endgroup$
    – Gorgious
    May 9 at 13:00

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .