What’s your take on AI-generated answers in programming forums?

[Dani]

Pretty much since ChatGPT first came onto the scene, we have had a strong policy of not allowing AI-generated content to be copy/pasted into forum posts here. I know that some members, like rproffitt, really disagreed with that sentiment at first. Now that it's been a few years, and you've all had some time to wrap your heads around AI, what’s your take on AI-generated answers in programming forums? Helpful or harmful?

[Dani]

PS This is not meant to be a loaded question. Promise!

[Salem]

AI is just hot garbage.

Pretty much every post here that has a URL and an exclamation mark is an AI generated mash-up of buzzwords extracted from the rest of the thread.

[rproffitt]

When it rolled out Dani and I spent time on this. I took it as no AI output could be posted so that meant I could not write any article demonstrating how to use the new tools. We've gone back and forth many times but that ship has sailed.

In reality I've worked with many others in my circle to introduce them to how these tools cut dev time and costs. And reports from those in my circle have confirmed savings in time and cost.

I vote that fully generated AI answers and content is not to be allowed. But as part of a tutorial when the human is showing a method I don't get why "a strong policy of not allowing AI-generated content to be copy/pasted into forum posts" still stands.

This policy meant zero articles from me about how to use the new tools. These are tools. Just like a Google search or Wikipedia which for years teachers clutched their pearls about.

[Reverend Jim]

In response to a user who had a SQL question that was beyond me, I forwarded it to a friend who wrote SQL for decades. He asked ChatGPT, then examined the answer and said "This is what I would have written if I hadn't asked ChatGPT." I posted his (verified) solution (giving him the credit) and got my hand slapped for posting AI generated content. Then recently Dani posted something that she said was AI generated. So the ruling seems to be "AI content is not allowed, unless it is".

[Dani]

AI is just hot garbage.

Pretty much every post here that has a URL and an exclamation mark is an AI generated mash-up of buzzwords extracted from the rest of the thread.

Absolutely fascinating point of observation! 🚀

[Dani]

When it rolled out Dani and I spent time on this. I took it as no AI output could be posted so that meant I could not write any article demonstrating how to use the new tools. We've gone back and forth many times but that ship has sailed.

In reality I've worked with many others in my circle to introduce them to how these tools cut dev time and costs. And reports from those in my circle have confirmed savings in time and cost.

I vote that fully generated AI answers and content is not to be allowed. But as part of a tutorial when the human is showing a method I don't get why "a strong policy of not allowing AI-generated content to be copy/pasted into forum posts" still stands.

This policy meant zero articles from me about how to use the new tools. These are tools. Just like a Google search or Wikipedia which for years teachers clutched their pearls about.

As mentioned before, you completely misunderstood me. I never said that there could/should not be any AI-related articles or tutorials. It would be silly for a tech site to not allow anyone to talk about pretty much the biggest innovation in tech in decades! Including bits of AI-responses in tutorials about AI makes complete sense, and that was always allowed. What was not, and still is not, allowed around here was/is copying/pasting AI responses into topics that are not related to AI.

It seems like you agree with that policy, so that's good to hear, I suppose. Again, when it came to AI-related tutorials, you completely misunderstood me at the time. By the time we were finally on the same page, you had said that you were no longer really interested in AI nor had an interest in writing AI tutorials anymore.

[Dani]

In response to a user who had a SQL question that was beyond me, I forwarded it to a friend who wrote SQL for decades. He asked ChatGPT, then examined the answer and said "This is what I would have written if I hadn't asked ChatGPT." I posted his (verified) solution (giving him the credit) and got my hand slapped for posting AI generated content. Then recently Dani posted something that she said was AI generated. So the ruling seems to be "AI content is not allowed, unless it is".

Per my above post that was in response to rproffitt, the policy was always to not include copy/pasted AI responses in any topics/tutorials not directly related to AI. At the time, I was deleting dozens of AI-generated posts every single day, and it was feeling like a constantly frustrating game of whack-a-mole. To then have a moderator, who is essentially a DaniWeb representative, do something largely indistinguishable from all these crap posts I felt like would have made it harder to legitimize deleting the AI garbage. For example, if our official community rule is, "Do not post spun or AI-generated content," and I use that rule to allow the deletion of AI garbage, then how could I allow a moderator to post AI-generated content, garbage or not?

Then recently Dani posted something that she said was AI generated. So the ruling seems to be "AI content is not allowed, unless it is".

I think you're referring to where I posted recently that I used ChatGPT to get some ideas for conversation and topic starters. However, not a single sentence I've ever posted was copied/pasted from AI nor AI-generated.

So, I think I understand where you stand, rproffitt. How about you, Jim? Do you think the current rule is sufficient? Should it be modified in any way given how you feel that sometimes AI-generated content is appropriate? How would you rewrite the official rule to make it clear-cut to all when it is appropriate from when it isn't?

[Dani]

Wow, this topic seems more decisive than even I perhaps initially thought.

[Reverend Jim]

Decisive, or divisive?

[Dani]

Decisive, or divisive?

Sorry, typo! Divisive

[Reverend Jim]

My feeling is that a good answer is a good answer regardless of whether it was AI generated or not. In the case of the SQL query, it was vetted by a professional. I don't want to open that can of worms again. You asked my opinion and this is it.

[Dani]

I just thought of a circumstance that I think that AI-generated content should be allowed. I think if someone is not a native English speaker, it should be allowed for them to type their post in their native language and use AI to translate it into English and/or correct spelling and grammar of a post written in broken English. Thoughts?

[Reverend Jim]

I think that is reasonable.

[Dani]

I suppose the devil's advocate to that is that, as a moderator in charge of upholding the forum rules, how can one tell if a post that reads as spun content is low effort AI-generated or just poorly translated? Forum rules must always be objective and clear cut and easy to discern what is and isn't allowed.

[Dani]

I think that is reasonable.

What's reasonable? Just this one bit about translating, or everything I've said?

[Reverend Jim]

What's reasonable? Just this one bit about translating, or everything I've said?

I'll say that one bit specifically ;-)

Ambiguity. The devil's volleyball.

• Emo Phillips

[trcooke]

As a moderator on another programming forum, my main objection to AI forum content is that which is posted purely for the purposes of posting something. I see a bunch of posts that are just summaries of the rest of the thread, which adds no value, or meandering statements roughly on topic which again add little value and certainly don't answer any questions that have been raised. Many times the end goal of users posting AI generated content is to create a perception of legitimacy in the hope the moderators won't notice that spam that starts to creep in.

In general I don't have an objection to using AI tools to answer questions but ultimately I wish to interact with humans on programming forums. I'd be happy for someone to answer my question with something like "I had half an idea about this but here's what Claude says about it and it sounds pretty spot on to me..... <quote AI output>". You're posting AI content but it's a human presenting and validating it.

The problem with AI tools in general is that they present answers with absolute confidence and authority, even if it's wrong. Because of this it's hard to trust it, which is why automated AI content in forums is so detrimental.

[Dani]

The problem with AI tools in general is that they present answers with absolute confidence and authority, even if it's wrong. Because of this it's hard to trust it, which is why automated AI content in forums is so detrimental.

My understanding is that this is why Stack Overflow was quick to ban AI.