Forum Rules for AI-Generated Content

[JColvin]

Hi All, 

We’ve been noticing an uptick recently in posts that appear to be generated by ChatGPT or other large language models. Some of these users have included links in their posts which appear to be associated with affiliate marketing scams, while others seem like they are out there to build SEO or just test and build up these tools. Spam is obviously not allowed, but the others are more of a gray area. 

From a moderation standpoint, we’re not entirely sure how to approach this, and wanted to solicit opinions from community members. For the time being, the policy is that users posting this kind of content will be treated the same as any other user. Moderation actions will be taken as appropriate based on Forum guidelines (available at the top of every Forum section and subForum, as well as here: https://digilent.com/reference/forum-guidelines), particularly when spam is involved. We’ve also updated guidelines to list what we would consider to be best practices while using these kinds of tools. 

Everyone is free to engage with any posts as they see fit, however, posting unverified outputs of these tools and not explicitly marking it as such is a potential waste of everyone’s brainpower and time. If you are reading a post or comment that seems off, please make use of the Report feature so that we can keep close eye on these posts and review for spam. 

Looking forward to hearing your thoughts. 

[Guest]

Can you point to some example posts with this kind of content? I haven't personally seen any of it.

Allowing these kinds of posts, when unchecked, will greatly diminish the value of your forum; I for one have no intention of perusing a forum that's is infested with AI-generated noise. This will be a big challenge in the years ahead for all forums that depend on human contributions, I suppose. Some of us have witnessed how Usenet fell from grace in a period of 10 years or so due to the influx of the rabble and bots; what is coming now with the LLMs is many times worse.

I think you should consider adding to your rules that (1) the intention of the forum is to foster interaction between humans; (2) AI-generated content should be marked as such and is only a valid contribution when a human being is at the wheel (i.e., it is okay to occasionally insert some AI-generated content, but don't let bots post autonomously), and the AI-generated content is clearly and unambiguously identified as such; and (3) you reserve the right to ban any and all accounts that you suspect of going against these rules.

Keeping the barbarians outside of the gates may require some rather unsavory policies, I'm afraid.

[vicentiu]

Here are some examples. Have a look at other posts from these users also.

https://forum.digilent.com/topic/26448-what-are-the-latest-trends-in-edtech-and-how-are-they-shaping-the-future-of-digital-education/#comment-79487
https://forum.digilent.com/topic/22580-dependency-on-libroborio_frc_chipobjectso19/#comment-76183
https://forum.digilent.com/topic/24066-jtag-hs3-stimulus-signals-inactive-during-autoconnect/?do=findComment&comment=79576

 

[asmi]

At the minimum, I would require AI-generated posts to be marked as such. It's hard enough to figure out what human user is actually trying to achieve (as opposed to what he says he wants to achieve, or often what he thinks he's trying to achieve), but wasting mental energy on unraveling AI posts is a bit too much to ask in my opinion considering many (including me) are trying to help people here voluntarily. With that, I don't really see how AI posts will help since anyone willing to get such responses can "ask" such AI system directly.

That said, I would simply forbid such posts altogether as in my opinion forums are meant for human beings.

[zygot]

It's hard to argue against requiring that all posts that involve AI generated content be identified a such. But how to you do that? From what I've read even AI tools designed to expose AI generated text is pretty bad at it.  How would Digilent even verify that a registered user is a human much less than that a human user isn't using AI tools?

I looked at the 'examples' of supposed AI content listed above and, while I have suspected AI posts, none of those were referenced and I wouldn't have suspected any of the ones listed as being AI content. There are plenty of bad (human) actors causing problems without thinking about computer generated issues.

Personally, I don't want to waste my time providing training data for an AI. I also don't want to waste my time trying to interact with an irrational human either, or do someone's homework assignment, or get caught up in other countless unproductive interactions. 

Digilent doesn't even control the management of it's website, so how can they resolve the AI or other issues that bother contributors?