Banning the use of AI and ChatGPT

[Mr. Bambu]

I think they mean that the initial claim (AI or not) and the actual sources for the claim are two different things.

For example, the AI might say that Thor did X feat in Issue Y. In that same vein, a random user can say that Thor did X feat in Issue Y.

In either case, a claim is made, but the actual scans and source links still need to be provided to support said claim.

AI and a random user can claim whatever they want. At the end of the day, it's the staff that are verifying the sources provided for that claim.

That's all fine and well, as I said previously in the thread I am in favor of allowing AI to be used as a tool, particularly in the finding of sources. If it helps people make accurate CRTs, it gets my sign of approval. But given AI's tendency to misinform or misrepresent information, we, the staff, should be forewarned. That's all I want. I really don't think it's an outlandish standard.

 

[Antvasima]

I still agree with Bambu here.

What is the staff tally here so far?

 

[TWILIGHT-OP]

@Niko, @Super_Ascended_Sean_Pazdera this is staff thread.

 

[Mr. Bambu]

I suppose that serves as a bump for a conversation left trailing off.

 

[Agnaa]

I think they mean that the initial claim (AI or not) and the actual sources for the claim are two different things.

For example, the AI might say that Thor did X feat in Issue Y. In that same vein, a random user can say that Thor did X feat in Issue Y.

In either case, a claim is made, but the actual scans and source links still need to be provided to support said claim.

AI and a random user can claim whatever they want. At the end of the day, it's the staff that are verifying the sources provided for that claim.

I think the important distinction there is that we can punish a user if they consistently make up things like that.

Should we do the same to users who post AI text?

Alright, here's the updated version.

Regarding ChatGPT and other uses of AI

vsbattles.fandom.com

This reads like LLM output, which seems highly inappropriate for our rules on that sort of thing.

And because of that, it's generally not how we structure instruction pages on our wiki, and has some points that are frankly incoherent within the context of our site. So I staunchly oppose this as-is.

---

My view on some specific points of concern are:

• Profile Images: We should not use AI-generated images on profiles at all. It is perfectly fine to leave a profile without an image if it doesn't have official or fan art.
• Translations: To be resolved in this thread.
• Other: Otherwise, AI text will be treated with the same level of authority as if that user wrote that text themselves. Someone saying "Concrete has a destruction value of 650 MPa" without a source should face comparable scrutiny to an AI saying that. If they post LLM-based text that invents things, they should be treated as if they were a user deliberately inventing evidence themselves. And so, they should use it at their own risk, but we would have no reason to chase down accusations of people using AI outside the realm of translations and profile images.

I think this could be easily and briefly incorporated into our editing and discussion/site rules.

 

[JPTheGamer]

This reads like LLM output, which seems highly inappropriate for our rules on that sort of thing.

And because of that, it's generally not how we structure instruction pages on our wiki, and has some points that are frankly incoherent within the context of our site. So I staunchly oppose this as-is.

Then what are your suggestions for a better layout?

 

[Agnaa]

Then what are your suggestions for a better layout?

I think this could be easily and briefly incorporated into our editing and discussion/site rules.

 

[JPTheGamer]

Okay... that doesn't really answer my question. If it reads like LLM, how can I change that?

 

[Agnaa]

There's no point in explaining that. I'd rather rewrite it from the ground up myself.

 

[Armorchompy]

I am at least of the opinion that all LLM usage should be clearly signaled. But frankly I don't think it should be utilized to make arguments - maybe research given you'd ultimately still need to source that yourself, but I'm not gonna waste hours of my limited time on Earth having a debate with ChatGPT. I would honestly be in favor of a full ban regarding its use in debates (and profiles), but if we're incapable of that we should at least be as strict as possible with its use.

For the record while I don't think LLM translations should be usable for profiles, I don't think those should be banned per se. Saying something like "I asked AI to translate this scan, do you guys think I should have a translator check it?" or "The official translation says this, but I tried translating the source and it seems very different" seems fine.

 

[Antvasima]

I think the important distinction there is that we can punish a user if they consistently make up things like that.

Should we do the same to users who post AI text?

This reads like LLM output, which seems highly inappropriate for our rules on that sort of thing.

And because of that, it's generally not how we structure instruction pages on our wiki, and has some points that are frankly incoherent within the context of our site. So I staunchly oppose this as-is.

---

My view on some specific points of concern are:

• Profile Images: We should not use AI-generated images on profiles at all. It is perfectly fine to leave a profile without an image if it doesn't have official or fan art.
• Translations: To be resolved in this thread.
• Other: Otherwise, AI text will be treated with the same level of authority as if that user wrote that text themselves. Someone saying "Concrete has a destruction value of 650 MPa" without a source should face comparable scrutiny to an AI saying that. If they post LLM-based text that invents things, they should be treated as if they were a user deliberately inventing evidence themselves. And so, they should use it at their own risk, but we would have no reason to chase down accusations of people using AI outside the realm of translations and profile images.

I think this could be easily and briefly incorporated into our editing and discussion/site rules.

This seems reasonable to me.

@DontTalkDT @Mr. Bambu @AKM sama @Qawsedf234 @GrathOfLux

What do you think about this?

 

[Armorchompy]

I think anybody using AI text should at least state clearly that they are doing so.

 

[Qawsedf234]

What do you think about this?

Having used AI stuff for some required college stuff my views are pretty close to Armor's.

• Arguments can be enhanced or clarified with AI, but it shouldn't be entirely made by AI
• AI should be clearly stated as being used in some way
• Just because AI is used doesn't make it correct

A blanket ban is probably to much and it is helpful with grammar or trouble shooting, but no one should argue with a LLM and translations are nearly always suspect in some way.

 

[JPTheGamer]

There's no point in explaining that. I'd rather rewrite it from the ground up myself.

What would you suggest, then?

 

[DontTalkDT]

This seems reasonable to me.

@DontTalkDT @Mr. Bambu @AKM sama @Qawsedf234 @GrathOfLux

What do you think about this?

That would be my approach as well.

 

[Agnaa]

What would you suggest, then?

In our Editing Rules, under the Images or The User of Art section:

AI-generated fanart should not be used, unless endorsed by the creator of the piece of media it is a fanart of.

Translation stuff should be determined based on the other thread.

And the other part would go in our Discussion Rules:

Any AI-generated text included in a user's post will be treated as if they wrote it. Uncited claims will not be given any more credence than they otherwise would be, and any falsehoods will be treated as lies written by the post's creator, potentially being treated as a rule violation if they happen consistently.

---

I think anybody using AI text should at least state clearly that they are doing so.

AI should be clearly stated as being used in some way

Other than the difficulty of enforcing it, I think that just becomes kind of an annoying rule.

If someone is using an LLM to translate their text from a language they know better to English for posts on the site, do we really need all of their thousands of posts saying "This post was written with the assistance of AI" at the start of it?

 

[Qawsedf234]

If someone is using an LLM to translate their text from a language they know better to English for posts on the site, do we really need all of their thousands of posts saying "This post was written with the assistance of AI" at the start of it?

For a CRT, yeah. For general comments, it probably would start getting to much.

 

[Ptumri9]

I think anybody using AI text should at least state clearly that they are doing so.

I strongly disagree with this. I have no problem admitting that I use AI to fix the grammar of my posts. Everything I write still comes from me, my ideas, my arguments, my reasoning. The only thing the AI does is correct the grammar and spelling. I don’t always have the time to write a perfectly polished essay every time, and I doubt my English is good enough to meet that standard on its own. But does that really delegitimize what I’m saying?

I always double-check what the AI fixes. I make sure it doesn’t change the meaning or tone of what I wrote. The content stays exactly the same. It’s still my work. The only difference is that it reads more clearly. And yes, grammar and presentation do matter because people do perceive well-written arguments as more credible.

What concerns me is that studies have shown people tend to rate AI-assisted or AI-generated texts as less credible, less thoughtful, and view the authors as less competent, even when the actual content is identical. Just labeling something as "AI-assisted" can lead to bias, even if the AI only fixed typos, or even if it wasn't written by AI.

If I have to disclose every time I use AI to correct grammar, people might wrongly assume I didn’t write the post myself or that I didn’t think it through. That is simply not true. This kind of policy would just unfairly disadvantage people who rely on AI for language support, especially non-native speakers. It would also discourage people from using a valuable tool that improves clarity and communication without altering the substance.

I'm sorry, but I strongly disagree, this would only end up disadvantaging people.

Got permission from Mr. Bambu

 

[Mr. Bambu]

Your links seem to indicate you're also getting your sources from ChatGPT.

Trusting the AI will lead to problems in of itself. Notifying a staff member that a post is AI can only help us do our job better. We recruit staff based on presumed good faith and while bias can be subtle, I would like to think we can trust them to operate on the same good faith they extend to other areas where bias may be a concern.

If someone is using an LLM to translate their text from a language they know better to English for posts on the site, do we really need all of their thousands of posts saying "This post was written with the assistance of AI" at the start of it?

If Armor is discussing this in relation to my suggestion, then the recommendation was that CRTs should be headed with a notice that they were made with the use of AI. I wouldn't expect someone using AI to translate their posts to constantly mention it, I think.

 

[Ptumri9]

Your links seem to indicate you're also getting your sources from ChatGPT.

Trusting the AI will lead to problems in of itself. Notifying a staff member that a post is AI can only help us do our job better. We recruit staff based on presumed good faith and while bias can be subtle, I would like to think we can trust them to operate on the same good faith they extend to other areas where bias may be a concern.


If Armor is discussing this in relation to my suggestion, then the recommendation was that CRTs should be headed with a notice that they were made with the use of AI. I wouldn't expect someone using AI to translate their posts to constantly mention it, I think.

Yes, for this one comment I did get sources with the help of AI. But as I mentioned, I always double-check everything before sharing. AI is a tool, like a search engine or a grammar checker, and when used responsibly it can save time and improve clarity.

Just to add, my CRTs and posts have always been well put together, with solid sources and reasoning. I don’t really see why using AI to help with grammar or finding sources should be an issue if the end result holds up and if its carefully reviewed by me. Most of the time I don’t even use AI to find sources, this was more of an exception.

I understand that transparency can be helpful for moderation and staff. However, humans are inherently subjective, and no matter how much we try to be objective and unbiased, we cannot fully remove bias. Most of our biases are traits of the structure of human mind: we have biases as long as we are human beings.
May I ask, did the fact that I used AI to help find the sources affect how you viewed what I said? I don’t mean to speak for you, of course, but given the context, it seems likely that it did, even if only subconsciously.

Then again, I’m just a member and a relatively new one too. Staff have more experience and are likely better positioned to make the right call on this matter.

 

[Armorchompy]

If it's a respectable source, then there's no reason not to share that you used it. If your only complaint is that you're afraid people might be biased against its use then quite frankly that just isn't a particularly important concern, it's not like we can just ignore a valid argument because we don't like it in a VS thread.

 

[Ptumri9]

If it's a respectable source, then there's no reason not to share that you used it. If your only complaint is that you're afraid people might be biased against its use then quite frankly that just isn't a particularly important concern, it's not like we can just ignore a valid argument because we don't like it in a VS thread.

Perception shapes how we receive and evaluate an argument. Even if an argument is logically sound and based on reliable sources, the way it is perceived can still influence its approval. Ideally, all decisions should be based solely on the content, but in practice, perception plays a significant role. The human mind isn’t capable of being fully objective, nor is it immune to bias. Staff and members are human, not robots. That’s exactly why decisions will always be, to some extent, subjective, and why bias inevitably affects the final outcome. This is how I see it. If this rule were added, it wouldn't be the end of the world, but it would end up disadvantaging certain members.

 

[Armorchompy]

Perception shapes how we receive and evaluate an argument. Even if an argument is logically sound and based on reliable sources, the way it is perceived can still influence its approval. Ideally, all decisions should be based solely on the content, but in practice, perception plays a significant role. The human mind isn’t capable of being fully objective, nor is it immune to bias. Staff and members are human, not robots. That’s exactly why decisions will always be, to some extent, subjective, and why bias inevitably affects the final outcome.

Verbosely explaining an obvious aspect of the human psyche doesn't turn this into more of an issue than it is.

If this rule were added, it wouldn't be the end of the world, but it would end up disadvantaging certain members.

If you're so worried about the vague possibility of certain staff users having a mild unconscious bias against AI posts, don't use AI. Either way, it's not really a concern worth changing our rules to avert.

 

[Armorchompy]

If Armor is discussing this in relation to my suggestion, then the recommendation was that CRTs should be headed with a notice that they were made with the use of AI. I wouldn't expect someone using AI to translate their posts to constantly mention it, I think.

I would extend this to arguments made with LLMs in CRTs and matches too.

 

[JPTheGamer]

In our Editing Rules, under the Images or The User of Art section:

Translation stuff should be determined based on the other thread.

And the other part would go in our Discussion Rules:

---

So, should we get someone to implement these rules in the targeted pages?

 

[Wonder_of_Fantasy]

@Pturmi9 When used, or presented as a source of information, if the source is AI it should still be cited. The same is true for when it is treated as a search engine, as you say. And even then, it technically is not even a search engine either. Those who market premium LLM's proclaim such, but I think that that's little more than a marketing strategy that has produced that perception: "Ask the [insert premium AI bot here] anything!"

Regardless of my perspective on it, I think not disseminating it's use as far as source-providing goes, is tantamount to misleading others - even if not malicious. I think nobody has issue with it being used as a grammar-fixing tool. I even see people use it as a Translation tool at times, though I think that Translator Helpers should have final say should they find anything iffy or lost in translation as a result of that AI tool's output. But that is yet another personal thought.

More to the point, it's evident to me that there's a shift in today's culture where AI is going to have household use; if not already. But I don’t see that as an argument against regulating its use in some capacity.

 

[Wonder_of_Fantasy]

Having used AI stuff for some required college stuff my views are pretty close to Armor's.

• Arguments can be enhanced or clarified with AI, but it shouldn't be entirely made by AI
• AI should be clearly stated as being used in some way
• Just because AI is used doesn't make it correct

A blanket ban is probably to much and it is helpful with grammar or trouble shooting, but no one should argue with a LLM and translations are nearly always suspect in some way.

For the record, I also agree with this sentiment.

I will step back for now and update votes in the OP later. Again, thank you all for participating in the discussion.

 

[Ptumri9]

Verbosely explaining an obvious aspect of the human psyche doesn't turn this into more of an issue than it is.

If you're so worried about the vague possibility of certain staff users having a mild unconscious bias against AI posts, don't use AI. Either way, it's not really a concern worth changing our rules to avert.

Dismissing legitimate concerns doesn’t turn this into less of an issue than it is. Saying it’s “just part of human psyche” or that “you should just not use AI” is, in my view, an oversimplification and not a fair way to approach the topic. The fact that bias is well-known doesn’t mean we should ignore it. If it’s truly an unavoidable part of how people think, then it’s worth taking seriously.

The core of powerscaling is to build a consistent and logical framework, an analysis that aims to be as objective as possible based on feats and abilities. Bias directly undermines that goal, and that’s exactly why it should be taken seriously in a forum like this.

And no, saying “just don’t use AI” to avoid that "hypothetical" bias (which isn’t even hypothetical, considering you acknowledge biases exist and I linked studies showing this exact effect) is a weak answer. AI is a tool, just like a search engine or a calculator. You wouldn’t tell a professional not to use a calculator when it’s a tool that improves clarity, speed, and accuracy.

I’m sorry, but your comment came off as dismissive rather than genuinely engaging constructively with the concerns raised. If we care about fair analysis on this forum, then acknowledging bias and addressing it should matter. And no, I won’t stop using a helpful tool just because there’s no willingness to talk about a real issue.

 

[Ptumri9]

@Pturmi9 When used, or presented as a source of information, if the source is AI it should still be cited. The same is true for when it is treated as a search engine, as you say. And even then, it technically is not even a search engine either. Those who market premium LLM's proclaim such, but I think that that's little more than a marketing strategy that has produced that perception: "Ask the [insert premium AI bot here] anything!"

Regardless of my perspective on it, I think not disseminating it's use as far as source-providing goes, is tantamount to misleading others - even if not malicious. I think nobody has issue with it being used as a grammar-fixing tool. I even see people use it as a Translation tool at times, though I think that Translator Helpers should have final say should they find anything iffy or lost in translation as a result of that AI tool's output. But that is yet another personal thought.

More to the point, it's evident to me that there's a shift in today's culture where AI is going to have household use; if not already. But I don’t see that as an argument against regulating its use in some capacity.

I understand that it should be cited, and Mr. Bambu has also brought up valid and interesting points about how transparency helps moderators, which makes sense. What I'm saying is that there's a real risk of disadvantaging certain members of the forum by default. I'm not saying there shouldn't be any regulation at all, but if I use AI solely to fix the grammar in my arguments, and that alone leads people to subconsciously delegitimize what I'm saying, then that's a problem, because it puts at a disadvantage those who aren't fluent in English, those who struggle with writing, or those who simply don't have much time.

Whether it's used to find sources, translate, or fix spelling and grammar, the mere fact that there's a required label will inevitably lead to some degree of bias. And in a forum like this one, that kind of bias isn't helpful at all.

The world isn't black and white, and this is a complex issue and discussion. Everyone in this thread has brought up important and interesting points. I'm simply trying to point out that labeling could bring its own set of problems.

---

I believe I’ve expressed all of my concerns at this point, so I’ll refrain from further involvement in a staff thread.

 

[Armorchompy]

To myself give a final statement on Ptumri9's points:

While the concern is in a vacuum legitimate, it is incredibly minor. Someone misusing AI (a tool that is essentially tailor-made to put in less than the bare minimum effort) in dozens of possible, very real ways, is a far more serious concern than the mere potential of bias.

The reason it's so minor is that staff members are already biased. In fact every member of the community is biased against and in favor of a hundred different things, as you said it's just human nature. Attempt to highlight what is essentially one example out of many of "things that could cause bias" can make it look like a real issue but it is in fact really not a major concern. I hate to sound dismissive but if it happens, it's really not such a big deal. We can't worry about every potential micro-factor that could potentially get in the way of the perfect debate, when in practice hundreds of these factors already take place and are unavoidable.

 

[Agnaa]

So, should we get someone to implement these rules in the targeted pages?

No! We need agreement on what the rules should be first, and right now, some people disagree with my suggestion.

 

[Antvasima]

Can somebody summarise which staff members who currently think what here please?

 

[Agnaa]

Can somebody summarise which staff members who currently think what here please?

I put off doing this since I knew it was such a quagmire, but here it is

LLM Usage for Text (Other Than Translated Scans)​

Banning LLMs is unfeasible, and there's no reason to do so: 1 (Firestorm808)

Treat LLM-written messages as if the user wrote it (they're responsible for flawed logic, and we don't accept assertions as fact without sources): 1 (Damage3245)

Treat LLM-written messages as if the user wrote it, outside of consideration for promotions: 2? (Agnaa, DontTalkDT (?))

While LLMs can be used to gather ideas or formulate thoughts, they generally shouldn't do both. Usage of LLMs needs to be declared: 1 (Qawsedf234)

LLM-written messages must be checked; if they're wrong and the user posting it neglected to check them, that is a rule violation. While LLMs can be used to gather ideas or formulate thoughts, they generally shouldn't do both. Posters have an obligation to not generate unreasonably long posts, and using LLMs to stonewall is prohibited. Staff members can prohibit users from using AI within a specific thread, or across the forum/wiki as a whole. Usage of LLMs should not need to be declared: 1? (DontTalkDT (?))

LLM-written messages must be checked; if they're wrong and the user posting it neglected to check them, that is a rule violation. While LLMs can be used to gather ideas or formulate thoughts, they generally shouldn't do both. Posters have an obligation to not generate unreasonably long posts, and using LLMs to stonewall is prohibited. Staff members can prohibit users from using AI within a specific thread, or across the forum/wiki as a whole. Usage of LLMs in creating CRTs needs to be declared: 2 (Mr. Bambu (1, 2), Antvasima)

LLMs should not be allowed to be used in debates at all: 1 (DarkDragonMedeus)

LLM Usage for Images​

AI-generated images should not be used on the wiki at all, unless that is officially-endorsed/made art: 2? (Agnaa, DontTalkDT (?))

AI-generated images can only be used on the wiki if approved by staff in a CRT, but can be used in forum threads just as any other image: 1? (DontTalkDT (?))

AI-generated images can only be used on the wiki if approved by a staff member in any capacity: 2 (Mr. Bambu, Antvasima)

 

[Antvasima]

As I mentioned elsewhere, I am currently strongly inclined to vote against using A.I.s for translations, as they notoriously inserts hallucinations into a small part of their scripts, and we wouldn't know which parts, or if they would catch nuances and choose the context-correct interpretations in cases of multiple meanings in Japanese or Chinese for example.

We would almost never know which 2% or more of the translations that an A.I. would just "lie" about.

Also, I am currently more inclined to agree with DontTalk and Agnaa about usage of A.I. images.

 

[Agnaa]

I'm trying to save the scan translation topic for its own dedicated thread.

 

[Armorchompy]

LLM-written messages must be checked; if they're wrong and the user posting it neglected to check them, that is a rule violation. While LLMs can be used to gather ideas or formulate thoughts, they generally shouldn't do both. Posters have an obligation to not generate unreasonably long posts, and using LLMs to stonewall is prohibited. Staff members can prohibit users from using AI within a specific thread, or across the forum/wiki as a whole. Usage of LLMs in creating CRTs needs to be declared: 2 (Mr. Bambu (1, 2), Antvasima)

I dunno if thread mod votes hold any sway but I agree with this

AI-generated images should not be used on the wiki at all, unless that is officially-endorsed/made art: 2? (Agnaa, DontTalkDT (?))

AI-generated images can only be used on the wiki if approved by staff in a CRT, but can be used in forum threads just as any other image: 1? (DontTalkDT (?))

Does the first one forbid use of pics on the forum? Because generally while distasteful to me I don't see an actual problem with sharing an AI meme or something, but I'd want AI pics to be always kept off the wiki.

 

[Agnaa]

Does the first one forbid use of pics on the forum?

It does not, but also has no provision for explicitly adding that to the rules, aiming instead to implicitly allow it.

(But I really don't care if y'all want me to add a sentence saying that's fine)

 

[Antvasima]

To answer Armorchompy, I don't think that using A.I. images in an unofficial manner is any problem. I did so myself for my old story after all.

 

[Mr. Bambu]

My stance is that AI should be low priority compared to any other artwork, and if it comes down to it we should sway in favor of the latter. If nothing else exists, I tolerate it.

 

[FinePoint]

My stance is that AI should be low priority compared to any other artwork, and if it comes down to it we should sway in favor of the latter. If nothing else exists, I tolerate it.

I personally agree with this, but perhaps we should consider just how violently unpopular AI art is.

If we feature it publicly it may come back to bite us socially.