AI guidelines

[ABCbits]

User @AnisEverRise created thread Solo Mining with Solar: My QSE Experiment with a Bitaxe Stack.. supposedly about his project, which include an image. Based on the thread text and image, there's implication it's real life photo taken by himself. But some people suspect it's AI generated, which later confirmed by himself.

Unless i misunderstood the normal rule, there's no requirement to mention source of non-text content. So i suggest this part of guideline should be updated, to also mention/disclose that the non-text content is AI generated.

That's going to be a tricky addition to the rules: many people embed images they found somewhere on the internet, which makes it increasingly difficult to know whether or not it's created by AI.

AnisEverRise's case basically comes down to lying, and as much as I don't like it, I don't think that should be governed by rules. It will be a slippery slope if lying is against the rules.

Fair point. How about more general guideline/suggestion, such as stating the image isn't created or taken by them (doesn't matter somewhere it's from internet, AI generated or other source they dont remember)?

[LoyceV]

Fair point. How about more general guideline/suggestion, such as stating the image isn't created or taken by them (doesn't matter somewhere it's from internet, AI generated or other source they dont remember)?

I'd vote against that too (not that I have any illusions of having a vote in this). In AnisEverRise's case: let's say there was such a rule. What would you want the forum to do about this? Deleting his post or banning him seems unnecessary to me. He received a few neutral tags, I'd say that's enough for pretending he took a photo that he didn't take.

[Alone055]

I'd vote against that too (not that I have any illusions of having a vote in this). In AnisEverRise's case: let's say there was such a rule. What would you want the forum to do about this? Deleting his post or banning him seems unnecessary to me. He received a few neutral tags, I'd say that's enough for pretending he took a photo that he didn't take.

I agree with that, because despite the fact that people might lie about some content to be their own but they actually aren't, but once they are caught for lying about it, they shouldn't be banned, either neutral or negative tags should do the work. I'll give an example about this.

We know that AI is developing at a rapid pace, sometimes, you can't even tell whether a content created using AI is actually AI-generated or an original. So, when there are contests on this forum where people need to submit designs and stuff, if someone uses AI to create a design and submits it, and they would obviously not say they used AI for it, but if they are later caught for using AI to participate in the contest, banning them shouldn't be the right thing to do, but they should surely get either a few negative or neutral tags telling everyone that the person is a liar and not trustworthy at all.

[nutildah]

Fair point. How about more general guideline/suggestion, such as stating the image isn't created or taken by them (doesn't matter somewhere it's from internet, AI generated or other source they dont remember)?

I'd vote against that too (not that I have any illusions of having a vote in this). In AnisEverRise's case: let's say there was such a rule. What would you want the forum to do about this? Deleting his post or banning him seems unnecessary to me. He received a few neutral tags, I'd say that's enough for pretending he took a photo that he didn't take.

Ultimately its up to the user to have some common sense when it comes how they are using AI on the forum. A simple application of the Golden Rule should suffice: would you feel deceived if someone else did the same thing you are doing?

This is how the trust system has sorted itself out over the years, without the need for administrative involvement.

We may all have differing opinions about what "deception" entails and that's a good thing.

Personally I'd prefer if someone clearly labeled an image or text as AI before posting it, unless its so obvious that it's not necessary. But I'm not gonna try to hold anyone to that, or expect others to. The image part is still a gray area for sure.

[Cricktor]

A simple application of the Golden Rule should suffice: would you feel deceived if someone else did the same thing you are doing?

I kinda feel this only works if people don't have malicious intents and have a society compatible social common sense.

Personally I'd prefer if someone clearly labeled an image or text as AI before posting it, unless its so obvious that it's not necessary. But I'm not gonna try to hold anyone to that, or expect others to. The image part is still a gray area for sure.

Definitely, I'd prefer having AI content of whatever form labeled as such. And if someone goes to the length to create such AI generated pictures, they should label them as such. But this is not going to happen unfortunately, as I believe and as you say, too.

[ABCbits]

You should not copy/paste text written by an AI into a post, with these exceptions:
 - If you clearly mention that something you're posting is AI-generated, and you only occasionally and appropriately post such AI-generated text, then that's OK.

Based on discussion on different thread, i have some question about this part.

1. What can be considered as mention? It is acceptable if someone mention it on their personal text or signature with very short or ambitious disclaimer?
2. How likely moderator take action (delete reply/thread or lock thread) only because there's no mention of AI usage if someone decide to make report?
3. What counts as only occasionally? Below 50% of all created post/thread in last month?

[LoyceV]

1. What can be considered as mention? It is acceptable if someone mention it on their personal text or signature with very short or ambitious disclaimer?

Why not just use quote-tags?

Hi I'm human

Your personal text doesn't apply to only one post, and if you ever change it, the post is left without "mention". Unless you intend to create a dedicated bot-account for it (ABCbots sounds kinda cool).

3. What counts as only occasionally? Below 50% of all created post/thread in last month?

I'd say 50% classifies as "frequently".

[Karl_3000]

3. What counts as only occasionally? Below 50% of all created post/thread in last month?

There will be different opinions in knowing who offended with AI posts. Someone that is having organic posts, making use of AI, indicating it to be AI to further explain something or to prove a point, which is how AI should be used, I do not think such a person will have AI posts up to 5% in his total post count.

[lovesmayfamilis]

3. What counts as only occasionally? Below 50% of all created post/thread in last month?

There will be different opinions in knowing who offended with AI posts. Someone that is having organic posts, making use of AI, indicating it to be AI to further explain something or to prove a point, which is how AI should be used, I do not think such a person will have AI posts up to 5% in his total lost count.

Proving your point of view using AI? Are you sure it will be your own point of view and not the millions of responses the AI ​​collects?

[Karl_3000]

Proving your point of view using AI? Are you sure it will be your own point of view and not the millions of responses the AI ​​collects?

I am alluding to, for instance to prove a statistical record or to bring out a valid information from external sources that are not from you but gotten from another source which can be through the help of AI. You can see that on some users posts which most users can put in a quote. You can use AI to get some answers and quote it, distinctly letting people know that you get the information with the help of AI.

[SeriouslyGiveaway]

I am alluding to, for instance to prove a statistical record or to bring out a valid information from external sources that are not from you but gotten from another source which can be through the help of AI. You can see that on some users posts which most users can put in a quote. You can use AI to get some answers and quote it, distinctly letting people know that you get the information with the help of AI.

If posters use AI content and put it in a quote, it's very encouraged but is it what AI users do with their posts in the forum?
Most of them did not post like this, but they tried to hide a fact that their content is AI generated by ignoring quotes, source, credit and this posting style is really discouraged by forum admins and community.

People want to discuss with other people, with human generated posts, but dislike to be lied and discuss with AI generated posts.

[Cricktor]

Based on discussion on different thread, i have some question about this part.

1. What can be considered as mention? It is acceptable if someone mention it on their personal text or signature with very short or ambitious disclaimer?

Some, currently temporarily banned, user tries to create this loophole:



There's an unreadable small font "Disclaimer:" in the upper left corner, there. I don't consider this as a free pass to use non-marked AI content as much as this user deems fit. This is a middle-finger to anybody else.

Using personal text or signature space to indicate AI usage is not acceptable in my opinion and should not be allowed as an AI free pass for obvious reasons.

As I've said before, I would want clear and unambiguous indication of AI generated or derived content, be it as a quote like LoyceV suggested or otherwise properly labeled. It simply should not be difficult to see what is human and what is AI generated or derived. Everything else is deception.

2. How likely moderator take action (delete reply/thread or lock thread) only because there's no mention of AI usage if someone decide to make report?
3. What counts as only occasionally? Below 50% of all created post/thread in last month?

I can't answer for moderators but I hope they'd take vigorous action against unmarked AI usage when reported.

"Occasional" is pretty ambiguous. For me "occasional" is more like less than 15-20%.

[theymos]

It is acceptable if someone mention it on their personal text or signature

No, it should be directly in the post, and not formatted to be unreadable.

2. How likely moderator take action (delete reply/thread or lock thread) only because there's no mention of AI usage if someone decide to make report?

I'd be strongly inclined to take some sort of action in all cases of undeclared AI usage, though in some cases the action might just be a warning, if the post is otherwise very constructive.

3. What counts as only occasionally? Below 50% of all created post/thread in last month?

Not many people have been banned for excessive declared AI yet, so I don't have enough experience to formulate more specific, reliable guidelines yet. It's at the moderator's discretion.

What I'd tend to ask is: if you were not using AI, how much less text would you be posting? If it'd be a reduction of more than 5-10% overall, then you're using AI excessively. For example, posting an AI's response to your 1-sentence prompt magnifies the amount of text you post by a lot, so you should do this only very rarely. But if you write a unique and on-topic 500-word post, give it to AI to clean up, and it produces ~500 words, then that counts against you a lot less, in my mind. (Though I wouldn't encourage doing that.)

[Mahiyammahi]

But if you write a unique and on-topic 500-word post, give it to AI to clean up, and it produces ~500 words, then that counts against you a lot less, in my mind. (Though I wouldn't encourage doing that.)

Since this idea is solely original to it's OP does it needs to be mentioned? I saw this types of refinement tends to lower AI detection scores . If AI don't change the tone of OP but just refined it as grammatical structure and more fluence it feels more real than AI.

[ABCbits]

1. What can be considered as mention? It is acceptable if someone mention it on their personal text or signature with very short or ambitious disclaimer?

Why not just use quote-tags?

Hi I'm human

Your personal text doesn't apply to only one post, and if you ever change it, the post is left without "mention". Unless you intend to create a dedicated bot-account for it (ABCbots sounds kinda cool).

I already use link, direct mention or quote, for example https://bitcointalk.org/index.php?topic=5561885.msg66650530#msg66650530. I asked it because i saw other account does such that. But i certainly didn't think consider someone would edit their personal text/signature in the future.

--snip--

Thanks for quick answer, let's see whether it'll encourage more people to report such cases.

[joker_josue]

But if you write a unique and on-topic 500-word post, give it to AI to clean up, and it produces ~500 words, then that counts against you a lot less, in my mind. (Though I wouldn't encourage doing that.)

Since this idea is solely original to it's OP does it needs to be mentioned? I saw this types of refinement tends to lower AI detection scores . If AI don't change the tone of OP but just refined it as grammatical structure and more fluence it feels more real than AI.

My question is: why refine your text, which will be used in the forum, using AI?

This forum is not a professional setting that demands rigorous writing. It is not a space for evaluating text or argumentation. Grammar, text quality, and other similar aspects are not analyzed.

At most, the person might be misunderstood due to some error in their text. But nothing that can't be resolved with a second text, a little simpler or clearer.

You could say that it might be because one isn't fluent in English, and uses it to translate the text or to improve the text. But to do this, you don't need to resort to generic AI tools. Simply use a text translator, which in 90% of cases translates the text correctly.

Yes, translation systems like Google Translate use AI. But they use it in a different way. An AI system, similar to ChatGPT, even with direct instructions not to alter the text, will always attempt to correct grammar, write based on the user's history, and eventually make informative adjustments.
In turn, a direct translation tool only translates the text that is entered into it at the time of translation. It does not analyze the user's history, does not interpret the text being written, nor does it have an opinion about it. It simply translates what is written as best as possible.

Therefore, be environmentally conscious and don't translate text using AI. Don't worry if a text has some grammatical errors when using it in the forum. If you want to translate text, use Google Translate or an equivalent.

This is an informal forum, from people to people. Even in more professional settings, there are no such formalities here. Be natural and yourselves.