AI guidelines

[Kazkaz27]

Theymos, I request your opinion regarding this Thread “Is AI Your Best Friend or Worst Enemy? 🤖”

I created the poll “Is AI Your Best Friend or Worst Enemy? 🤖” just a handful of hours before you created this thread and I believe my post may have prompted you to make this topic as it was in regards to getting community feedback on the topic of AI and rules for the forum.

Let me know if you have issue with it. It would be very enlightening to hear your response.

But doing this sort of evaluation (i.e. reading through hundreds of mediocre-to-bad posts) is really soul-sucking, so nobody wants to do it very much. I'm still thinking about ways to resolve this issue, possibly involving rejiggering how merit works.

I'll do it. I feel like I could be pretty objective about it.

Personally, I do not trust nutildah at all to be objective in regards to AI enforcment or creating a program to filter such posts. He is actually quite bias about the entire thing. If you read through his past comments and my responses on the very link I provided you, that alone proves it.

[notocactus]

The problem is our current AI detectors cannot distinguish between the nature of or motivation behind the prompt, so AI-translated posts often show up as 100% AI generated, even if the thought behind it was human. Some shitposters know this and this is their first defense for using AI: "I just use it to translate my text." The thing is their posts wouldn't be good even if written in their native language.

Even if their native language is not English, they can make their posts naturally in English even with mistakes in grammar, punctuation, word choices. They are not in any hurry to make posts for getting enough activity point, hunt merit or fill post quota. If a newbie, for example, just posts to ask questions, ask for help, it's not challenging to write some posts for that purpose even English is not good.

It's interesting that people who use AI for generating posts, or they said to translate their ideas from native language to English, are from nations where they can use English quite well.   For users who are from such nations - where English can be used well - but still use AI for posting, a permanent ban is well deserved.

In other words, there is no need or pressure of using AI to impress anyone in this forum.

[lovesmayfamilis]

AI written text is mostly generic and it lacks human emotions, and even if some AI models try to add emotions to the written text, still that's AI like emotion not human emotion. So, directly copy-pasting from AI should be avoided at any cost and now it's not allowed to do such a thing. Even if done, it should be properly mentioned with in quotation so others might not report it.

No, and not always. For quite some time now, there have been certain "humanizers," whose scripts you can simply indicate to the tool. For example, avoid "universal" phrases like "should be noted," "must be emphasized," "important to point out," and so on. Or use a friendly, informal communication style, ultimately setting a tone that perfectly aligns with human communication. Such scripts have long been circulating in online public groups, and many users are now actively using them. Again, the mega-thread, which regularly rehashes the same thoughts and ideas about buying, selling, holding Bitcoin, and so on, is the very swamp where such scripts are spread.
They evade detection by detectors, they paraphrase the presented text and maintain the topic of discussion, but the fact that these are written not by people, but by AI, is already ineradicable, unfortunately. As long as there are these huge mega-themes where users just need to take a post and paste a script into it, AI will be the "king" of such topics.

I write so many freaking posts and I mostly do it in James Joyce stream of continuous [spelled that wrong]

In the very near future, we will see AI with a communication style similar to philipma1957, or anyone else whose posting style is quite recognizable and individual.

[Don Pedro Dinero]

I haven't tested it, but I feel like a direct translation should not trigger AI detectors.

It does
I just checked it out with this your post
I translated the above to Spanish

And it came out as 100% Ai generated.

If it's not an original local language then I think it will be detected as Ai.

If we compare most self written local text on the forum with those translated by Ai, there would be much difference and a local language speaker would notice it.

I agree that such should somehow be quoted as an Ai translation to avoid any issues.

People can use what I proposed here:

How to use AI for translation without getting tagged or posts deleted.

If anyone hasn't read the thread, they can take the opportunity to read the stupid things that were said to me when I proposed that.

[PrimeNumber7]

* I acknowledge that not nearly enough people are banned for low-quality posts. The fundamental issue is that we don't want to ban someone just for posting one or two stupid posts, since that's subjective and might just be due to the person having weird opinions or being a poor writer. So to evaluate a user properly, we really want to be reading a sampling of 20-40 of their posts. But doing this sort of evaluation (i.e. reading through hundreds of mediocre-to-bad posts) is really soul-sucking, so nobody wants to do it very much. I'm still thinking about ways to resolve this issue, possibly involving rejiggering how merit works.

In all seriousness, have you considered using AI for this? At a minimum, it could help identify low-quality posters, without having to look at an excessive number of false positives. This could be especially helpful when dealing with newer accounts (and those with low post/activity counts), when the threshold for banning is generally lower.

AI could also be used to better enforce rules about posts being on topic.

I have long wondered if the source of large portions of low-quality posts over the years has been from AI. I get that some people put in very low effort for sig deals, but the level of intelligence (or the lack thereof) seems excessive for some of the spammers.

[SamReomo]

Your recommendation is good, I agree, as if they can spend time to interact with AI for learning and improve their grammar, it's better for their own written English but if they don't do that, they won't be banned here if the post content is from their own mind.

If you check the OP thread then it clearly says that the ones who even post AI generated text occasionally won't be banned either if they do it with proper mentioning. But, that's not the point, if someone fixes his/her spelling mistakes with the help of AI then that's not a wrong approach either but if that includes grammar and punctuation fixes or a rewrite by AI then such content isn't accepted if I'm not wrong.

No, and not always. For quite some time now, there have been certain "humanizers," whose scripts you can simply indicate to the tool. For example, avoid "universal" phrases like "should be noted," "must be emphasized," "important to point out," and so on. Or use a friendly, informal communication style, ultimately setting a tone that perfectly aligns with human communication. Such scripts have long been circulating in online public groups, and many users are now actively using them.

Those humanizers might work sometimes but those won't work all the times and the ones who rely on those will certainly gets banned for such type of posts if I'm not wrong. I even know that it's possible for AI to use same tone of a genuine person but the tone will be same, however it will still make worst type of posts that no one would like to read.

The AI generated posts are not enough for a healthy discussion especially on a forum like this where healthy discussion matters a lot. Some people who're good at training AI could train it in style of any human and that way it will be hard to find for AI detectors but still if they feed more data to that AI aside from the posts of that human or text of that human then still there'll be some chance for AI detectors to work on the generated text.

[Xylber]

* I acknowledge that not nearly enough people are banned for low-quality posts. The fundamental issue is that we don't want to ban someone just for posting one or two stupid posts, since that's subjective and might just be due to the person having weird opinions or being a poor writer. So to evaluate a user properly, we really want to be reading a sampling of 20-40 of their posts. But doing this sort of evaluation (i.e. reading through hundreds of mediocre-to-bad posts) is really soul-sucking, so nobody wants to do it very much. I'm still thinking about ways to resolve this issue, possibly involving rejiggering how merit works.

Some campaigns have humans reading and rating posts every week, with users making between 20-70 posts per week.
Do you think campaigns can help the forum in any way? It is a job that it is being already done anyway.

But keep in mind that even when there is a lot of souless posting, it is really hard to proove they use AI: Newest AI tools sometimes pass as "100% human". And when you catch them, they say "I'm using an AI translator".

[SamReomo]

But keep in mind that even when there is a lot of souless posting, it is really hard to proove they use AI: Newest AI tools sometimes pass as "100% human". And when you catch them, they say "I'm using an AI translator".

Well, even the newest AI tools of current time will not be able to make posts 100% human and similarly the AI detectors might sometimes do false detection. Even, if that's the case then how would the ones who wrote in English would come with words like they used AI translators? For them such words might not work and they won't be safe by breaking the rules.

[mindrust]

The problem imo isn’t copy pasting ai text. It is a bit deeper than that. Thanks to ai people can outsource thinking to llm models now. Before ai, people had to show effort, make a websearch, find the right and relevant content, make sense out of it, educate themselves and then answer… now you can copy paste someone else’s post to a llm, ask it how to counter this message, ai gives you instant ideas that has high probability of success and you don’t even have to copy paste text to make it work. You copy paste its brain. And no amount of restriction can stop it. The genie is out of the bottle.

[ABCbits]

So to evaluate a user properly, we really want to be reading a sampling of 20-40 of their posts.

My question is; is the ban going to be temporary or permanent ban like plagiarized posts.

It depends on the person's intent and history. If the person seems to only care about boosting their post-count, then that should typically be a permaban. Otherwise (eg. if it's just a misunderstanding of moderation policy), a warning or temporary ban would be appropriate.

I have some major skepticism about the enforcement. For example,

1. Some account mentioned on I'm seeing some sketchy stuff going on receive one or more accusation of spamming with AI/chatbot, where i'm sure many of their posts also reported. I even made a mass report for one of the account on https://bitcointalk.org/index.php?topic=5456516.msg66211874#msg66211874. But last time i checked, none of them are banned.
2. 77/81 posts made by account SilentEcho deleted by moderator (because of spamming with AI/chatbot), but this user still not receive permaban.

[LoyceV]

you can tell the AI to "Directly translate this text into Spanish: <something you wrote>"

How does this align with this rule:

27. Using automated translation tools to post translated content in Local boards is not allowed.

I assume this rule was created to stop people from posting their automated translations on many boards, and I don't think this should be allowed again. But I have found myself in the situation a few times where I wanted to respond to a post on a local board (usually after I was mentioned there), but couldn't do it, or only with a quote or in my shittiest handwritten German.

* I acknowledge that not nearly enough people are banned for low-quality posts. The fundamental issue is that we don't want to ban someone just for posting one or two stupid posts, since that's subjective and might just be due to the person having weird opinions or being a poor writer. So to evaluate a user properly, we really want to be reading a sampling of 20-40 of their posts. But doing this sort of evaluation (i.e. reading through hundreds of mediocre-to-bad posts) is really soul-sucking, so nobody wants to do it very much.

Good description: "soul-sucking". That's why I now often place them on Ignore after just one report, but if they're not banned, they'll keep spamming. Due to the lack of bans reporting their posts feels like such a massive waste of time, as it's takes much more time for a human to read than it takes for a bot to create it.

I'm still thinking about ways to resolve this issue, possibly involving rejiggering how merit works.

Just yesterday, I wrote down my realization that the Merit system is now luring shitposters to the tech boards:

When I visit the tech boards now, I see rows of posts made by shitposters who are already on my Ignore list. The tech boards used to be the best boards on this forum, free from spam, but thanks to chatbots and Merit, it's now getting spammed. In this case, I'd say the Merit system is counter productive.

I'm very optimistic about AI on the whole

I wish I was as optimistic as you are. Apart from having seen far too many movies that show how this can go wrong, I see it as more annoying in real life already. When I call a company for example, I get to talk to a machine because it's cheaper than an employee. I read more and more articles that don't make sense because it's obviously generated by a bot. I see it as mostly a very big money maker for very few people, and used for cheating by many others.

But doing this sort of evaluation (i.e. reading through hundreds of mediocre-to-bad posts) is really soul-sucking, so nobody wants to do it very much.

I'll do it. I feel like I could be pretty objective about it.

Most of the time, I can't even get myself to read the whole post if it's - as I like to call it - chatbot verbal diarrhae. It's just too terrible to process.

[CryptoHeadlineNews]

you can tell the AI to "Directly translate this text into Spanish: <something you wrote>"

How does this align with this rule:

27. Using automated translation tools to post translated content in Local boards is not allowed.

I assume this rule was created to stop people from posting their automated translations on many boards, and I don't think this should be allowed again. But I have found myself in the situation a few times where I wanted to respond to a post on a local board (usually after I was mentioned there), but couldn't do it, or only with a quote or in my shittiest handwritten German.

Yes, you have a good and solid point on what you just stated above regarding the new guide Sir Theymos just gave by allowing the use of A.I by forum members to be able to translate their own words from one language to another. Because by so doing, you are indirectly just allowing the use of A.I, because I'm 100% sure that many people will abuse and misused this opportunity, and when been caught, they will just claim and argue that they just used A.I to translate it but never did copy and paste. Since Theymos just clearly stated that it is allowed.
So in other to avoid this, there should be zero tolerance on the use of A.I, unless if the post/thread clearly states it was generated with the use of A.I. And such post shouldn't be counted as eligible for signature campaigns post. Because that's the only way to discourage people from posting A.I garbage.

If someone wants to translate a post , he or she should go hire a translator on the service board, rather than using A.I. Because by so doing, it will make the forum look far more professional. That's just my humble opinion.

[joker_josue]

In this area I agree with @theymos
A forum where technology is a central theme, distancing it from our daily lives, doesn't make much sense.

Logically, I don't agree with posts generated 100% by AI. Doing so by copying and pasting, without citing the source, is really wrong. Now, this shouldn't stop people from using it to extract ideas or informative details.

So I think you really shouldn't use AI to write our posts. But, if someone uses it to translate some text, I should use it as a true translator, and not as a writing assistant.

At most, take the ideas he gives and write them in your own words!

The problem is our current AI detectors cannot distinguish between the nature of or motivation behind the prompt, so AI-translated posts often show up as 100% AI generated, even if the thought behind it was human.

I haven't tested it, but I feel like a direct translation should not trigger AI detectors. If it does, we'll just have to use our best judgement. Ultimately, it's about evaluating whether the user is a net-positive or net-negative for the forum (while also keeping fairness in mind).

I have been pointing out for some time that these AI tools are highly prone to errors in analysis. There are several studies that show how these detectors fail even with ancient texts, claiming they were written by AI – texts thousands of years old.

We have many writing patterns, and AI writes based on an analysis of these human patterns. Therefore, it's not so difficult for a human to write originally, with a pattern that appears to be AI-generated. Therefore, I think it's a mistake to evaluate a text simply by looking at these AI text detection systems.

The text needs to be analyzed in another way, at least here in the forum. Look at several of the user's posts, as theymos says, and observe if their writing style has significant differences. Sometimes it doesn't take much to notice; just the way they respond to a question raised about the suspicious post is enough.

you can tell the AI to "Directly translate this text into Spanish: <something you wrote>"

How does this align with this rule:

27. Using automated translation tools to post translated content in Local boards is not allowed.

I assume this rule was created to stop people from posting their automated translations on many boards, and I don't think this should be allowed again. But I have found myself in the situation a few times where I wanted to respond to a post on a local board (usually after I was mentioned there), but couldn't do it, or only with a quote or in my shittiest handwritten German.

I don't want to turn this into a debate about these "rules"... I think that this "rule 27" (like others...) is often misinterpreted.

The idea behind this rule is to prevent someone like me, a Portuguese speaker, from going to the Russian tab to chat, because there are more topics there than in the Portuguese tab. But that shouldn't stop me from responding to something that's there, where I was quoted, or perhaps about a subject very close to me.

But that shouldn't stop me from responding to something that's there, where I was quoted, or perhaps about a subject very close to me. I've done this myself, sometimes even leaving the text in the local language, in English (and sometimes even in Portuguese), thus allowing readers to do their own analysis. I did this punctually, and I never felt any kind of problem or discomfort from native speakers of that language. However, the person should not do this systematically and routinely.

On the other hand, this rule does not apply to the global English forum. I think that's the point theymos is referring to, where we can translate a text into English and participate in the general forum discussion.

Another point is that I think it's excessive for people to use AI models for translation. It's better to use tools specifically designed for that purpose, like Google Translate, even if those tools also use AI. I say this for a very simple reason. Just as it doesn't make sense to spread butter on bread with a cleaver, it doesn't make sense to use an AI model to translate text (in most cases used here on the forum).

Ultimately, everything deserves common sense from everyone: from those who write using translations; from those who read; and through moderation.

[Free Market Capitalist]

Haha!

I'm very optimistic about AI on the whole...

This beginning of the post has sent a few forum members to the emergency room with panic attacks. Today is a bad day for the AI Luddites on the forum, and I would like to take this opportunity to express my condolences.

That's normal, theymos is a reasonable guy who allows a certain amount of legitimate AI use on the forum, but he won't allow AI to turn this into a forum of bots interacting with each other.

Hey biggest moron on the forum hurry up, you still haven't red-tagged theymos for allowing certain uses of AI on the forum. We know you like to red tag people for BS like that.

[philipma1957]

AI written text is mostly generic and it lacks human emotions, and even if some AI models try to add emotions to the written text, still that's AI like emotion not human emotion. So, directly copy-pasting from AI should be avoided at any cost and now it's not allowed to do such a thing. Even if done, it should be properly mentioned with in quotation so others might not report it.

No, and not always. For quite some time now, there have been certain "humanizers," whose scripts you can simply indicate to the tool. For example, avoid "universal" phrases like "should be noted," "must be emphasized," "important to point out," and so on. Or use a friendly, informal communication style, ultimately setting a tone that perfectly aligns with human communication. Such scripts have long been circulating in online public groups, and many users are now actively using them. Again, the mega-thread, which regularly rehashes the same thoughts and ideas about buying, selling, holding Bitcoin, and so on, is the very swamp where such scripts are spread.
They evade detection by detectors, they paraphrase the presented text and maintain the topic of discussion, but the fact that these are written not by people, but by AI, is already ineradicable, unfortunately. As long as there are these huge mega-themes where users just need to take a post and paste a script into it, AI will be the "king" of such topics.

I write so many freaking posts and I mostly do it in James Joyce stream of continuous [spelled that wrong]

In the very near future, we will see AI with a communication style similar to philipma1957, or anyone else whose posting style is quite recognizable and individual.

I am sure that with 50,000 posts to study my style of writing can be made into A.i. script

my annoyance is the coder won't ask permission

he or she will download my post's and some James Joyce novels and make the software essentially robbing the two of us.  But that is this world is it not?

[Lucius]

~snip~
That's normal, theymos is a reasonable guy who allows a certain amount of legitimate AI use on the forum, but he won't allow AI to turn this into a forum of bots interacting with each other.

Do you believe that years of doing nothing have already resulted in at least 25% of all posts being completely or partially generated with the help of AI? The forum is full of artificial intelligence be it bots or humans acting like AI bots. I've already written that the only legitimate use of AI on a forum should be if it's used for some kind of data analysis or to generate images for fun, everything else is just another version of plagiarism.

Instead of a ban, AI shitters were given instructions on how to use AI effectively to avoid punishment.


For those who think I'm talking nonsense, look in the next topic - that's just part of the iceberg called AI, a bunch of crap that flows through this forum.

AI Spam Report Reference Thread    

[MarryWithBTC]

can we get maybe new [AI] [/AI] tags that hide blocks like the code tag? automagically formats a short preview of the AI stuff in the tags, and is scrollable inside the message. so the thread doesnt turn into a scroll wheel wrecking endless AI drivel of garbage.

oh! what an excellent suggestion. having an ai tag is definitely going to solve alot of problems.

• it will keep formatting neat
• it will also help the Ai models training in the forum.

This is fine since theymos doesn't want our little corner of humanity being misrepresented or omitted by AIs.

... Also, I'm concerned that putting too much unmarked AI-generated text on the forum might confuse AIs which try to train on forum posts, which is undesirable in my book because it makes AI worse/incomplete and it prevents our little corner of humanity from being reflected within AIs. (That's why we don't block or paywall AIs trying to crawl the forum.)

[philipma1957]

~snip~
That's normal, theymos is a reasonable guy who allows a certain amount of legitimate AI use on the forum, but he won't allow AI to turn this into a forum of bots interacting with each other.

Do you believe that years of doing nothing have already resulted in at least 25% of all posts being completely or partially generated with the help of AI? The forum is full of artificial intelligence be it bots or humans acting like AI bots. I've already written that the only legitimate use of AI on a forum should be if it's used for some kind of data analysis or to generate images for fun, everything else is just another version of plagiarism.

Instead of a ban, AI shitters were given instructions on how to use AI effectively to avoid punishment.


For those who think I'm talking nonsense, look in the next topic - that's just part of the iceberg called AI, a bunch of crap that flows through this forum.

AI Spam Report Reference Thread    

There are issues about how much one can weave AI into a thread.

1) I frankly think any AI used without mentioning it is a wrong use of AI
2) I also think and translating without mentioning it is a wrong use of translating.

But even if those two things are obeyed perfectly the problem is a human can simply say this is AI in his post

this is translate in his post

thus making the content a self admitted mix of human and AI all transparent.

and what is worse a post that is 70% AI and 30% human all transparency included can be a great post.


----------------------------------------------------------------------------------------------------------
All of the below is AI from google

"AI Overview
As of late 2023, approximately 263.5 million people in the United States are aged 18 and older. This adult population represents about 78% of the total U.S. population, which was estimated to be around 336.8 million during that time. The number of adults has grown significantly from 258.3 million in 2020.
Adult Population (18+): ~263.5 million (2023)
Total Population: ~336.8 million (2023)
Demographic Detail: The adult population (18–64) was roughly 202.7 million in early 2025, with an additional 55.9 million aged 65 and over. "

this is google search:

"Mental Health Treatment Among Adults: United States, 2020

Centers for Disease Control and Prevention | CDC (.gov)
https://www.cdc.gov › nchs › products › databriefs
by EP Terlizzi · Cited by 233 — In 2020, 20.3% of U.S. adults received any mental health treatment in the past 12 months, including 16.5% who had taken prescription medication for their mental ...Read more"




this is me:

So as much as 20.3% of Americans are on psych meds

or 263,500,000x0.203=53,490,500

53.49 million people in the USA are on Psych meds and we all know more people are refusing treatment so possibly 60 to 70 million people have problems with their minds in America.

My conclusion is

This is why bitcointalk suffers from poor posting" As much as 25% of the posters suffer from mental illness and maybe only 20% of

the 25% are being treated.

-------------------------------------------------------------------------------------

So look at that post it is mostly AI but all transparency is there.

based on the goole ai and google search it is easy to see

1) why America is fucked
2) why bitcointalk suffers greatly from bad posting

If I made that post into a poll it would get a lot of traffic and it is more ai then not ai

some people would actually think is was not a good post because it uses ai.
yet it is exactly the right way to use ai.


like I said I am willing to help sort ai posts out.

but
1) I have zero issues if you use ai
2) you must say you used ai

like I said this will be fun.

oh I was inspired to do a poll

https://bitcointalk.org/index.php?topic=5574293.msg66400632#msg66400632

[DireWolfM14]

My question is; is the ban going to be temporary or permanent ban like plagiarized posts.

It depends on the person's intent and history. If the person seems to only care about boosting their post-count, then that should typically be a permaban. Otherwise (eg. if it's just a misunderstanding of moderation policy), a warning or temporary ban would be appropriate.

Will this policy be applied retroactively?  Assuming an obvious AI post was written before this verdict was issued by, should it be reported?

The problem is our current AI detectors cannot distinguish between the nature of or motivation behind the prompt, so AI-translated posts often show up as 100% AI generated, even if the thought behind it was human.

I haven't tested it, but I feel like a direct translation should not trigger AI detectors. If it does, we'll just have to use our best judgement. Ultimately, it's about evaluating whether the user is a net-positive or net-negative for the forum (while also keeping fairness in mind).

I would still like to see AI translations identifies as such.  Having a little experience translating my writings to Arabic I've notice much of the authentic nature of my writing style is obliterated when doing so.  Being fluent in the Levant dialect of Arabic (but shitty at writing) it makes easy for me detect the obvious soulless transcriptions.

[philipma1957]

My question is; is the ban going to be temporary or permanent ban like plagiarized posts.

It depends on the person's intent and history. If the person seems to only care about boosting their post-count, then that should typically be a permaban. Otherwise (eg. if it's just a misunderstanding of moderation policy), a warning or temporary ban would be appropriate.

Will this policy be applied retroactively?  Assuming an obvious AI post was written before this verdict was issued by, should it be reported?

The problem is our current AI detectors cannot distinguish between the nature of or motivation behind the prompt, so AI-translated posts often show up as 100% AI generated, even if the thought behind it was human.

I haven't tested it, but I feel like a direct translation should not trigger AI detectors. If it does, we'll just have to use our best judgement. Ultimately, it's about evaluating whether the user is a net-positive or net-negative for the forum (while also keeping fairness in mind).

I would still like to see AI translations identifies as such.  Having a little experience translating my writings to Arabic I've notice much of the authentic nature of my writing style is obliterated when doing so.  Being fluent in the Levant dialect of Arabic (but shitty at writing) it makes easy for me detect the obvious soleless transcriptions.

It will be very hard to police and regulate AI but maybe we can do it.

My opinion is transparency should be the gold standard as AI when use clearly is a good tool.