Shouldn't there be a forum rule against obvious AI-generated content?

[Vod]

umm, I don't think so. The content was originally mine, I just fixed it. 

That's what AI does too.   The content is originally mine, it's just fixing it for me.

[LoyceV]

In programming things will be a little different.

First, as mentioned, a lot of content can be developed based on Open Source code. Some of this code requires credits to be given, others do not.

If the code you use, edit and publish doesn't require to credit the original author, that's fine when it comes to copyright laws. But if you pretend you wrote all of it, it's still plagiarism.

[joker_josue]

In programming things will be a little different.

First, as mentioned, a lot of content can be developed based on Open Source code. Some of this code requires credits to be given, others do not.

If the code you use, edit and publish doesn't require to credit the original author, that's fine when it comes to copyright laws. But if you pretend you wrote all of it, it's still plagiarism.

That is true.

But for a programmer to say that was simply ruining his career, because saying that he did something that in reality he is not capable of doing, is not very positive. It's better to say that you used a certain tool and receive credit for knowing how to use it very well, than in the end losing all credibility.

[Vod]

If the code you use, edit and publish doesn't require to credit the original author, that's fine when it comes to copyright laws. But if you pretend you wrote all of it, it's still plagiarism.

Open source code, which most of these packages are, give you full rights to edit without credit.

[LoyceV]

Open source code, which most of these packages are, give you full rights to edit without credit.

I checked The MIT License, which is used by Bitcoin Core, and indeed, it only requires to include the License, not the credits. I didn't expect that.

[Alone055]

That's what AI does too.   The content is originally mine, it's just fixing it for me.

Not specifically, only if you ask it for that. When you use a prompt to generate something (textual) from an AI model, the idea is yours, not the content.

What Grammarly or other tools do is correct your mistakes, for example, if I write, "He have a good sense of humor.", now, there is a mistake in my sentence, I'm supposed to use "has" with "He" because it's a third person singular and based on English grammar rules, we are supposed to use "has" with third person singular. Grammarly will show me the mistakes in my sentences and help me correct them. In this case, the written content is my own, but it's just helping me have no mistakes in it.

There is absolutely nothing wrong with using AI to rectify your mistakes or help you write your sentences in correct grammar. If a user is writing the content and then using an AI to correct the mistakes, there is nothing wrong with it, but if someone is generating the whole content using an AI model, and then passing it on as their own, that is wrong.

[joker_josue]

Open source code, which most of these packages are, give you full rights to edit without credit.

I checked The MIT License, which is used by Bitcoin Core, and indeed, it only requires to include the License, not the credits. I didn't expect that.

It's true, using open source really allows you to do whatever you want with it, even if it means taking it as it is being sold.

I went to ask ChatGPT if it was considered plagiarism, using AI code in projects, and after a "conversation" he highlighted:
"1. AI-Generated Code is Original - AI-generated code is not a direct copy of other existing code unless it is explicitly asked to replicate something specific.
2. AI Acts as a Tool - Using AI to create code is similar to using other software development tools, such as frameworks, libraries, and code generators.
3. User Control and Supervision - The user has control over the generated code, guiding the process and refining solutions as needed. You are responsible for the logic, structure and final implementation, which means the end result is a co-creation under your direction. The final code is the result of your decisions and adjustments, with the AI ​​serving as a technical assistant."

I think point 3 is the key point in this issue. If the idea, supervision, guidance, analysis, adjustments are made by the person, then the final result belongs to that person.


But this logic can no longer be applied to the written text, without modifications. In the same "conversation", I asked what it would be like in the case of texts and the answer was:
"If you use an AI like ChatGPT to create text and submit it without modifications or without indicating that it was generated by an AI, the concept of plagiarism may apply depending on the context."


In short, using source code via AI is very different from using text.

[GeorgeJohn]

For me AI generated content in forum fits nicely under rule number 33:

Posting plagiarized content is not allowed.

You can actually use AI generated text for something and post it in forum, but you should post it with quotes and add note that this text was generated with xyz AI.
Anything else is going to be considered as zero or low value post and cheating.
If detected post gets reported to mods.
Simple.

You have said it all, that AI rules should be under or considered as plagiarism, And many persons I have come across their cases concerning AI has been treating same way as plagiarised post is been treated..I have not seen any value why someone will put in quote of what it extracted with AI and post in the forum, since the post will look as low value post, i don't think it's needful, the only thing is that anyone that's been caught with AI post should face the consequences by the mods [ban he/her acct] and i think that will limit the rate of using artificial intelligent for forum, if many accts is been brought down, people will desist  using AI.

[The Sceptical Chymist]

Yes, yes, a thousand times YES.

Members using AI to make posts should be permabanned just like they would if they plagiarized something.  And I know there are tools to detect AI usage, but I don't know how to use them and frankly I don't have the patience to scrutinize every post I think was generated or assisted by AI.  OP's example is pretty convincing, but every time I see a Newbie or Jr. Member either creating threads or just making posts in impeccable English, all the alarm bells go off in my head because I know such a phenomenon was almost unheard of prior to idiots gaining access to AI tools.  In other words, most new members' first language isn't English (and it's probably not even in the top five) and hasn't been for years.

The reason I think the punishment should be so harsh is that if this shit isn't nipped in the bud, bitcointalk is going to turn into a forum consisting of bots talking nonsense to each other and will have a flood of threads with moronic topics in which the OP doesn't really say much of anything.

Actually, has anyone checked out the Economics section lately?  Armageddon might have already arrived.

[Alone055]

OP's example is pretty convincing, but every time I see a Newbie or Jr. Member either creating threads or just making posts in impeccable English, all the alarm bells go off in my head because I know such a phenomenon was almost unheard of prior to idiots gaining access to AI tools.  In other words, most new members' first language isn't English (and it's probably not even in the top five) and hasn't been for years.

That's the basic way to detect that a post has been generated through an AI model. You can't expect someone having broken English in their previous posts which they possibly wrote themselves to become a completely perfect English writer in their latter posts which makes you realize that there is something wrong. The AI uses perfect punctuation, sentence starters as I mentioned in the OP, and most importantly, perfect grammar. Even a native speaker might make mistakes sometimes but AI doesn't, and that makes it easy to detect unless someone edits the text after generating it but those who can't write a few sentences themselves would barely have enough brain to do this.

The reason I think the punishment should be so harsh is that if this shit isn't nipped in the bud, bitcointalk is going to turn into a forum consisting of bots talking nonsense to each other and will have a flood of threads with moronic topics in which the OP doesn't really say much of anything.

That's why I suggested there should be an official rule about it, and if there isn't and AI content falls under plagiarism, the culprits should get the same punishment as plagiarizers.

Actually, has anyone checked out the Economics section lately?  Armageddon might have already arrived.

I do, every day, but you know what? Discussion boards are worse than that. Bitcoin Discussion is still better, but Altcoin Discussion is so full of spam and spammers that if you report one post, two more are posted.

[nakamura12]

That's different and shouldn't be punishable just as how you can provide reference with copied content and don't get punished for plagiarism.

It is the same as plagiarism since the one who posted it isn't the one who write it or who made it so it is possible to consider it as plagiarism although the data isn't taken from somewhere else so i'd say it is a partial plagiarism and not plagiarism at the same time. Since it is like that then it should just be considered as spam. Also, people would have to check it to see if it is AI generated or not at all. For punishment, I think restricting an account to post for a certain amount of time like unable to post for 1 day as an example.

[FinneysTrueVision]

I’ve noticed an increase in AI-generated posts. It is not just newbies that are doing it, but also established members who are active in paid signature campaigns. I’ve submitted reports in the AI spam reporting thread but it didn’t seem worth the effort to continue investigating and making reports. Some of the more annoying spammers do get banned but moderators are lenient enough that AI spam is still rampant.

[Alone055]

I’ve noticed an increase in AI-generated posts. It is not just newbies that are doing it, but also established members who are active in paid signature campaigns. I’ve submitted reports in the AI spam reporting thread but it didn’t seem worth the effort to continue investigating and making reports. Some of the more annoying spammers do get banned but moderators are lenient enough that AI spam is still rampant.

I believe the reason why mods aren't banning users for making AI posts is obvious, they don't consider this to be a rule-breaking offense because we don't have an official rule about it yet. So when you report AI-generated posts, the posts get deleted but the users are left without any punishment, that is at least what I have noticed so far and that is the reason why I created this thread.

I have also reported a lot of AI-generated content, most of them were deleted, some unhandled, but I have not seen users getting nuked for it. If a mod comments about it here that would make things more clear. If they consider this plagiarism, there should be a punishment for it, if they don't, there should be a separate rule for it.

[joker_josue]

I believe the reason why mods aren't banning users for making AI posts is obvious, they don't consider this to be a rule-breaking offense because we don't have an official rule about it yet. So when you report AI-generated posts, the posts get deleted but the users are left without any punishment, that is at least what I have noticed so far and that is the reason why I created this thread.

They may not be doing this, not for lack of rules, but for lack of evidence.

I can use Google Translate to write here. Or I can write to an IA and ask him to translate it into perfect English. This is not plagiarism.

Plagiarism, would be you asking "what is Bitcoin?", and I go to IA to write the question and ask him to answer, and then I post the answer here.

The problem now is, which of the two scenarios did I use? Did I write and translate or ask to write?

[Alone055]

They may not be doing this, not for lack of rules, but for lack of evidence.

I can use Google Translate to write here. Or I can write to an IA and ask him to translate it into perfect English. This is not plagiarism.

Plagiarism, would be you asking "what is Bitcoin?", and I go to IA to write the question and ask him to answer, and then I post the answer here.

The problem now is, which of the two scenarios did I use? Did I write and translate or ask to write?

Usually, when you use Google Translate, the text wouldn't be too good to be detected as AI-generated because Google Translate sucks most of the time, however, if you use an AI model, that might happen. But, there will always be a slight difference between a text that has been translated and a text that has been generated. I have explained some patterns that AI models use when generating textual content. Mostly, with translated texts, not all of those things will be present.

Besides, I have mentioned that we are talking about obvious cases, cases where it's obvious that a user is using an AI model to generate the posts they are making in the forum, this could be based on the results of AI detection tools or their post history where they might have not yet started using an AI to generate posts.

There can be exceptions in moderation in cases where there is no convincing evidence about a user being reported for using AI-generated content.

[Cricktor]

I think the reason why there's no specific forum rule to address AI content has to do with fool-proof detection and hard enough evidence. If someone just copy/pastes and plagiarizes it's easy to convict.

I don't have experience with AI content detectors and how well they actually work. (Side question: have those AI content detectors been scientifically well checked?)

What threshold do you want to apply to convict someone of plagarizing with or using AI content? How many decent detectors should flag suspected content as AI diarrhea? Who defines what a decent AI detector is?

My personal opinion is that any AI-generated content should be clearly designated as such, no exception. Omission of this designation should be punishable^[1]. I'm not interested to read any diarrhea of large language models and particularly not when posters make it up as their own stuff.

Computer aided language translation is a different thing and by itself usually no plagiarism.

My proposal for a forum rule would therefore be:
Any AI-generated or large language model content has to be designated as such, no exception allowed. Failure to designate such content will result in temporary bans when convicted. After three increasingly temporary bans the next ban is permanent.


[1] I would be fine with increasingly temporary bans for first few occurances and finally a permaban when the suspect doesn't learn and improve.

[nutildah]

University professors are getting pissed off at the use of ChatGPT for thesis creation:



He brings up some good points on the subject; here are some relevant snippets:

I cannot reject a candidate because I believe their application is AI-generated, as it’s impossible to know for certain whether a text is AI-generated.

It’s true that no algorithm can tell for sure  if a text is AI-generated. However, we are discussing a specific piece of text, not a general one. The text contained dozens red flags, including words typically used by ChatGPT (intricate, delve…) and a suspicious change of font in the last sentence. Additionally, the statement continued in this manner for three pages, displaying red flag after red flag. The strongest indicator was the shape of the apostrophes: standard Calibri font has round apostrophes, GPT-generated text has straight ones.
...
It’s alarming to see how many of these are written by generative AI. Millions of people use ChatGPT to replace their work.
...
Unfortunately, many people use ChatGPT to replace their work. This is bad for them (because they won’t improve) and to society overall, as we risk a regression to the mean.

The message of my post was: don’t replace your unique contributions with average outputs. Be unique.

It is worth differentiating ChatGPT users:

 - those using ChatGPT to enhance their output, vs
 - those using ChatGPT to replace their output.

I'm fine with people using it to fix grammar & spelling errors, but not OK with it when used to generate non-substantive blather (aka spam).

[LoyceV]

I'm fine with people using it to fix grammar & spelling errors

That's like using the software for proofreading, instead of having a human do it. When I was writing my thesis, I asked someone to proofread it and use a red marker an anything that needed my attention. A piece of software may be able to do that for grammar, but it won't be able to do that for anything more serious than that (such as interpreting data incorrectly).

I'd never do this though: I wouldn't want some software company to use my work to train it's software. It's bad enough they already use anything I've ever posted online without my consent.

[mirakal]

That's different and shouldn't be punishable just as how you can provide reference with copied content and don't get punished for plagiarism.

It is the same as plagiarism since the one who posted it isn't the one who write it or who made it so it is possible to consider it as plagiarism although the data isn't taken from somewhere else so i'd say it is a partial plagiarism and not plagiarism at the same time. Since it is like that then it should just be considered as spam. Also, people would have to check it to see if it is AI generated or not at all. For punishment, I think restricting an account to post for a certain amount of time like unable to post for 1 day as an example.

As long as you are posting that’s not originally yours, then it will still fall to plagiarism, irrespective if it’s partial or the whole part. And the forum is very clear that plagiarism seriously is breaking the forum’s rule. Same with this AI generated content, you are not originally the source of your post, so it will still fall to plagiarism, something that should always be avoided if you want to build your own reputation and credibility within the forum.

[Stablexcoin]

OP, some users are well-professional writers or they come from countries with good English and writing. Yes, anyone who plagiarised or used AI to generate responses should be punished. But some make use of AI to correct spelling errors, place punctuation marks, commas, or full atop where necessary, and other assistance in writing is not an offense. I know that ost people use grammar as a tool to help them write good English. Do you consider that to be cheating?

It's annoying if someone writes in terrible English and posts, members will complain, and if the Engish is clean and excellent members still will complain. Then what do we expect from ourselves?