Topic: I require community feedback (Thanks) (Read 529 times)

22 members participated on this topic excluding me but it seems none was able to conclude the case. Everyone has doubt and not 100% convinced that AI was used. These tools are useless too. The results are not accurate at all. In other two cases that happened recently, there were no confusion - not even from my investigation too so I did not need to ask others opinion. But in this case I was in doubt too. I just did not want to act harshly without investigating and taking enough opinions.

@Bitcoin_people, sorry for any inconvenience, I would like to give the benefit of the doubt in your favor.

Thanks everyone.

PS: Fixed that typo 2023 :-P

Well, it is usually clear enough when you just read what is written with AI usage, detectors can just reaffirm what is suspected. Last times I used https://copyleaks.com/ai-content-detector and it probably not each time correctly detects posts which are likely written with AI, but as for now it never gave a false positive result. But the sample is small, so I'm still not sure if we can believe its positive results each time, even if it didn't make such kind of mistake yet (in my practice).

As for Bitcoin_people, I'd say that AI usually does better, so it is probably not AI written. Of course I can not be sure 100%. Like nutildah said, "bland, uninformative and perfunctory". May be Bitcoin_people should think why some of his posts for some look like AI written. I tend to believe him, but it is not the 100% obvious case. By the way, copyleaks doesn't consider his posts as AI written, but it can give false negative results.

The text tested in random AI content detectors,


1. https://contentatscale.ai/ai-content-detector
"Highly likely to be Human!"

2. https://writer.com/ai-content-detector/
"5% Human-Generated content"

3. https://www.zerogpt.com/
"Your Text is human written"

4. https://copyleaks.com/ai-content-detector
"This is human text"

5. https://gptradar.com/
"Likely Human Generated"

4 out of 5 - Human

Who introduced this checking tool to bitcointalk users? What was s/he thinking to discover this crap

Hold!
This is a bug...
If there is no reliable tool to detect an AI written content, there's actually no way to punish offenders. This is because 99% of the offenders will not admit. So, theymos is correct for keeping mute about this AI trend.

Challenge accepted...


https://sapling.ai/ai-content-detector/4b39093a543b9eedbc80ed7672e754f0




Maybe Royse should remove you for using AI too since what works for Alice must work for Bob 

I'm not saying that you are wrong but Royse is right in creating this thread, this is in aid of how he will manage his campaign, AI abuse, and how to track users who abuse the privilege of being in a campaign, this discussion opens us about the unreality of some AI this is not a concern of bounty managers only but the forum as well.

Nice, this could settle it for me. Visit my profile and run the test, it it returns 5% AI generated results, then sapling.ai should be banned in the forum.


No! Your sense of understanding failed here instead(apologies). The analogy rather is, if it worked for Alice, it should work for Bob. Humans are grossly inconsistent, but programs aren't unless some certain lines of codes are altered.

The offense of using AI is only limited to the signature campaigns rules and not included on the forum's rules, if not anyone found using such to post should be allowed to face the consequences, i believe those particular posts that shows using AI can be rechecked by members who are very good at these and verify on different tools if it's truly AI geerated post or not, if found guilty then the manager should not hesitate in doing the appropriate thing, some participants were also discovered with the use of AI while making their post and the two of them were kicked out of that same campaign, it's an offense that's more worst as plagiarism just that the forum have not taken action to kicked against it, which means it can be permitted outside participating in a signature campaign but not when you're a participant.

Yes, I think so. I am confident that if I use the same tool on your post history, it will return 100% AI generated results on some of your posts to show how inaccurate these AI detector tools are. As I and Lucius have stated on this thread, there are no accurate AI detectors out there that can guarantee up to 60-70% accuracy on whether a post is AI generated or not; We are still in the early stages of this development, and AI created posts tend to leave some phrases behind that indicates that the content was generated by AI. While some smart individuals always remove such phrases, gullible ones always leave them.

It should not be used on Alice just because it worked on Bob- Your sense of judgement is worrisome.

Well this response from you certainly puts a new angle on the subject matter.

I have not determined or concluded whether or not AI was used but well done for you to come here and to stand your ground. If you say you did not use AI and there is no certain way to prove you did or you did not then creating this thread was probably a waste of time when the OP could have sent you a PM to discuss the issue beforehand.

Hello Royse777, this is my feed back.

1. There are some controversies in detecting AI written articles or posts in the forum and in the other hand, victims do not agree they use AI unless proven beyond doubt.

2. But there is a way out. There need to be a standard. Recently you kicked a user out of your campaign @bbigtart. He admitted that he was using AI for his posts. https://bitcointalksearch.org/topic/m.62312182

3. The best thing to do is to use same tool that exposed @bbigtart to measure @bitcoin_people. The tool cannot be correct for @bbigtart and wrong for @Bitcoin_people.

4. I think sapling.ai was used for @bbigtart and that same sapling.ai is used for @bbigtart. The question is, will sapling be so inconsistent such that he will be correct in case one and totally wrong in case two. I don't think so... This is my feedback.

At the moment, there is no tool that could determine with 100% certainty whether someone is using AI or not, and accordingly anyone who makes decisions based on such tools risks accusing someone without any basis of having done something wrong. The only indication that someone is using AI is when they leave a part of the text in which it can be seen that the post was made with AI, and we have seen that there are such cases.

Unfortunately, the plague of AI creeps into every pore of society, and the fight with it on the forum currently looks like a fight with windmills.

Sapling.ai is one of the worst tools available; it scores every random post as 100% AI. Just a few days ago, I was resolving an issue with a campaign manager that removed someone from the campaign without payment due to the exact message you received, and after much investigation, we realized that sapling.aI was classifying nearly all of the posts we examined as 100% AI. At this point, there was nothing he could do; he knew he messed up by acting without adequate investigation. Both the reporter and the campaign manager later apologized, and the user received his payment, but his slot had already been given to someone else. I also asked the reporter who reported the case to quit playing with people's reputations using some shady unproven AI detector, but based on the message you received, he appears to be the same person who sent you the report. People don't learn do they? not him.

Looking at the posts he reported, they had grammatical and phrasing issues, which AI will not generate. There is no AI detector out there that is 60% accurate. AI generated posts are easily to detect merely looking at the common phrases in the post.

These posts are not AI generated .......



Guys Royse is moving us back to sand of time.

There was a report on the usage of AI by one participant in the Stake.com campaign

https://bitcointalksearch.org/topic/m.62156867

The user is still in the campaign and one of the Stake's management addressed this concern


We caught one because he admitted to using AIs'


Everyone has a concern that they are going to be flagged for using AI although they are not using AI, we need a reliable tool to detect AI, I support Nutilda's finding with so many AI in the market some of it will show false positives.

C'mon man, i think its too early to start saying such things about the user as its just an accusation yet and still haven't been confirmed yet as to if truly his posts are AI generated and also it might interest you to follow up the comments above as most community members are still not sure if actually the post are Proven AI written and secondly this AI technology is still at its early stages and a lot of flaws have been found in them so counting on most of them to  prove if someone's post is AI generated can be still be reconsidered.

Believe me I never use AI. I always try to follow the signature campaign rules and post my own way. Monitor your posts carefully with AI detector. I always write by my own hand.
And I respect my campaign manager a lot.
Basically I am very active and post for several weeks to get bonus but you can trust me I never post AI generated. I follow @Royse777 manager's campaign rules all the time and also know that AI generated posts in his campaign lead to account bans. Why would I knowingly harm myself? Please check my posts again and judge.

I ran these posts through my 3 favorite AI detectors and I do not think the results are conclusive enough to assume they were written by AI.

1.

Hivemoderation.com: 0% AI
Writer.com: 95% AI
Contentscale.ai: 0% AI

2.

Hivemoderation.com: 0% AI
Writer.com: 0% AI
Contentscale.ai: 12% AI

3.

Hivemoderation.com: 6% AI
Writer.com: 85% AI
Contentscale.ai: 0% AI

And let's choose another one at random:


Hivemoderation.com: 0% AI
Writer.com: 73% AI
Contentscale.ai: 27% AI


So I mean the posts aren't good. They are bland, uninformative and perfunctory -- par for the shitposter course. But I don't think they are necessarily AI-generated. If they were, the poster changed enough of the wording to make them not appear that they were and undetectable to most (some?) of the detectors.

Not strong enough evidence to assume these are AI-generated posts IMO.

I don't think asking a majority opinions will work, you're the campaign manager so you're the one who can make a decision. Majority will say he was using AI, but your heart say the user wasn't using AI, this only lead a conflict of interest.

If you still in doubt to take an action, it's better to leave it and watch closer on the next of his post, sooner or later you could find the answer.

No one will admit he was using AI, that's really stupid, similar like you expecting a scammer will admit if he is a scammer.

Perhaps, it would be possible to reduce AI domination by allowing users state in quote that the text was generated by an AI tech.
I think if one doesn't have a good question to ask, the AI won't be able to provide a reasonable answer. So, it is only logical to assume that the person who used AI tech knew what he wanted but didn't know how to articulate it.
This is still early stages of AI development and anyone who readily uses it to create a thread is just plain lazy to think and should be cautioned sternly.

By pure chance, I know about a similar recent case, where someone was accused for same thing and the accuser used same website - sapling.ai.

However, checking, first mentioned post of Bitcoin_people with another AI detector (https://copyleaks.com/ai-content-detector) it seems the text has 73.8% chances to be written by a human:



Checking second post with another AI detector leads to a 71.8% chance for the text to be written by a human:



Last days a forum user was excluded from a campaign after an accuse of using AI for his posts because the reporter checked the text with sapling.ai. That user was also denied his last campaign payment. I checked the text with 2 other AI detectors and both said his reported posts were written by a human. The user sent this proof to the manager, the manager accepted to send him the last payment but, however, the user remained excluded from the campaign.

Seeing that different websites give totally different results makes me wonder about the accuracy of some of them. Perhaps, excepting the website results we should also take into consideration the user's reputation, in such cases?

---

Perhaps you want to update the highlighted part?

Why is that some people are not serious in a work that pays them, what happens to his hand to type letters with his own idea's,  tell what should warrant people to use IA, from what I noticed is like this people have different account and they use IA to make sure they have meet up with the number's of post they needed weekly, if I'm the person in control of the sector it promote I will take my decision by removing the person without notice since it doesn't want to take its engagement serious

Well not being able to participate til January 2023 isn't a big deal since that was 6 months ago lol. Your campaign your rules man, your job isn't to please the community with whatever standards you think they want to see, just run the campaign however you feel it need ran. If people don't like it, fuck em they don't have to join. They can try to get a project or 2 and manage in the way they feel.

I have passed a detector and I see that I get some posts as humanly generated at 100% and others as AI generated at 90% or so. The problem is that I do not see much difference between one and the other, as they are generic posts in general, most of them in Gambling or Bitcoin Discussion.

The problem with AI is that it will be increasingly difficult to distinguish what is written by AI from what is written by humans, we are still in early stages.

 

I checked a couple of his post on two platforms and they all passed with flying colours. Posting the screenshot of the first post that seems to be written by an AI. Detecting AI content is not that easy. Sometime you can easily detect them by reading the content and sometime you need to use detection platform. Those two I used are considered good.

Another thing as per my understanding if some is using AI then they would use it to create topics and not for replying to a topic. The user in question is replying in this instance.

Using AI is not yet moderated by the staff. Could it be considered plagiarism if they post those content genereated by AI?  Or who knows he is as good as how AI thinks thats why its more than 90%. But 100%?

It is not good to posts them though since basically, its not him who created the context. I guess a wiggle room for it will work like if the AI Detector only sees less than 50%, then he must have contributed.

I also tried to copy some of my latest posts on the website that anonymous reporter has shared (thanks for sharing anyway), and most of them are less than 1% fake (some are even 0%), but there are 2 of my posts that has more than 30%. I also don't know if this AI detector is reliable or not as well.

Anyway, I don't know if it's only me, but I despise users who are using ChatGPT for their advantage here in the forum just to get a few bucks in their signature campaign. IMO, it's obvious that the OP has used AI apps to maximize the quality of his posts. I just don't know why people are using it here? To gain merits? Self-validation? To show here that you are a good poster? I mean it doesn't make sense (for me).

We've seen previous report from a user who has been caught using AI tools in his posts, but what he only got is a neutral trust, and a ban from campaigns that Royse777 is managing, and will be managing in the future. Deleting his AI posts will be useless as well because he has been exposed already, and I guess it's almost certain that he will be out of Yomix anytime soon (if proven he's using AI).

P.S. Well, I guess that the site that the reporter has posted isn't as reliable as it is because @pakhitheboss tried it on another AI detector, and it seems that it says there that it wasn't an AI written post.
I tried using the AI detector website that @nutildah has shared here: LINK
LINK LINK LINK LINK
On the first post that is suspected to be AI written, 2 of the sites said that it is created by a human, while the 2 said that it has made thru AI.
On the second post that is suspected to be AI written, 4 of the same websites that I used said that it is created by human.
On the third post that is suspected to be AI written, 2 of the sites again said that it is created by a human while the 2 said that it has made thru AI.

TBH, it's hard to detect which posts are created by human, and which are created by AI knowing that these AI detector tools have different outputs for the same posts. The only time that we will know that the suspected user is using AI is when he, himself says it. If the accused is using or not, I still despise using AI to earn here.

Accidentally you exposed the reporter. You removed his name, but it forwards to his Bitcointalk profile when someone clicks on the quote link. You may want to remove his UID as well.

About this matter. This is a shitty thing people do. I guess he did it to reach the maximum post limit to get weekly from the campaign. I have seen he received the bonus multiple times. It's not easy for someone to write over 90+ posts each week. Maybe he decided to get help from AI. I suggest he delete all his AI posts before moderators punish him.

Edit: I've Checked some of his posts as well. It doesn't seem AI-Generated. Some AI Detector tools giving false information, I guess?

I don't think you need community feedback about this matter. You already mentioned this in your campaign thread;

I think if you can find direct evidence of using chat GPT or any other AI application I agree people should be banned from the campaign. Obviously many people are in this for the Bitcoin/sats, but honestly I don't count myself among them as I don't go out of my way to make posts for money. I look at a lot of people in campaigns and it seems all they do all day long is shitpost to get as much BTC as possible. If I started doing that for example it would be pretty obvious. Just saying how I personally feel of course.

As you can see the reporter don't want his profile to be in public so respecting that I am posting it on behalf of him. I have done some of my investigations but it seems fighting against AI is not as easy as it was for plagiarism.

@Bitcoin_people, if the result comes negative from majority opinions then please accept my apology in advance but if you found guilty then you will be banned participating from any of my campaigns until 2nd January 2023 2024.

Thanks to you all in advance.

I would like you know your feedback.