AI Spam Report Reference Thread

[execijutiere]

I’m not saying my posts are perfect, but based on what you’ve developed, it’s clear they’re not bad. After all, PostHealthScore shows the result as “good,” doesn’t it?

AI posters frequently have "good" results. This is a problem for which there is no solution. That is why I made this caveat:

The system was not developed to measure how good a poster's posts are, but rather how not good they are. So long as you are in that middle 50% or higher, you are fine.* Unless you are using AI to write your posts for you, in which case all bets are off.
...
* "most likely" fine; I don't want to offer definite certainty... just don't plagiarize or cheat while writing your posts.

Reading through your posts I get the feeling that you are mainly getting your material from elsewhere; whether its from AI or plagiarization of another source, I'm not entirely sure, but the end result is the same.

Additionally, I find your posts in large fail to satisfy any of these criteria:

Here's another honest word of advice. Before hitting the "Post" button, ask yourself:

- is this something I would read all the way through in my spare time?
- am I saying anything compelling, original or insightful?
- when I make these kinds of posts, does anyone seem to read them or care?

If the answer is "no" to 2 or more of these questions, then that's a good sign you probably need to improve your post before hitting the button.

It’s so terrible when a person can’t prove themselves.
I don’t use AI. You can easily see this just by translating the sentences into Turkish and re-entering them into the program, but you choose not to see it.
I also don’t understand where or based on what you concluded there was plagiarism.
In short, I don’t use AI and my post percentage is good, and I have my own sigban. The moderators aren’t responding. It’s just you and me now

[AakZaki]

After all, PostHealthScore shows the result as “good,” doesn’t it?
You're contradicting yourself right now. You say the app you developed is “good,” but then you say it isn't.

As far as I know, the PostHealthScore tool is not for detecting AI-generated posts. Rather, it is for providing a thorough assessment of all posts created. Those are two different things bro.

Regardless of whether your post is created using AI or not, the tool will still judge according to its quality.
So I advise you not to mention that tool, to prove that you are not using AI.

[Sterlino]

After all, PostHealthScore shows the result as “good,” doesn’t it?
You're contradicting yourself right now. You say the app you developed is “good,” but then you say it isn't.

As far as I know, the PostHealthScore tool is not for detecting AI-generated posts. Rather, it is for providing a thorough assessment of all posts created. Those are two different things bro.

Regardless of whether your post is created using AI or not, the tool will still judge according to its quality.
So I advise you not to mention that tool, to prove that you are not using AI.

Yes, PostHealthScore is not for AI detecting. Basically, sigban is given for not making quality full posts, for not making contributed posts in the forum. Just copying and pasting from AI does not make it a contributed post. That is why he was given sigban. And nutildah clearly said that sigban is not given only based on PostHealthScore. The sigban is given by manually checking.

I needed to tack on one more summary of issues to this thread before once again locking it.

First, in the time since this thread was opened, I started receiving payment for sigban work. It feels weird not to mention it, but also weird to open a new thread just to say it, so I'm mentioning it here.

Here are some general points that need addressing or re-addressing:

• Every sigban is issued only after a thorough manual review. There is no way to automatically trigger a sigban.
  
• Sigbans are never issued for a single weakness or potential weakness. They are the result of an account's consistent, perpetual underperformance in terms of general forum expectations.
• The appeals process will remain as is (granted at my discretion) and accounts seeking to appeal should open their own appeal topic in Meta. This way each case can have a more focused discussion.
• Its impossible for anyone to have intimate familiarity with every board, which is why community input is crucial to the development of this process. If you value the contributions of a sigbanned user, don't be shy to let your opinions be known in their appeal thread. I do appreciate it if you can do so respectfully.
• The case of arallmuus should demonstrate that I am willing to be reasonable. I made an incorrect assessment about the value of this user's posts. The humility in the defense of this user expressed by a few independent voices was what helped me realize this.
• That said, community pressure will only work in cases where I don't strongly believe a user produces more noise than signal. If I think the same conclusion could be made by reasonable people, the sigban will remain.
• Sigbans are reversible by an admin or global moderator, who are monitoring the progress of this new process. If you truly believe you have been wronged, they can overturn the sigban.

Finally, if you have an unresolved appeal in this thread, please create a new appeal thread in Meta.

Thank you.

PostHealthScore is a tool, where the score is shown depending on some criteria in particular. But it is definitely useful. those who post low quality posts, post in repeatable threads, and repeatedly post short posts just to increase the post count are detected and considered as low quality posters. But having a high score does not mean that he is a quality full poster, it is not proven. So the decision to give sigban by manually checking is definitely logical and that is what should be done.

user: 50 Cent


His score has come to Elite Poster (Top 10% of active posters).
But he was given a sigban because his post was maximum copy pasting, where he only provided a link to the source to be safe from plagiarism. Clearly proven he is not quality poster  though he has good score in tools.

[jokers10]

You can easily see this just by translating the sentences into Turkish and re-entering them into the program, but you choose not to see it.

How can it prove anything? It just shows that some program changes the text the way it is not detected. Some humanizers do the same. It doesn't prove that the original text wasn't written by AI. If you know the automated translators which don't make translation via AI (totally rewriting the original text), then you should use them or translate by your own. When you use the translator which rewrites a text with AI, it becomes indistinguishable from AI to detectors. And it is really hard to try to prove the opposite.

[execijutiere]

After all, PostHealthScore shows the result as “good,” doesn’t it?
You're contradicting yourself right now. You say the app you developed is “good,” but then you say it isn't.

As far as I know, the PostHealthScore tool is not for detecting AI-generated posts. Rather, it is for providing a thorough assessment of all posts created. Those are two different things bro.

Regardless of whether your post is created using AI or not, the tool will still judge according to its quality.
So I advise you not to mention that tool, to prove that you are not using AI.

I actually showed this as an example because I got a sig ban even though I wasn't using AI. After all, I can prove it's not AI by translating it into Turkish.

You can easily see this just by translating the sentences into Turkish and re-entering them into the program, but you choose not to see it.

How can it prove anything? It just shows that some program changes the text the way it is not detected. Some humanizers do the same. It doesn't prove that the original text wasn't written by AI. If you know the automated translators which don't make translation via AI (totally rewriting the original text), then you should use them or translate by your own. When you use the translator which rewrites a text with AI, it becomes indistinguishable from AI to detectors. And it is really hard to try to prove the opposite.

When I translate the text into Turkish and paste it into various programs, every program flags it as 100% human-written. The fact that I use an English translation—even if it’s AI-powered—shouldn’t make me look guilty. If I’d known this, I wouldn’t have used it in the first place, but theymos had mentioned something like this before—saying you can use AI for translation—but I just can’t see how it improves the quality that much.

Beşiktaş didn’t play poorly, but managed to win on the road. We can say they put on a performance against Hradec Králové that at least guaranteed they would advance to the next round. After Kassoum Ouattara was sent off, Beşiktaş played the final 20 minutes with 10 men but at least avoided returning home with a loss. Semih Kılıçsoy came off the bench and scored just 3–4 minutes later, giving his team the lead.Beşiktaş is not going to lose this home game anyway, and the fact that they managed to come away with a win from this away match—even while down to 10 men—is an added advantage for them.

Sapling: 83.2% AI
Pangram: 90% AI

I ran the message again—the one where Nutildah accused me of using AI and, when I checked it, it showed an 83% AI score.
It’s showing 7% now
How reliable are AI detectors?
This is just one example, of course; there are others in different posts as well.

[jokers10]

When I translate the text into Turkish and paste it into various programs, every program flags it as 100% human-written. The fact that I use an English translation—even if it’s AI-powered—shouldn’t make me look guilty. If I’d known this, I wouldn’t have used it in the first place, but theymos had mentioned something like this before—saying you can use AI for translation—but I just can’t see how it improves the quality that much.

I made a translation of that text for my local section, so I remember it good enough and it doesn't allow the AI translation, which "improves" your text. So is it possible that it is a result of some automated translator usage? I'm not sure, but okay, probably. But it is still not what was allowed.

- If you have the AI do almost a direct translation of something you wrote, then that's OK. For example, you can tell the AI to "Directly translate this text into Spanish: <something you wrote>" or "Output the following text exactly as-is, except with any clear spelling or grammar mistakes fixed: <something you wrote>". It's not OK to tell an AI, "Improve this text: <something you wrote>", since then it will mostly rewrite it, and it risks becoming AI slop in the process.

If the result is indistinguishable from AI to detectors, the impact of AI rewriting was too significant.

[execijutiere]

When I translate the text into Turkish and paste it into various programs, every program flags it as 100% human-written. The fact that I use an English translation—even if it’s AI-powered—shouldn’t make me look guilty. If I’d known this, I wouldn’t have used it in the first place, but theymos had mentioned something like this before—saying you can use AI for translation—but I just can’t see how it improves the quality that much.

I made a translation of that text for my local section, so I remember it good enough and it doesn't allow the AI translation, which "improves" your text. So is it possible that it is a result of some automated translator usage? I'm not sure, but okay, probably. But it is still not what was allowed.

- If you have the AI do almost a direct translation of something you wrote, then that's OK. For example, you can tell the AI to "Directly translate this text into Spanish: <something you wrote>" or "Output the following text exactly as-is, except with any clear spelling or grammar mistakes fixed: <something you wrote>". It's not OK to tell an AI, "Improve this text: <something you wrote>", since then it will mostly rewrite it, and it risks becoming AI slop in the process.

If the result is indistinguishable from AI to detectors, the impact of AI rewriting was too significant.

I was using the DeepL programme. I’m still using it. I’m not even aware that it improves the text as it translates, and perhaps that’s why they call it AI.
Moving on to the ‘Local section’, there are loads of people using the local translation feature – are all of them going to get sigbanned?

[jokers10]

I was using the DeepL programme. I’m still using it. I’m not even aware that it improves the text as it translates, and perhaps that’s why they call it AI.
Moving on to the ‘Local section’, there are loads of people using the local translation feature – are all of them going to get sigbanned?

Try to warm them, inspire them to check if the tool they use for the translation is using AI during its work. Better not to do so. It is not about any automated translators, but about some.

But when you say that you didn't know that your tool uses AI and anyway use an argument of admin's words from some old topic, which have a slightly different meaning, when you say what happens if to use some detectors the way you like and also try to prove that detectors are not reliable themselves... I'm not sure that all of the users you are talking about have the same problems as you.

[lovesmayfamilis]

Cryptobanco

From the payments side, a few reasons this rarely happens:

- There's nothing to reject. An on-chain transfer lands in the exchange's address before anyone looks at it. The screening (KYT) runs after the fact.
- Sending it back is a decision too. If the funds really are tied to a sanctioned or stolen source, returning them means the exchange just processed a payment to that source. Compliance teams would rather freeze and ask than take that risk.
- "Return to sender" isn't always clear. If the deposit came from another exchange or a shared wallet, the sending address often isn't the person who owns the funds.

What helps on the sender side is what BitMaxz said: check the address risk score before you send, and don't keep reusing an address that has picked up a bad history. Most business payment tools now run this check automatically before a payout leaves.

Copyleaks: 100%
GPTZero: 100%
pangram.: 100%

It's annoying, but from the network's side it's mostly about who they're paying. Once a network sends hundreds of payouts a week to publishers in different countries, its own bank or payment provider starts asking: who are these recipients, is anyone sanctioned, where does the money end up. Without KYC on publishers they risk losing the payment rails altogether.

That's also why the "address only" networks usually pay in BTC directly and keep minimums low, while the ones paying by bank or USDC tend to ask for ID. Worth deciding up front which payout method you actually want, because that decides how much KYC you'll face.

GPTZero: 100%
pangram.: 100%

[Yorubasrm]

How many checkers should be used to check whether a post is written by AI?

After checking with how many checkers can it be confirmed that a post is written by AI?

If I want to check whether a post is written by AI, which checkers should I use?

[[btc]]

How many checkers should be used to check whether a post is written by AI?

After checking with how many checkers can it be confirmed that a post is written by AI?

If I want to check whether a post is written by AI, which checkers should I use?

Hi, you can read the OP of this thread (look at the page #'s on the bottom and go to page one, at the top) to learn more, but it basically says to use at least 2. I personally like to use the whole list, and then some, if I really care enough about wanting to learn if text is AI or not from someone on this forum.

Here are the criteria that will be used to determine whether or not a post was either wholly or partially made using AI (as originally referenced in this thread):

• At least 2 of the 7 AI content detectors mentioned below must give a result of the post having a strong likelihood of containing AI-written material (with "strong likelihood" defined for each one):
  
  1. GPTZero - 70% probability to contain AI generated text or higher
  2. Copyleaks AI Content Detector - 70% probability for AI or higher
  3. Sapling AI Detector - 70% Fake or higher
  4. Quillbot - 70% of text is likely to come from AI or higher
  5. Zerogpt - 70% likely to include parts generated by AI/GPT or higher
  6. StealthWriter - 70% AI probability or higher
  7. Originality.ai - 70% confident that text is AI generated or higher
  
  (Pay-to-use AI detectors have been excluded to keep verification accessible to all)

.

• There must be at least 3 posts by the account that meet the requirements specified in A. If its just a one-off or even two-off, the poster may deserve some leniency, but if they make a habit of it, their posts should be reported as "AI spam."

[memehunter]

How many checkers should be used to check whether a post is written by AI?

After checking with how many checkers can it be confirmed that a post is written by AI?

If I want to check whether a post is written by AI, which checkers should I use?

You should have read the OP. Here are all the details you want: https://bitcointalk.org/index.php?topic=5456516
I also suggest you to read some reports here and the discussion followed. It will give you more context and enable you to be more efficient in detecting AI spam.

[LoyceV]

What's the verdict on Newbie bill_colton, who's first topic was on the Tech board?
What are the most common mistakes made when Setting Up an offline wallet.

There are a number of things to take into consideration when Setting Up a secure offline wallet.

Key management, seed backups, signing, and verifying addresses.

For example, if you have an air gapped signing machine you will need to send unsigned and signed transactions to your online watch only wallet. This can cause issues when verifying addresses if you don’t understand the process.

What do experienced users do?

What security measures would you implement to ensure the wallet is secure and you can store valuable funds in it?

Do you favour using a hardware wallet, air gapped computer, multi-signature or any combination of these?

Update: it was moved away from the Tech board.

[Shishir99]

What's the verdict on Newbie bill_colton, who's first topic was on the Tech board?
What are the most common mistakes made when Setting Up an offline wallet.

Copyleaks 0% AI
Originality shows 96%, original.
Pangram shows 100%, possibly human.

I also ran it through a plagiarism checker, but couldn't find any match.

[[btc]]

What's the verdict

Without using any detectors/quick feeling, I'm leaning towards no, because grammar error(s?) and it looks like he browsed the forum a bit to find areas he's interested in to post before doing so.

• "Setting Up" uses a capital S and a capital U. Should be lowercase.
• "favour" instead of "favor" (Ex: Color vs. Colour), LLM's usually, unless prompted otherwise, default to American English spelling, no?
• Post was made 4 days after registration. Slop usually posts very close to the time of registration.

His profile's local forum time shows his matches up with UK time, though - so who knows.

[Yorubasrm]

Hi, you can read the OP of this thread (look at the page #'s on the bottom and go to page one, at the top) to learn more, but it basically says to use at least 2. I personally like to use the whole list, and then some, if I really care enough about wanting to learn if text is AI or not from someone on this forum.

You asked to read the points in the first post and it was said to confirm based on the majority.

But if we want to use tools other than the ones listed here to check for AI, there are other AI checkers besides these few that give very good results.

Why are those not added here?

[execijutiere]

I was using the DeepL programme. I’m still using it. I’m not even aware that it improves the text as it translates, and perhaps that’s why they call it AI.
Moving on to the ‘Local section’, there are loads of people using the local translation feature – are all of them going to get sigbanned?

Try to warm them, inspire them to check if the tool they use for the translation is using AI during its work. Better not to do so. It is not about any automated translators, but about some.

But when you say that you didn't know that your tool uses AI and anyway use an argument of admin's words from some old topic, which have a slightly different meaning, when you say what happens if to use some detectors the way you like and also try to prove that detectors are not reliable themselves... I'm not sure that all of the users you are talking about have the same problems as you.

I don’t understand what you mean, mate.
If I’d known the translation app I was using made corrections like this, would I have let it label me as AI?
I’m talking about how the detectors sometimes flag certain sentences as AI and then stop doing so. If it was AI before, what’s changed now? It’s the same sentence, after all—so why the change?

[macson]

How many checkers should be used to check whether a post is written by AI?

After checking with how many checkers can it be confirmed that a post is written by AI?

If I want to check whether a post is written by AI, which checkers should I use?

I don't know why you're so lazy to refer to page 1. What you're asking has already been explained so clearly. Also, these days a lot of newbies deliberately use AI checker tools to make sure their AI-generated text isn't detected by those tools.

[Yorubasrm]

I don't know why you're so lazy to refer to page 1. What you're asking has already been explained so clearly.

Now I understand this matter. The post was made because I didn't understand the matter and later, after few people answered my question beautifully, I saw this and was able to learn the details of the topics mentioned in the first page's first post.

Also, these days a lot of newbies deliberately use AI checker tools to make sure their AI-generated text isn't detected by those tools.

Isn't it better to check and verify a post after writing it, before publishing? I don't know much about these things, but it's better to check the post after writing it so there's no chance of mistakes.

[macson]

Now I understand this matter. The post was made because I didn't understand the matter and later, after few people answered my question beautifully, I saw this and was able to learn the details of the topics mentioned in the first page's first post.

Are you an LLM model? What LLM model are you?

Isn't it better to check and verify a post after writing it, before publishing? I don't know much about these things, but it's better to check the post after writing it so there's no chance of mistakes.

Yes, it's better to check it first and then paraphrase if you are an LLM model. That way you can avoid the AI checker. But you can't fool a real human reading your writing. 

Besides, if it's a real user and not an LLM, then checking with an AI checker is just a waste of time because why bother. If you are confident with your writing, then there shouldn't be any problem.