Topic: How Can We Identify Someone Who Uses AI? - page 2. (Read 1043 times)

Yes, you're exactly right about this instance. ChatGPT's use of these phrases is exceedingly common, more than a regular human being would ever use.



And when there's one, there's usually more, so I added him to this thread to get his posts deleted.

The technology works, the problem is the funding and the board of the company is as retarded as it can be when it comes to choosing which direction they want the technology to go, they don't seem to know what to do so they end up collapsing from the inside. It's a sad reality that technology that could benefit the humanity will almost always perish before it can be available for public use because capitalists use them first and dismantle it as soon as they see that they're not going to make a generous amount of profit there. Of course they're going to have a hard time with identification, it would be a bad AI if it doesn't cover itself well against detection right? If that was the case, I'd be ecstatic that my project works as I intend it to be.

It's mostly pretty evident. There is a general difference in tone between a text written by a human and a text written by an AI, and that tone stays even if someone tries to tweak their prompts and asks the AI model to write a text that should sound like a human has written it, the AI can't perfectly imitate the tone and the way of writing of a human.

Another notable thing is that an AI model will barely make any mistakes, punctuation, spelling, grammar, and even sentence structure will probably be perfectly accurate which makes it look a bit unreal, though I know there can be people having enough knowledge and experience to be able to do that without using an AI, you can't expect an absolute newbie to have that sort of perfection, right?

Last but not least, an AI language model will have a certain writing pattern. It will separate paragraphs, and start paragraphs with words such as However, Lastly, Additionally, etc.

Here is an example of a post that I feel is written by using an AI model: https://bitcointalksearch.org/topic/i-need-help-to-generate-my-own-key-pair-5495149

Are you trying to say you use an App or something more to detect if a post is AI? Because I myself I can't seem to figure it out or identify an AI post by reading or looking at it... In what way can we find it suspicious before we conclude ? To b honest I don't think is 95% possible by looking at a post without running it to check, this is confusing you know.

firstly, the major problem isn't using AI, but mindless usage of it. For example, chat GPT has made research very easy for people. they get great insights from the near-accurate solutions it proffers. Now, that doesn't mean you'll mindlessly copy whatever you see there to a professional or educational setting. Those are for gathering knowledge and is not intended in any way to be a replacement for your brain or your intentional delivery of concepts.

Secondly, we still have to make use of our intuition and wealth of experiences to make accountable and educating posts and comments in this prestigious forum. Most people who use AI for creating contents to post here are letting this forum down as they are not upholding the integrity of this forum which I know is centered on enlightening the members with well thought, rich, Educating and self-coined contents. or crediting the source of the content if copied.

thirdly, AI is a research tool and never a content writing tool, and is never a replacement for our thinking and creativity. Therefore, mindless usage of AI should be highly discouraged to reduce lack of creativity in its apex and uphold the integrity of this forum.


Finally, there are various tools to detect mindless use of AI eg:

• https://www.scribbr.com/ai-detector/
• https://copyleaks.com/ai-content-detector

Well, yeah, it works It works so well that Open AI itself decided to shut down its service based on their neural network, which determined whether the text was written by an AI or a human. Because they themselves admitted that today there is simply no algorithm for determining whether the text is written by an AI or a human.

You can't detect AI produced text, but newbies who just want to make a few bucks are easily detectable, because they don't utilize it properly. For instance, they don't request to use some different dialect.

Usually, I notice AI posts when they respond with much more content than required. A newbie making four paragraph long posts, is quite suspicious.

These two statements seem to be at odds with each other.

Anyway, what matters first and foremost as far as forum reports are concerned is if a post is spam or not. If it contributes nothing of substance and is suspected to be written by AI, it stands a much higher chance of being deleted than if it is just suspected AI.


Yes but there will still be telltale signs that give them away. It hasn't happened yet so I don't know what they are, but I guarantee there will be.

We can't actually. That's one of the reasons people are afraid of deep learning. The better it is, the less capable we are at distinguishing it. If you write a program, that learns how I'm writing, responding, which are my preferences, interests etc., then you can clone BlackHatCoiner. At that point, only my digital signature will be sufficient for proving I'm indeed me.

But at the moment it isn't so concerning. I'm myself detecting and reporting newbies who use ChatGPT. But we won't distinguish that forever. I suspect that, at some point in the future, there will be users talking, who will not exist.

One thing we can do to spot an AI is to copy the text and ask ChatGPT if it provided the text itself. I've actually done this before and it's working. But sometimes other uses ChatGPT and uses other platforms like Quillbot and other paraphrasing tools to make the text original.

From my observation, text from ChatGPT are like this;

- uses a lot of flowery words
- particular with capitalization and such
- written in formal manner ( strict with punctuations and such)
- extensive and repetitive words


Agree on this, banning accounts who uses ChatGPT is a must because it completely ruins the essence of this forum. But as what you've said, it's really hard to identify who uses AI-generated tools.

There is already a forum rule restricting forum users against the habit of plagiarism, the punishment for breaking this rule is outright ban without warning (could be tem ban or perma ban, all depends on the user involved), the habit of posting contents copied from Ai is no different from a user copy pasting content from other sources, and the user fail to properly reference the source where the content comes from, be it from Ai or an article or someone else's post from any social media, that becomes plagiarism, and the punishment for the offence like I've said before is a ban.

And as for how to know if someone's post is Ai generated, there are many sources online by which this can be known, you can search on google if other users have not already mentioned some of those sources, simply coping and pasting the post on that source will tell you if the post is Ai generated or an original post.

I would love to see this AI ChatGPT ban in bitcointalk forum, but I think it's not always easy to identify AI super-plagiarism.
From my testing of ChatGPT I found out that AI is stupid and it doesn't know what is right or wrong, but it's very good in plagiarism and repeating stuff developers feed it.

I think it is, similar to the way that using a text spinner to rewrite something is considered plagiarism. If someone mentions that they are using AI to write their post, then I have no problems with it. Then its clear that they're acknowledging its not their work.

Another reason why some forums (such as Stack Overflow) have banned the use of AI is they often give wrong answers, which I do see happening from time to time...


Though its not like people can't be wrong without AI. Some make a whole career of it.


Well, not exactly, and I doubt that it ever will... at least not until it starts complaining about not having any rights.

nutildah's post is one of the most coherent ones from this topic and I also consider his topic very useful. Perhaps not many have his skill, to detect a post written by AI simply by reading it, so this must be a special perk

At same time, the highlighted part is the one most frustrating and deceptive. I would certainly not like at all to read some posts, to consider them valuable and, in the end, to find out that the author used an AI to generate them. Regarding the mention of plagiarism -- this is a sort of plagiarism indeed... After all, the author plagiarized an AI (lol?) since he did not write the post by himself... It's like if we were on SF forum and someone would come up with quotes from Jules Verne or Isaac Asimov and he would receive tons of merit. But those merits were supposed to be for the real authors of his lines -- Jules Verne and Asimov. Yet people which did not read those authors or which did not recognize the lines as belonging to them would merit a simple user which would try to take credit on someone else's work.

Now this leads to an interesting debate: can copy-pasting from an AI be considered plagiarism, if the source is not mentioned? Does the AI have the same rights as a human, in order to be considered to be prejudiced by a plagiarist? (This also reminds me of Asimov's novelette The Bicentennial Man (there was made also a film adaptation), the subject being a robot which struggled to be considered human. In the end, he could prove he is a human being only after he made some changes to his body in order to be able to die.)

Most of these people who use AI to generate posts do so because they are too lazy to write what they know and live the rest, thinking that it will make their post look excellent, but that is not the case. They need to understand that the forum does not support that, and if someone will wait and focus on what brought them here, they will seriously learn and be able to write their own post without using the AI. However, some of them are mostly newbies that are yet to know how to post correctly on the forum, so those newbies think that they can use AI to post comments or create new threads, and they believe that using AI will make it look excellent and then deserve merits, but there is a link that they use to check whether a post was created using AI here

I think there is difference we can indentify from a post made my computer and person himself as sometimes we can sense it from the writing style or better there are tools for it.The ChatGPT company also have plagiarism detecting tools developed to indentify such contents so people don't misuse it imposing as it their own content.

Although I personally view the use of GhatGPT, GPT-3, etc. here as plagiarism (someone is taking credit for something they didn't write), at the end of the day AI posts are likely only to be deleted by mods if:

a) they are obviously written by AI, and/or
b) they are spam that adds nothing of value or substance to the forum

with more weight given to b).


In this thread there's a lot of mentioning instances where a particular detector returned a false positive, and some people are keen to use that as an excuse to dismiss all detectors altogether, but some are obviously more accurate than others. What I've noticed is that if a post is short (2-4 lines), the detectors are more likely to return a false positive. The more data the detectors have to analyze, the better job they're going to do (which is why some have a minimum character count).

But ultimately what matters is whether or not the post is useless spam, which is what a lot of AI-assisted posts come across as. I can personally tell if someone is using AI within the first few lines of a post, and then my eyes kind of just gloss over the rest. Don't know about you guys but I simply don't care to read the rest of a post knowing its machine-generated.

I have mixed feelings about the possibility of detecting correctly AI generated text in all cases. I know 2 forum users which were incorrectly accused of using AI for their posts (one case is the one from here). While a (one) website was detecting AI generated posts in their cases, 3 other websites were saying that their posts were written by a human.

AI generated posts are, indeed, a problem, since many can use it for reaching their campaign quotas and also for obtaining not deserved merits. So some action against such cases is more than welcome, such as nutildah did. Besides, some campaign managers are also actively looking for AI generated posts.

However, since this madness technology is somehow recent, the tools for properly detecting it are not flawless and they may lead to false accusations. At same time, a broken tool is better than no tool at all so, at least, there is a way for combating this new way of cheating the system. Therefore, since the tools used are not flawless, I believe that best approach is to treat these accusations case by case, try each time using more websites for detecting the AI and not just one, allow the users to defend themselves and to present proof in their favor and so on.

What I'm trying to say is that, since this technology is here and there is nothing that can be done about that, let's not always jump to conclusions and let's make a thorough analysis for each accusation.

Ultegra123, you understood me very well. I do not actually mean that any newbie that makes grammatically correct posts uses AI. I only stated my own way of dictating an AI generated post. If a newbie who has no merits or fewer than 5 merits, whose post history has been 2 to 3 lines, suddenly starts making lengthy posts like wall of text with higher level of grammatical connectivity or correctness, it's a red flag.

Besides, I do not hunt AI users, but most times I do test random posts to quench my curiosity. I don't have the intention to hunt AI users for now until theymos makes any kind of rule about AI usage.

I am not asking to stop AI, but action is needed against those who are misusing AI and using AI in their posts to show everyone that he is a quality full post data.But if someone by using AI and share something informative info with the acknowledgment he is using Ai, I don't think there is anything wrong here.
Moreover, already many members are skillfully using AI to catch the posters. And by strengthening it I think the abuse of this AI can be stopped in this forum. For your kind information in this topic already described how can you identify those Ai post generators. --> AI-generated post discussion thread: how to identify & report
Now if you can identify those posters with the help of this topic, just report them in to this thread- https://bitcointalksearch.org/topic/report-plagiarism-copypaste-here-mods-please-give-temp-or-permban-as-needed-1926895