AI-generated post discussion thread: how to identify & report

[Dr.Bitcoin_Strange]

For now, the few ways that have been discovered can be helpful in phishing out some AI-generated posts, but some AI posts may not be noticed because humans, being so logical, can get assisted by the AI to generate them a reply to post on the forum, but after the AI has done its work, they can further modify the post to look more like it was created by a real human.

As Loyce and Suchmoo have mentioned, the AI will just continue to get better, which is my fear too (particular what Loyce said, if it happens like that, then signature and the forum will just be bump with AI contents). If all this AI stuff really gets worse, then having these discussions will be pointless. because it can become so difficult to even detect an AI-generated post.


What if these AIs get modified in a way that they deny what they have already said? for example, someone posted a reply that was created by AI, and you really want to find out if it was AI-generated, so you decide to run an AI word search, and the AI tells you a different thing and denies it really said the first statement? This can be possible if these AIs keep getting better.

[1714697931]

1714697931

[1714697931]

1714697931

[1714697931]

1714697931

[1714697931]

1714697931

[NotATether]

At the end of the day I think we'll lose this battle either way. AI bots will get better, humans will get tired.

We are already losing it. ChatGPT is being integrated into our lives and work environments in the same way that the Internet began to be integrated 25 years ago and cell phones 15 years ago. We can swim against the tide all we want in this forum to stay in the dinosaur era as the world continues to progress thanks to the internet, mobile and AI.

I am all for AI, especially if it is used as leverage, where the human rewrites and adapts, and not just copy-pastes.

I don't think people here are necessarily anti-AI, but as you just said, this is literally a copy-paste problem.

It's like having the internet and the "information superhighway", but your first exposure to all of this is just spam content on the web. Because that's exactly what we are seeing on this forum. Nobody's implementing an "AI-assisted FAQ thread" where you can just ask it a question about Bitcoin and it answers it (thereby eliminating 30% of all posts on this forum), but rather, they are just using it for sophisticated "SEO spamming" or in this case signature spamming.

[cryptoaddictchie]

What if these AIs get modified in a way that they deny what they have already said? for example, someone posted a reply that was created by AI, and you really want to find out if it was AI-generated, so you decide to run an AI word search, and the AI tells you a different thing and denies it really said the first statement? This can be possible if these AIs keep getting better.

Thats a bit hard or if ever it is. Its gonna be a chaos I think. Since even AI can now detect a content written by an AI. Honestly an advance tech could be good but so advance can also not beneficial either. See OP needed to have a standard method for these users using AI which technically cheating the idea of a healthy forum discussion, since instead of creating their own thought and putting here some relying already on AI to finish the job.

[giammangiato]

thanks for these very interesting tools
I think it's easy to notice if one writes with the help of ai or not
if you ask specific questions in my opinion they fail to answer well
a ai cannot respond well to specific sentences

[jokers10]

thanks for these very interesting tools
I think it's easy to notice if one writes with the help of ai or not
if you ask specific questions in my opinion they fail to answer well
a ai cannot respond well to specific sentences

If they answer anyhow you'll hardly guess if it was AI or just a person which have some problems with formulating of their ideas. Those who ignore any questions look like bots or at least impolite ones so we can suppose that there is a bot posting. But when there is any answer it makes everything harder. We have some examples in the topic where I can hardly guess if a post was made right by person or with a ChatGPT help (to be honest, I don't). So I guess it depends mostly on how good are the settings.

[leonair]

thanks for these very interesting tools
I think it's easy to notice if one writes with the help of ai or not
if you ask specific questions in my opinion they fail to answer well
a ai cannot respond well to specific sentences

If they answer anyhow you'll hardly guess if it was AI or just a person which have some problems with formulating of their ideas. Those who ignore any questions look like bots or at least impolite ones so we can suppose that there is a bot posting. But when there is any answer it makes everything harder. We have some examples in the topic where I can hardly guess if a post was made right by person or with a ChatGPT help (to be honest, I don't). So I guess it depends mostly on how good are the settings.

Bot postings, like AI, are always in plain English.  And bot never uses a shortcut English word that humans do. But for English many people use different tools like google translator to identify bot or AI posts, it becomes very difficult. the means shown by the Op might help us find AI posts but it doesn't seem to work Guaranteed. Because in this case the person who posts will be reported against.  If he explains anything, we'll have a hard time proof it a full AI post

[jokers10]

Bot postings, like AI, are always in plain English.  And bot never uses a shortcut English word that humans do. But for English many people use different tools like google translator to identify bot or AI posts, it becomes very difficult. the means shown by the Op might help us find AI posts but it doesn't seem to work Guaranteed. Because in this case the person who posts will be reported against.  If he explains anything, we'll have a hard time proof it a full AI post

It depends on a bot settings. You can ask ChatGPT to use different style or to correct a little by yourself. Bots of previous generations were very obvious and as for new bots... it is much harder to be sure if it is a bot at the other side. Yes, we still can see that there are bots on the forum but just because their owners are too lazy to make good settings or correct the text, IMO.

[FatFork]

thanks for these very interesting tools
I think it's easy to notice if one writes with the help of ai or not

I disagree. A lot of people nowadays are using AI-powered writing assistants like spell and grammar correction tools or AI-powered translation tools to enhance their writing skills. Therefore, it has become quite challenging to tell apart content that's been generated by AI and content that's been aided by AI without taking into account the wider context.

if you ask specific questions in my opinion they fail to answer well
a ai cannot respond well to specific sentences

I don't completely agree with that either. They surely can give acceptable and relevant responses to specific questions. The areas where AI language models may fall behind is understanding the bigger picture of the conversation or interpreting the emotional tone of the text. This is where human intuition and comprehension still play a crucial role.

I think you are confusing between AI generative language models and "simpler" pre-trained chat bots or voice assistants. While both use AI technology, generative language models like ChatGPT are more advanced and capable of producing complex and creative language outputs, whereas pre-trained chatbots typically rely on scripted responses to specific queries or commands.

[dkbit98]

I don't think people here are necessarily anti-AI, but as you just said, this is literally a copy-paste problem.

I am on verge of becoming anti-AI.
In hands of malicious people this can be used as a weapon.

AI crap generation is getting seriously dangerous, now ChatGPT created AI-generated version of Joe Rogan podcast, you can listen entire podcast on youtube, it's around 50 minutes long.
Joe Rogan is obviously concerned where all this AI experiments are going and he tweeted that this is going to get very slippery.
If they can generate this realistic audio podcast AI conversation than they sure can generate AI forum conversation, and I wonder what is next, Skynet, Terminators, etc.  
I really don't know how to identify that something is AI after having reasonable conversation with it, but maybe asking controversial questions can help to identify AI.
Joe Rogan AI Experience:
https://www.youtube.com/watch?v=meu0CoYv3z8

[satscraper]

Forum is  on the edge of collision with outside world.

While Elsevier  has published scientific paper writen in  coauthor with ChatGPT,    forum tends to ban any outcome from this AI.

A possible "out" my be the relevant  reference to source of content published on forum if this content came from  ChatGPT, IMHO

[nutildah]

OK well I'm prepared to concede that it matters more whether a post is spam rather than if it was written by AI.

Just seems like cheating, and for the most part they are low-information posts that don't really add anything substantial to the conversation. At least they are well-written, lol.

Most people using ChatGPT to post on the forum wouldn't be writing anything of quality regardless, pumping out posts that are usually just coherent & on-topic enough to not be deleted as spam.

So basically we're back to where we were before.

Spammers gonna spam, AI or no.

[Poker Player]

Forum is  on the edge of collision with outside world.

The forum is staying in the time of the Australopithecus, going hunting with a club, while the world is on its way to the fifth industrial revolution.

While Elsevier  has published scientific paper writen in  coauthor with ChatGPT,    forum tends to ban any outcome from this AI.

AI is being used to improve the work in various fields right now, and the change it is generating is among the fastest in history. Those who worry that AI will jeopardize their income from signature campaign would do better to worry that AI will make this forum obsolete if they insist on fighting it instead of embracing it.

[nutildah]

Forum is  on the edge of collision with outside world.

The forum is staying in the time of the Australopithecus, going hunting with a club, while the world is on its way to the fifth industrial revolution.

Yes but TBH I kind of like it that way. I don't think the forum should be overhauled -- especially the design. The forum's primary purpose now is to act as a repository / archive for satoshi's posts, along with Hal and a handful of other early contributors / devs. As well as anything that could be regarded as "historically significant" in the timeline of bitcoin or cryptocurrency in general. It really doesn't matter if thick layers of spam get added on top so long as those early, formative posts are still visible an intact. And it would be weird to view satoshi posts in a new forum design that wasn't around when he was here.

[Poker Player]

Yes but TBH I kind of like it that way. I don't think the forum should be overhauled -- especially the design. The forum's primary purpose now is to act as a repository / archive for satoshi's posts, along with Hal and a handful of other early contributors / devs. As well as anything that could be regarded as "historically significant" in the timeline of bitcoin or cryptocurrency in general. It really doesn't matter if thick layers of spam get added on top so long as those early, formative posts are still visible an intact. And it would be weird to view satoshi posts in a new forum design that wasn't around when he was here.

The original old, dusty, hole-punched books can be consulted thousands of years from now thanks to digitization. To preserve something you don't have to keep it in its original format.

Besides I think they are two different things. I would leave the forum format as it is. However, for innovations such as IA, I am clear, if it were up to me all text that is a mixture of IA with human content and results in a quality post, welcome.

[Cryptomultiplier]

For me, I think AI generated text would sound almost too perfect with tenses and wordings to begin with. It would also touch every point as in a summary of what text books chapters or web pages would provide.
No one is beyond mistakes, but most times these human text points can even be made better with the use of text apps like Grammarly to correct errors and appropriate tenses in order. Hence, why it is reviewed on second note.
Unless one is foolish enough to not write on a few points, wherein more details is inputted than necessary, it could be easily detected as AI generated. Thank God for AI text detection tools, else who could really note the difference.

[satscraper]

Thank God for AI text detection tools, else who could really note the difference.

To test level  of scrutiny of AI-text-detection-tool developed by openai I have fed it with ChatGPT-generated- text which was acknowledged by one of forum member.

and got the following responce:




To whom should I trust?

[NotATether]

Thank God for AI text detection tools, else who could really note the difference.

To test level  of scrutiny of AI-text-detection-tool developed by openai I have fed it with ChatGPT-generated- text which was acknowledged by one of forum member.

and got the following responce:

~

To whom should I trust?

ChatGPT can only analyze english generated text at the moment, and it can't detect auto-generated text translated from one language to another (and possibly back again) using something like Google Translate.

Forum is  on the edge of collision with outside world.

The forum is staying in the time of the Australopithecus, going hunting with a club, while the world is on its way to the fifth industrial revolution.

Yes but TBH I kind of like it that way. I don't think the forum should be overhauled -- especially the design. The forum's primary purpose now is to act as a repository / archive for satoshi's posts, along with Hal and a handful of other early contributors / devs. As well as anything that could be regarded as "historically significant" in the timeline of bitcoin or cryptocurrency in general. It really doesn't matter if thick layers of spam get added on top so long as those early, formative posts are still visible an intact. And it would be weird to view satoshi posts in a new forum design that wasn't around when he was here.

Well, look at our website layout. Considering that it looks like a 2005 relic, I don't see how any advancements are going to change life around here (barring absolutely necessary additions like CAPTCHAs and CloudFlare).

[Synchronice]

AI definitely follows some logic, algorithm but at the same time it takes data from internet and some texts, written by humans, might be exactly copied as it was written, but at the same time we have Grammarly with it's Premium plan that features check for plagiarism, makes sentences to look like it's written by someone fluent in English, has ability to change sentences and still keep the main idea.
I think it's a waste of time to focus on whether every random posts are written by AI or not, I don't even think it's bad idea to post an AI generated answer if it truly solves the problem of the user who asked a question.

But I think we can detect users who use AI frequently or abuse it. One scheme, that I often see on this forum is this:
User A creates a thread with titles like: Can bitcoin end poverty?, Bitcoin Vs Gold, Bitcoin Vs Fiat, How to trade safely, Is money root of all evil? Why did you have faith in Bitcoin?
After that, a lot of spammy users come and post tons of shit posts because these subjects are easy to abuse. So, the scheme is: One creates spammy thread, hundreds of spammy users come and spam. It's like, one user feeds other users to earn money.

This user called Susano is a perfect example of what I talk about right now.

So, I think, our priority should be to eliminate these type of users who feed spammers. Because of these users, we also catch spammers in their threads. Usually, these users and their spammy posters are the ones who use an AI.

And to my mind, if you look at user's post history, it's easy to understand whether they use an AI or not. When they use an AI, their posts and answers are general, you can feel that the answer is only answered according to thread title and it had no clue about the OPs post.

[satscraper]

ChatGPT can only analyze english generated text at the moment, and it can't detect auto-generated text translated from one language to another (and possibly back again) using something like Google Translate.

To be precise text-classifier "is likely to get things wrong on text written by children and on text not in English"

but it seems   it is likely to get things wrong on text written also in English, according to

that LoyceV did a test of its own text and those detectors said it was AI text.

[be.open]

For me, I think AI generated text would sound almost too perfect with tenses and wordings to begin with. It would also touch every point as in a summary of what text books chapters or web pages would provide.
No one is beyond mistakes, but most times these human text points can even be made better with the use of text apps like Grammarly to correct errors and appropriate tenses in order. Hence, why it is reviewed on second note.
Unless one is foolish enough to not write on a few points, wherein more details is inputted than necessary, it could be easily detected as AI generated. Thank God for AI text detection tools, else who could really note the difference.

The US Constitution is 92.26% generated with AI/GPT.