Use of AI on Bitcoin talk | Bitcointalksearch.org

dkbit98

legendary

Activity: 2212

Merit: 7064

I think members who are using AI will be easier to recognize in the future, because I am predicting all AI will become worse in giving any correct answers.
Brave browser recently released AI inside their search engine and users now have the option to get answers from this (stupid) AI.
So I decided to test it with random questions and it was giving me wildly wrong answers most of the times.

Few examples:

Quote

Me: How many active members Bitcointalk forum has in 2024?

Brave AI:
According to the provided search results, there are currently 0 members online and 5 guests on the Bitcointalk forum.

Quote

Me: Who is loycev?

Brave AI:
Based on the provided search results, it appears that LoyceV is a user on the BPIP.org platform, which is a media platform for the global development community. Additionally, LoyceV has sent sMerit, which is a digital token of appreciation, to other users on the platform.

https://search.brave.com/

nutildah

legendary

Activity: 3010

Merit: 8114

Quote from: BlackHatCoiner on March 05, 2024, 08:48:36 AM

I think mods are in a quite difficult situation when they evaluate AI reports. First things first, there is no rule which prohibits the use of AI. So, reporting it and using "AI" as excuse, does not satisfy the conditions for deletion. Only if the post is completely meaningless, it'd be removed, but that's because being "completely meaningless" is against the rules. So, if someone used accurate AI responses, I'm not entirely sure what a mod should do.

Pretty sure I'm up to over 100 "good" reports where I put "AI spam" in the report description along with a link to the post that details the detector results in this thread. But yes, simply being AI-generated isn't enough to have it deleted. I find that threads that open with an AI post aren't necessarily deleted, especially if its an advertisement for a new coin, token, or service. A post can contain "accurate responses" and still be spam at the same time.

Quote from: BlackHatCoiner on March 05, 2024, 08:48:36 AM

Quote from: alani123 on March 03, 2024, 04:13:20 AM

Should there be more mobilization of users in that regard? Perhaps start giving neg trust to those abusing AI to make posts... Or we just don't care about it anymore?

If it's totally apparent, then yes. Any account which is essentially run by a chatbot, presented as a human, cannot be trusted.

I've been neutral tagging users for it -- I don't think red tag is appropriate as it doesn't necessarily have anything to do with trade trustworthiness. If others want to red tag offending accounts for it, I'm not gonna try to stop them.

FinneysTrueVision

sr. member

Activity: 1680

Merit: 379

Top Crypto Casino

Quote from: ?? on ??

Why would anyone use ai for a forum lol. Never though about that. However spammers use ai to generate texts but it looks obvious.

Because they can earn money from it. There is a campaign in particular that has a very high maximum number of posts they will pay you for. It is unreasonable to expect anyone to create 50+ quality posts per week so a lot of members in that signature campaign resort to either spamming generic nonsense or using AI to write their posts. There are also some users participating in campaigns with multiple accounts, which would obviously require a lot of effort if they were creating posts organically, so they use AI to cheat more efficiently.

It is almost always obvious when somebody is using AI, but it is time consuming to investigate every case, and with how unreliable AI detecting tools are, there are many cases which go overlooked or unpunished.

jokers10

legendary

Activity: 1974

Merit: 3049

Quote from: Kavelj22 on March 05, 2024, 08:47:55 AM

I was not going to care about the issue until I noticed the interest of Bounty campaign managers to avoid assigning members who use AI to produce publications and in other cases punish those found guilty. But what determines whether or not there is a conviction if there is no accurate way to discover this? In many examples, I am certain that a particular post was made by AI, but I cannot make an official accusation because I do not have enough evidence other than my innate intuition. I really hope no one is wronged because of this.

This is in addition to the great possibility that there will be a similarity between one of these texts and what a human mind might produce. The issue is getting more complicated, in my opinion.

You can use this topic and try to detect with several reliable detectors. If there are several posts detected by Copyleaks, Hive Moderation and Sapling, then any chance that a user didn't use any AI is about zero. If they weren't caught this time, let them make more mistakes. Many when slightly editing AI texts forget that AI takes his words from Internet, so once they will be caught on plagiarism when big enough parts of texts from some outer source will be found in what they post. Those who are too lazy to write by themselves or who have too many accounts for that, will make many mistakes and being caught for them is just a matter of time.

decodx

hero member

Activity: 1456

Merit: 940

🇺🇦 Glory to Ukraine!

Quote from: smelody on March 05, 2024, 01:54:48 PM

Quote from: lovesmayfamilis on March 05, 2024, 08:26:13 AM

Don't keep to yourself behind Kids, Keep walking hand in hand. This job doesn't suit you. Angry

Your wording is confusing. Can you try saying it again in a way that makes sense?

And by the way, lovesmayfamilis is right. You can't just accuse reputed members of using AI-generated content if you're not competent enough to use AI detection tools properly. You gotta learn to tie your shoes before you win a marathon!

smelody

member

Activity: 89

Merit: 8

Quote from: lovesmayfamilis on March 05, 2024, 08:26:13 AM

Don't keep to yourself behind Kids, Keep walking hand in hand. This job doesn't suit you. Angry

JollyGood

legendary

Activity: 2534

Merit: 1713

Top Crypto Casino

There should be a deterrent in the form of an official announcement that would make those using AI on a continuous basis beyond grammar correction face a permanent ban. At the moment the appropriate acceptable course of action to take is to either report the post and/or leave a neutral tag. As this problem will only increase with time (because of signature spammers), an announcement will probably be made to address it.

Quote from: pattrick on March 05, 2024, 07:44:39 AM

If you are using a tool like grammarly and your primary intention is to improve the grammar, it should not be problem. But if you are spamming with AI generated text, this will not be tolerated. You should also report such posts if you come across them. This could have become a major problem for a big forum for bitcointalk but the good thing is that there are many active members who report these kind of posts on time.

BlackHatCoiner

legendary

Activity: 1512

Merit: 7340

Farewell, Leo

Quote from: alani123 on March 03, 2024, 04:13:20 AM

Even in the aforementioned thread there's responses from over a week ago that came out as AI generated text with 98% certainty, and the mods still did not remove it.

I think mods are in a quite difficult situation when they evaluate AI reports. First things first, there is no rule which prohibits the use of AI. So, reporting it and using "AI" as excuse, does not satisfy the conditions for deletion. Only if the post is completely meaningless, it'd be removed, but that's because being "completely meaningless" is against the rules. So, if someone used accurate AI responses, I'm not entirely sure what a mod should do.

Quote from: alani123 on March 03, 2024, 04:13:20 AM

Should there be more mobilization of users in that regard? Perhaps start giving neg trust to those abusing AI to make posts... Or we just don't care about it anymore?

If it's totally apparent, then yes. Any account which is essentially run by a chatbot, presented as a human, cannot be trusted.

Kavelj22

legendary

Activity: 1778

Merit: 1474

🔃EN>>AR Translator🔃

Quote from: AnonBitCoiner on March 02, 2024, 11:39:27 AM

Quote from: Kavelj22 on March 01, 2024, 05:57:50 PM

I do not support the effectiveness of fraud detection applications because they are still inaccurate to use in judging any suspicious content.

Services and online applications that provide AI detection that offer the service for free are barely trustworthy. Sad

Some of them might work sometimes but at times, they are inaccurate as you say. I say this because I have tried most of them with actual AI-generated texts and most of them aren't providing accurate results. The percentage they provide for the text to be AI-generated is very low whereas one can understand that the text is AI-generated just by reading through it.

I was not going to care about the issue until I noticed the interest of Bounty campaign managers to avoid assigning members who use AI to produce publications and in other cases punish those found guilty. But what determines whether or not there is a conviction if there is no accurate way to discover this? In many examples, I am certain that a particular post was made by AI, but I cannot make an official accusation because I do not have enough evidence other than my innate intuition. I really hope no one is wronged because of this.

This is in addition to the great possibility that there will be a similarity between one of these texts and what a human mind might produce. The issue is getting more complicated, in my opinion.

examplens

legendary

Activity: 3472

Merit: 3507

Crypto Swap Exchange

Quote from: lovesmayfamilis on March 05, 2024, 08:26:13 AM

You accused the participant of using an AI tool without reading the rules at all and without delving into the correctness of the assessment you made. And the only reason is that examplens made you a good and fair comment.

Just a few days ago, he sent me a PM "Just for Thanking" with a lot of appreciation and pleasure. Now I'm not sure that exposing my use of AI is an expression of dissatisfaction because I didn't delete the questionable post.
It's amazing how obsessed people are with AI-text tools.

lovesmayfamilis

legendary

Activity: 2072

Merit: 4265

✿♥‿♥✿

Quote from: smelody on March 05, 2024, 06:26:49 AM

If anyone used the news portal copy and paste, In this stages will account banned ?
If account holder realized it's violation of forum rule and remove the related post still be banned the account?

The fact that you have removed evidence of your plagiarism does not mean that the forum will forget it. Don’t forget that archives exist, and if necessary, interested people will always be able to check whether your story was plagiarized or not since you will have a tag as a reminder.
But yes, the moderators gave you a chance to improve your posting, and you were not banned this time. But you should not be vindictive when your mistakes are pointed out to you. Look what you did today!

Quote from: smelody on March 05, 2024, 07:28:58 AM

The user examplens posting AI content and I suspect it.

sapling.ai Fake: 18.9%:
https://sapling.ai/ai-content-detector/e5dc28b08011e3873a9da1980615e942

Post link:
https://bitcointalksearch.org/topic/m.63243288

sapling.ai Fake: 100%:
https://sapling.ai/ai-content-detector/adb96eb4dbc0ca2c746f9271ae3da2ec

Post link:
https://bitcointalksearch.org/topic/m.62281780

You accused the participant of using an AI tool without reading the rules at all and without delving into the correctness of the assessment you made. And the only reason is that examplens made you a good and fair comment.

Quote from: examplens on February 29, 2024, 04:23:11 PM

Quote from: ?? on ??

My countdown to junior member. Anyone review my post and give only 1 Merit(s) please...
Number of Merit(s) left: only 1

Don't beg for merit. It's hard for you to get merit from pity, and it's against the forum's rules.
I looked at your posts and didn't see a single one worth one merit. Also, the feedback you have on the profile, refers to plagiarism, and that is certainly not helpful.
You can do it much better.

Quote from: https://bitcointalk.org/index.php?action=trust;u=3606704

lovesmayfamilis

2024-02-22

Reference

Plagiarism

Your desire to gain merit in any way seems very obsessive and funny. Grin

pattrick

newbie

Activity: 3

Merit: 0

If you are using a tool like grammarly and your primary intention is to improve the grammar, it should not be problem. But if you are spamming with AI generated text, this will not be tolerated. You should also report such posts if you come across them. This could have become a major problem for a big forum for bitcointalk but the good thing is that there are many active members who report these kind of posts on time.

FatFork

legendary

Activity: 1624

Merit: 2594

Top Crypto Casino

Quote from: smelody on March 05, 2024, 06:26:49 AM

If anyone used the news portal copy and paste, In this stages will account banned ?

First offense? Maybe you'll only get a slap on the wrist (post deleted by a moderator). But keep plagiarizing content, and you're outta here. Plagiarism is a no-go according to the forum rules, and the mods decide how harsh the penalty gets.

Quote from: smelody on March 05, 2024, 06:26:49 AM

If account holder realized it's violation of forum rule and remove the related post still be banned the account?

Probably not! Most mods are cool with folks cleaning up their own messes.

smelody

member

Activity: 89

Merit: 8

If anyone used the news portal copy and paste, In this stages will account banned ?
If account holder realized it's violation of forum rule and remove the related post still be banned the account?

jokers10

legendary

Activity: 1974

Merit: 3049

Quote from: alani123 on March 03, 2024, 04:13:20 AM

Even in the aforementioned thread there's responses from over a week ago that came out as AI generated text with 98% certainty, and the mods still did not remove it.

If you show that there are several posts of the same user which are detected by several reliable AI detectors, then there are over 95% probability that moderators will delete these posts. Probability of ban is yet tiny, but it also exists. And having a post with several examples of AI usage in that topic is a good reference for leaving a neutral tag about AI usage, what can be important for those who plan to take part in different campaigns.

It is not as effective as when we report plagiarism, but it gives some not so little result anyway. The more active AI posters are caught, the less motivated others are to use AI.

alani123

legendary

Activity: 2422

Merit: 1451

Leading Crypto Sports Betting & Casino Platform

Quote from: FatFork on March 03, 2024, 04:05:11 AM

In my experience, most reports that are submitted there and follow the guidelines tend to get addressed by the moderators.

That's a good thread and thanks for bringing it to my attention. However I'm very doubtful on what moderators expect to delete a post or thread response based on the AI accusation. Sometimes even for an AI system, there can be bad outputs (as I mentioned it being a trash in trash out system). I've reported obvious word salads that reshuffle words in the OP or from other responses at best adding a couple original words. Even those didn't get taken down.

Even in the aforementioned thread there's responses from over a week ago that came out as AI generated text with 98% certainty, and the mods still did not remove it.

Since mods are acting in this manned, not being strict or outright ignoring AI posts... I am wondering. Should there be more mobilization of users in that regard? Perhaps start giving neg trust to those abusing AI to make posts... Or we just don't care about it anymore?

FatFork

legendary

Activity: 1624

Merit: 2594

Top Crypto Casino

Quote from: alani123 on March 03, 2024, 03:54:26 AM

There should be some framework to address AI posts.

You can use the framework outlined in nutildah's 'AI Spam Report Reference Thread' for reporting AI spam. In my experience, most reports that are submitted there and follow the guidelines tend to get addressed by the moderators.

alani123

legendary

Activity: 2422

Merit: 1451

Leading Crypto Sports Betting & Casino Platform

There should be some framework to address AI posts.

The other day I reported a post reply that was clearly AI generated. For instance, when you put into ChatGPT a piece of text and say to it "write a response in similar words" it will just reshuffle the words and make a response that barely makes any sense, recycling all the input. It's a trash in trash out system after all.

So instead of at least providing some incentive to make such reports mods labeled them as bad... I almost never have my reports labeled as bad and I've made thousands of reports. So if mods treat reports against AI like this, then I think the forum is doomed to be flooded with AI messages, which mods are seemingly giving a greenlight too.

It's obvious even to me that many participants in signature campaigns have started using AI models for their thread responses, but very rarely are they called out about it and even if they are, signature campaign managers that pay them get no consequences...

AnonBitCoiner

sr. member

Activity: 1456

Merit: 326

Eloncoin.org - Mars, here we come!

Quote from: Kavelj22 on March 01, 2024, 05:57:50 PM

I do not support the effectiveness of fraud detection applications because they are still inaccurate to use in judging any suspicious content.

Services and online applications that provide AI detection that offer the service for free are barely trustworthy. Sad

Some of them might work sometimes but at times, they are inaccurate as you say. I say this because I have tried most of them with actual AI-generated texts and most of them aren't providing accurate results. The percentage they provide for the text to be AI-generated is very low whereas one can understand that the text is AI-generated just by reading through it.

Ultegra134

hero member

Activity: 1680

Merit: 845

Quote from: MFahad on March 01, 2024, 10:08:09 PM

I have seen a lot of high-ranked members constantly posting AI-generated posts and their posts are always identical and you can see that they aren't written by a human, what's worst is that they don't even edit them and you can sometimes find sentences and words that give it away that they are using an AI language model to generate those posts.
I wonder how difficult they find it to write posts themselves that they do this, I mean, if you have nothing to contribute, just don't make posts at all, but then again, they need to earn money from here. Signature campaign managers need to up their game identifying and catching AI users.

Well, do you have something to contribute? Because to me, it sounds like you're simply trying to overexaggerate a situation that isn't as bad as you make it look while also increasing your post count for your excessive quota of 60 posts. I imagine it must be hard to reach such a high quota. I personally cannot think of me writing 60 posts per week and being constructive at the same time.

There's a thread to report AI usage, just like lovesmayfamilis told you, and if you're so downright that a large number of high-ranked members are using AI, then you would have known. It's highly possible that this statement is derived from your pure imagination. I've been on the lookout for AI content for quite some time now; the majority is newbies and generally newer accounts who attempt to appear knowledgeable. I've personally found only one Sr. Member and reported him.

So, if you have any evidence to back your statements, why didn't you make a report? A quick look around the forum, and you would have found the corresponding thread.

Topic: Use of AI on Bitcoin talk (Read 1692 times)