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

[stompix]

Testing a few of those detectors on a text I just stumbled upon

https://writer.com/ai-content-detector/   - 100% HUMAN-GENERATED CONTENT Fantastic!
https://gptzero.me/                                  Your text is likely to be written entirely by a human
https://contentatscale.ai/ai-content-detector/  96%  Highly likely to be Human!
https://x.writefull.com/gpt-detector  https://x.writefull.com/gpt-detector
https://hivemoderation.com/ai-generated-content-detection The input is: not likely to contain AI Generated Text
https://paraphrasingtool.ai/ai-content-detector/   Text may contain AI-written sentences   
https://detector.dng.ai/                               Based on our Analysis your test has been most likely written by a Human.

ChatGBT:



 

See? I interact and exchange ideas based on what is in my mind

This is not interaction you just copy-pasted stuff.

This is helpful and unique content

It's not unique content because anyone trying to do the same would come with the same result.
Unique content is generated by unique persons and ways of thinking, in this case, you're giving a standard response everyone could come up with, just like looking at a Wikipedia article and copy-pasting that paragraph.

[1714498891]

1714498891

[1714498891]

1714498891

[1714498891]

1714498891

[aseev]

This is not interaction you just copy-pasted stuff.
It's not unique content because anyone trying to do the same would come with the same result.
Unique content is generated by unique persons and ways of thinking, in this case, you're giving a standard response everyone could come up with, just like looking at a Wikipedia article and copy-pasting that paragraph.

It seems like you replied without giving much thought to my comments. If you use AI as a tool to supplement and speed up your process to create something helpful in response to others or use AI to communicate in your native language - it's an interaction. If you copy-paste AI responses then - yes, there is not much value to it.

I'm not talking about using answers directly from AI and pasting them here. And by the way, it is not the standard response for everyone. The response will depend on the prompt you supply to AI. The more knowledgeable you are and the more information you put in the prompt, the more interesting and unique response you will get.

Sometimes talking to a human can be worse than talking to the bots, and not the AI-powered ones, just those primitive spinner bots, not capable of going a little beyond the script and just repeating what the other guy said just a few posts before. People give too much credit to their creativity and uniqueness.

[nutildah]

It is a tool. Blaming a tool entirely for a problem is not fair. How you use it determines the outcome.

Nobody is blaming the tool, we're blaming the users who use the tool to generate spam and get paid for it.

But you can easily change that by asking to write the responses, for example, in Ernest Hemingway's literary style. Try it.

Yes but spammers aren't going to do that because, again, they are lazy and simply looking for time-saving measures to mass-generate spam.

Testing a few of those detectors on a text I just stumbled upon
...
https://hivemoderation.com/ai-generated-content-detection The input is: not likely to contain AI Generated Text
...
ChatGBT:



 

Can I see what your example was here? I've yet to come across a text mis-identified by the Hive Moderation tool.

[stompix]

Can I see what your example was here? I've yet to come across a text mis-identified by the Hive Moderation tool.

One thing to make sure first, the user didn't post it as his own creation, he just made the topic linking to this article

Top 10 Cryptocurrency Secrets Unleashed
Be Prepared for the Price Swings
Every investment has its own set of features that keeps them afloat., Similarly, the cryptocurrency market is filled with volatility. When you are buying a cryptocurrency, it indirectly means that you are signing up for the ups and downs it leverages. Sometimes, short-term steep falls are rise might even shock the investors. However, these are very common in the cryptocurrency ecosystem and you might actually yield profits from this. If you keep a close track of the growth and follow experts who are accurate about predictions, then you can invest in certain digital tokens and get your hand on good profits. 

Keep a Constant Check on Its Performance
Even if you are not a full-time cryptocurrency investor and is doing it on the sidelines for profit, you need to keep a constant tab on the price swings. Although a 24/7 observation is not needed, checking them at constant intervals is a good thing. The more complicated your investing strategy becomes, the more you should review it. While this might seem to be a thing for short-term investors, long-run investors can also follow these criteria to keep a healthy investment. 

Do Your Own Research and Manage the Risk Factors
A global fact is that cryptocurrency trading is a high-risk business and more traders lose than not. Therefore, don’t get tempted to add more value to your investment portfolio once you see a profit. Most importantly, don’t take other investors’ advise on what cryptocurrencies to invest in. Every digital token has its day! Bitcoin was at an all-time high just last November and now it is down like crazy. Therefore, do your own research before investing.
 
Opt for the Long-Term Investment
Fear of Not Missing Out (FOMO) and panic selling is very common things in the cryptocurrency market. Currently, people are trying to get rid of their Bitcoin investments before they could fall further and eat up their potential money. Similarly, six months back, investors went crazy over the Shiba Inu rally and many beginners tried their hand on the memecoin. Although these are the factors by which the cryptocurrency market functions, opting for the long-term investment plan is the best way to yield profit. 

Choose the Right Platform to Trade
Even if you have picked a cryptocurrency in mind to trade, choosing the right platform also matters. While picking the platform, make sure it abides by all the regulations of the country you are living in. Other factors like exchange liquidity, asset liquidity, and fees need to be clarified before investment. Explore the other features in the platform while you are trading in it. 

Use Trading Bots
Trading bots are becoming a common thing in the cryptocurrency market. They are automated software tools investors use to buy and sell financial instruments at a preconfigured time or when predefined conditions are met to maximize profits. Generally, trading bots identify the market trends and suggest investors invest in cryptocurrencies that give increased profits and reduced loses and risks.
 
Cryptocurrencies Can’t be Completely Wiped Out
Although countries might try to bring regulations on cryptocurrencies and their trading, a complete ban is impossible because anybody can own a wallet. Even if the country bans digital tokens, people can still use foreign accounts to trade them. Tech-savvy investors are also in the top of following such tactics. However, major countries won’t even consider banning cryptocurrencies as they know how much people have invested in them. 

Digital Tokens can Come Under Taxation
When Bitcoin made its debut in 2009, the whole concept of cryptocurrency was new to even government agencies. But over the decade, central authorities have become more aware of digital tokens and the profit investors yield. Therefore, many countries are coming up with effective taxation regulations that could come into effect in the coming months or years. 

Keep the Asset Safe at the Right Place
Currently, there are two types of storage in the market. One is hot storage and the other is cold storage. While hot storage refers to an online digital wallet, cold is an offline wallet that is typically stored on a hard drive. According to experts, keeping your cryptocurrency in a cold wallet is the safest way to protect it from hacks and mishaps.
 
Try Out Technical Analysis
If you are a pro in investing, then try out technical analysis tools to choose the right cryptocurrency. Technical analysis involves using mathematical indicators and chart patterns to try and predict which way the process will move next.

In this case the author most likely added a few phrases himself, probably that's why the spelling errors in it and the detectors fail, but it's AI generated for sure.

[nutildah]

In this case the author most likely added a few phrases himself, probably that's why the spelling errors in it and the detectors fail, but it's AI generated for sure.

On the contrary, its completely human written for sure. It was written before ChatGPT was launched:



The detectors all returned that it was human written because it was indeed written by a human. If anything its an example of the detectors doing their job accurately.

[stompix]

On the contrary, its completely human written for sure. It was written before ChatGPT was launched:
The detectors all returned that it was human written because it was indeed written by a human. If anything its an example of the detectors doing their job accurately.

So, does this mean that we have a situation where chatGPT was caught both plagiarizing and lying about its claim?  
I've tried asking it for 10 tips and indeed, the reply had the same 10 "secrets" in almost the same order and the same content, but..this is really getting weird, the AI is generating content that was previously written by a human, not so much intelligence as paraphrasing.

Sometimes talking to a human can be worse than talking to the bots,

You have no idea how much worse it can get but keep on..
 

[aseev]

ChatGPT is not a paraphrasing tool, it's a language processing model. You asked it to give you something, it gave you what you asked. If you ask what something means, it may pull a definition from Encyclopedia Britannica. There is nothing wrong with that. You may even want to request a definition from a specific source.

If you need to paraphrase something, you need to specifically ask it to do so. You caught nothing.

Edit: If your intent is to obfuscate the source, right after it generates the list, type "Rewrite in conversational style as if it was written as a forum post and change the tip's order" and see what happens.

[Synchronice]

It's strange but no one has mentioned ZeroGPT. I have checked a lot of texts on different AI text detectors and this one is probably the best out there.

I'm trying to understand what's the logic behind ChatGPTs writing capabilities. It may sound funny but I came up with one idea. Probably all of you know popular anime called Naruto and Naruto Shippuden. I asked ChatGPT to use its own imagination and and write a story of Sasuke Uchiha vs Nagato Uzumaki, this is the result I got:

Sasuke Uchiha, a former member of the Konohagakure's Uchiha clan, had set his sights on the elusive Nagato, the leader of the criminal organization, Akatsuki. Sasuke had a score to settle with Nagato after learning of his involvement in the death of his brother, Itachi.
The two warriors clashed in a remote location, surrounded by towering trees and craggy rocks. Sasuke's Sharingan eyes glowed red as he launched a barrage of lightning-infused attacks at Nagato, who evaded them effortlessly.
Nagato, a master of the Rinnegan, unleashed a flurry of jutsu at Sasuke, who countered with his own techniques. The battlefield was soon littered with craters and debris, as the two combatants exchanged blows with brutal force.
Sasuke attempted to land a decisive blow, but Nagato's Rinnegan allowed him to absorb and repel Sasuke's attack. Undeterred, Sasuke continued to fight with all his might, determined to avenge his brother's death.
As the battle raged on, Sasuke gradually gained the upper hand. His speed and agility allowed him to dodge Nagato's attacks and launch swift counterattacks. Nagato, sensing that he was losing ground, unleashed his most powerful jutsu, the Six Paths of Pain.
Six grotesque creatures appeared, each with its own set of unique abilities. Sasuke was momentarily taken aback by their appearance, but quickly regained his composure and engaged them in combat. The fight was fierce, with Sasuke narrowly avoiding the attacks of the six creatures while dealing them deadly blows.
Finally, after a long and grueling battle, Sasuke emerged victorious. Nagato, wounded and exhausted, lay on the ground, defeated. Sasuke stood over him, his Sharingan eyes blazing with anger and triumph.
"You were no match for me," Sasuke said, his voice dripping with contempt. "Now tell me everything you know about my brother's death."
Nagato, his face twisted in pain, whispered something inaudible. Sasuke leaned in closer, eager to hear his words. But before Nagato could say anything else, he closed his eyes and breathed his last.
Sasuke stood there for a moment, gazing at Nagato's lifeless body. He had won the battle, but his heart was heavy with the knowledge that his brother's death had been part of a larger conspiracy.

With a heavy heart, Sasuke turned and began the long journey back to Konohagakure, determined to uncover the truth about his brother's death and put an end to the shadowy forces that had conspired to bring it about.

While it's not the best story, still sounds interesting from a fresh AI.

[nutildah]

So, does this mean that we have a situation where chatGPT was caught both plagiarizing and lying about its claim?  

I don't think its lying so much as it just happens to be wrong a good portion of the time. For example, I caught this the other day -- I'm personally certain this text was AI generated although cannot "prove" that it was:

Many cryptocurrencies, such as Bitcoin, are based on decentralized blockchain technology, which means that transactions can still be processed and recorded even if the internet is not available. While it's true that accessing cryptocurrency wallets and conducting transactions would be more difficult without internet access, it's possible that alternative methods of transmitting data could be developed in times of war.

 

I know that is just anecdotal but Stack Overflow has actually banned the use of ChatGPT on its site because it was generating too many wrong answers:

Overall, because the average rate of getting correct answers from ChatGPT is too low, the posting of answers created by ChatGPT is substantially harmful to the site and to users who are asking and looking for correct answers.

The primary problem is that while the answers which ChatGPT produces have a high rate of being incorrect, they typically look like they might be good and the answers are very easy to produce. There are also many people trying out ChatGPT to create answers, without the expertise or willingness to verify that the answer is correct prior to posting. Because such answers are so easy to produce, a large number of people are posting a lot of answers. The volume of these answers (thousands) and the fact that the answers often require a detailed read by someone with at least some subject matter expertise in order to determine that the answer is actually bad has effectively swamped our volunteer-based quality curation infrastructure.

I think the same rationale applies to this forum and if they're doing it, we should also pursue a set of criteria to establish what constitutes an "AI-generated post."

[aseev]

I think the same rationale applies to this forum and if they're doing it, we should also pursue a set of criteria to establish what constitutes an "AI-generated post."

The forum soon will be overrun with low quality AI generated posts, just because there is a financial gain in it, it is easy to do and the tools are available to everyone.

The first step is to include in the forum policy that AI generated content is not allowed. Most people spamming the forum won't care about the quality of the output, won't spend much time crafting their prompts, so it will be easy to spot and ban the offenders. As others mentioned, if someone doesn't know what they are talking about - AI will not help them hide that forever.

More complicated cases will have to be left to the mods' discretion (human gut feeling). Because the technology is changing rapidly, today's criteria may not be applicable in a couple of months. Using the technology can be one of the reasons for the ban, but not the main criteria. The criteria could be usefulness of the content, whether there are other policy violations, like plagiarism, and whether the account is created solely for posting generated content for self-promotion or capital gain.

Tools will be helpful with only the obvious cases, where they are not really required.

[Baofeng]

Another AI generated post, here. ^archived

But the thing is that somewhat is it being hide under the pretense of a Poll. 

[Flexystar]

In this case the author most likely added a few phrases himself, probably that's why the spelling errors in it and the detectors fail, but it's AI generated for sure.

On the contrary, its completely human written for sure. It was written before ChatGPT was launched:



The detectors all returned that it was human written because it was indeed written by a human. If anything its an example of the detectors doing their job accurately.

You guys discussing my thread and I wasn’t invited to the party? 

Interesting to see this discussion. I actually noticed it after reading @stompix comment on my thread.

I am not sure what’s the war about but is this any blame game regarding AI generated content? I mean if I’m making a post in the forum while sipping my morning coffee is it now mandatory to check the article written by someone or published by other news sites?

I mean this is ridiculous. I hope I haven’t broken any rules here (funny).

I think it’s advisable that Whatever is being referred from outside forum MUST have references and links to avoid getting direct allegations.

@stompix and @nutildah, thank you guys for running the checks and mentioning it wasn’t my original creation. 

***
After seeing the whole thread, I’m pretty sure AI has already taken over the “minds” of posters.

I just hope that our forum doesn’t become place to discuss what’s being shitting around the world. Lolz.

[lovesmayfamilis]

I will add one more resource, familiar to everyone who used it to search for plagiarism. Recently, he added the ability to determine whether the text is written by AI or humans.
In addition, there is a browser extension that, after highlighting the text, determines whether AI was used or not.
I checked the posts of some users, and although the extension is still in the beta version, it defines itself quite well.

The extension works in the Chrome browser!

https://copyleaks.com/ai-content-detector

https://chrome.google.com/webstore/detail/ai-content-detector-copyl/gplcmncpklkdjiccbknjjkoidpgkcakd?utm_source=website&utm_medium=landingPage&utm_campaign=installExtension

[aseev]

Can I see what your example was here? I've yet to come across a text mis-identified by the Hive Moderation tool.

Sorry, I missed your question.

I used the tool on my last reply (https://bitcointalk.org/index.php?topic=5448111.msg62129192#msg62129192) that contains generated content and this is the result (had to paste the text twice though because of the limit):



I also tried to generate longer text in "normal" English with prompts designed to produce output that is not identifiable by these detection tools (without putting too much work, no more than a spammer would) and it catches most of them pretty well. The lowest percentage I got was 60%. This is one of the better tools. Interested to see how it performs when GPT 5 comes out.

[nutildah]

I also tried to generate longer text in "normal" English with prompts designed to produce output that is not identifiable by these detection tools (without putting too much work, no more than a spammer would) and it catches most of them pretty well. The lowest percentage I got was 60%. This is one of the better tools. Interested to see how it performs when GPT 5 comes out.

Well, again, most spammers using this tool are too lazy to create a prompt to make the output look different from the usual AI blather. That's why they're using AI: they are lazy and uncreative.

At the end of the day what really matters whether the post is considered "spam" or not. If its not adding anything to the conversation then its most likely spam and should be deleted either way.

[nutildah]

It occurred to me that ChatGPT often uses the same phrases when generating answers, and I was wondering how to ascertain which phrases are verifiably part of ChatGPT's response formula. Then it occurred to me, why not ask ChatGPT itself? So here it is, straight from the horses mouth:

Prompt:

Can you give me a list of phrases commonly included in ChatGPT responses? For example, I noticed that "It's worth noting," "In essence," "In summary," "Finally," "Overall," and "On the other hand" are all commonly found in ChatGPT-generated responses.

Response:

Certainly! While it's important to note that the language generated by ChatGPT can vary, here is a list of phrases that are commonly found in ChatGPT-generated responses:

1. It's worth noting
2. In essence
3. In summary
4. Finally
5. Overall
6. On the other hand
7. However
8. Moreover
9. Furthermore
10. Additionally
11. In addition
12. In conclusion
13. That being said
14. In light of
15. Nevertheless
16. In other words
17. As a result
18. Therefore
19. Consequently
20. To illustrate
21. For instance
22. In fact
23. As mentioned earlier
24. It is important to emphasize
25. Notably
26. Interestingly
27. Conversely
28. In contrast
29. Similarly
30. In a similar vein
31. On a related note
32. It is worth mentioning
33. Given these points
34. To summarize
35. In brief

Remember that the use of these phrases may vary depending on the context and style of the generated response.

I asked it to regenerate the response and it came back with a very similar list, although sometimes it gets tired and only lists 20 phrases or so. It is funny how the first 6 are ones I gave it as examples.

Basically I believe that posters that frequently use a combination of these phrases, along with

- writing posts in perfect English (especially if they previously did not)
- sometimes writing long posts in rapid succession of one another, and
- writing posts that come back as "likely" or "very likely" to contain AI-written material from 3 or more AI text detectors

exhibit a strong likelihood of using AI to write their posts for them.

Feel free to ask it the same prompt to verify the correctness of these results.

Why continue to harp on this issue, you ask? I'll demonstrate the answer shortly.

[joker_josue]

It occurred to me that ChatGPT often uses the same phrases when generating answers, and I was wondering how to ascertain which phrases are verifiably part of ChatGPT's response formula. Then it occurred to me, why not ask ChatGPT itself? So here it is, straight from the horses mouth:

This approach was very interesting, it will help to more easily try to understand what is written in AI and what is not.
Of course, it will never be 100% viable, but these phrases/words combined with other known elements will certainly help you to make better decisions.

[hilariousandco]

I think the same rationale applies to this forum and if they're doing it, we should also pursue a set of criteria to establish what constitutes an "AI-generated post."

The forum soon will be overrun with low quality AI generated posts, just because there is a financial gain in it, it is easy to do and the tools are available to everyone.

The first step is to include in the forum policy that AI generated content is not allowed. Most people spamming the forum won't care about the quality of the output, won't spend much time crafting their prompts, so it will be easy to spot and ban the offenders. As others mentioned, if someone doesn't know what they are talking about - AI will not help them hide that forever.

It should be treated the same as plagiarism. Asking an AI something then copy and pasting the response is just as bad if not worse than plagiarism because it's not as easy to detect, though the posts do kind of stand out. I just noticed this user after one of their posts was reported for bumping a thread: https://bitcointalk.org/index.php?action=profile;u=335943

They immediately look AI written to me:

Identifying the best coin and the right time to invest can be a challenging task. It requires research, analysis, and a bit of intuition. One way to identify the best coin is to look at its market cap, trading volume and historical performance. These metrics can give you an idea of how popular and valuable the coin is. Additionally, you can look at the technology behind the coin, its use cases and the team behind it. As for the right time to invest, it is important to keep an eye on market trends and news. Look for any major announcements or events that could impact the price of the coin. It's also important to have a long term investment strategy and not get caught up in short term price fluctuations. Investing in crypto currency can be risky so its important to do your due diligence and invest only what you can afford to lose.

To increase capital investment for a business try to attract investors or partners who can contribute funds into the business. This can be done through networking, advertising or pitching to potential investors. Improve profitability of the business by increasing sales, reducing costs or increasing efficiency. This can be achieve through marketing initiatives, cost cutting measures or process improvements. Business can explore financing options such as loans, lines of credit or crowd funding. These options can provide immediate funds to support business growth. Business can consider expanding their product or service offerings, entering new markets or diversifying their revenue streams. This can help to increase revenue and attract new investors.

And you can see the stark difference from the user's first posts:

facebook is still alive. Me myself i can't live without opening my account every single day. Facebook is a source of communication specially to family, friends who are in distant places. Facebook will remain for a long time unless the owner will give-up.

i would rather to be rich to buy all the necessity in the world and share it to the people who are in need.

I think theymos should probably make an announcement about prohibiting this sort of use of AI or at the very least we add it to the "unofficial" forum rules but users should be alerted not to use it in such a way. If a user if just going to use it to copy and paste the title of a thread or the OP into an AI then copy and paste the results that should be treated the same as plagiarism. If it isn't then this place isn't going to be a discussion forum for much longer but just a get paid to copy and post AI responses forum.

[aseev]

If it isn't then this place isn't going to be a discussion forum for much longer but just a get paid to copy and post AI responses forum.

The person you used as an example has fully embraced the technology - almost all of their recent posts are AI-generated. Soon, almost everyone will use language models or at least try them in their online communications. This is just to support my other comment about the future ease of access to large language models and why they will be very different from the common spam tools. These LLM apps will be everywhere, in browsers, email clients, virtual assistants, etc, available to every forum user, regardless of their level of technical skill.

[Findingnemo]

.
.
.

I think theymos should probably make an announcement about prohibiting this sort of use of AI or at the very least we add it to the "unofficial" forum rules but users should be alerted not to use it in such a way. If a user if just going to use it to copy and paste the title of a thread or the OP into an AI then copy and paste the results that should be treated the same as plagiarism. If it isn't then this place isn't going to be a discussion forum for much longer but just a get paid to copy and post AI responses forum.

Personally I tried many AI detection tools and none of them able to find the contents written by myself but enhanced via AI tools so the big problem is hiw we are going to find those trickster who actually play smart by not simply copy pasting the response generated by AI but they alter something and can make it look genuine.

That's why Theymos is waiting for some kind of tool to be invented to differentiate with 100% accuracy of whether the content written by human or AI.