It is bad to use AI to help my post quality improve?

[UmerIdrees]

If I wanted to read AI diarrhea, I wouldn't need Bitcointalk. I want to read from humans, that's what forums were created for.
What's next? Only AIs posting and only AIs reading? What's the point?

Lately, I'm getting vibes that someone is going to create a script that can write AI-generated answers under any post from the content available on this forum or from the internet. Honestly, It will be better for posts that were answered hundreds of times before 

That still can be done if you put the post that you want to reply to on ChatGPT and it will generate an AI answer for that, however, it can be very easily detected because there is a lot of difference between a human written post and the AI written post.

The point is, AI detectors aren't that accurate so don't fully rely on it.

This is what the majority of the people think and they start to use AI to write posts for them and think that since AI tools aren't accurate, they won't be caught.
AI tools just prove that the content is AI written, it can be detected with the naked eye too.
Remember You are talking about the tools to detect AI, people can judge whether the content is written by humans or AI, just by reading the post.

[1714961592]

1714961592

[1714961592]

1714961592

[jhonjhon]

The point is, AI detectors aren't that accurate so don't fully rely on it.

And I would add: lest you end up making a fool of yourself.

You're not quite getting the point. I was referring to the post that was published before AI was even invented. It raises a valid question: how can AI detect that it's created by AI when it was made before AI even arrived? It doesn't quite add up, you know? I think posts created on or after the AI era need to be reviewed, as it's obvious that posts from before the AI era had nothing to do with AI. Hope that clarifies things.


Please don't make it "The point is, AI detectors aren't that accurate so don't fully rely on it" as a general statement as I was referring to a specific reply.

[nutildah]

The point is, AI detectors aren't that accurate so don't fully rely on it.

You've oversimplified it. Because of these types of grammar/capitalization errors, I would never assume this was an AI-generated post:

I believe you should take a look at bitcoin. As you can see, many bitcoiners have demonstrated that they have made significant profits from the currency. It all depends on whether you are willing to take the risk. Many investors choose Bitcoin because of its rapid growth, but if you'd rather something more secure and low-risk, consider investing in gold. Here's the idea: To decide which of the two investments you should make, start with a small sum of money. For instance, you could invest the same amount in gold and bitcoin, then watch to see which one is more profitable and able to quickly recoup your initial investment.

Just about every AI detector gives false positives at some point, so its a good idea to check multiple detectors. Some are better than others.

For the text in question, here are the results from all the detectors:

Sapling.ai - 100% fake (occasionally produces false positives)
Hivemoderation (the most reliable detector) - 4.7% likely to contain AI Generated Text
Copyleaks - AI content detected (this one no longer provides a % so I'm not sure it should be used anymore)
Writer.com - 28% human-generated content
Writefull.com - 65% likely this comes from GPT-3, GPT-4 or ChatGPT (not a high enough score, threshold for reporting based on a detector should be 70%)

So yes, based on the results of Hivemoderation and Writer.com, this post could be reported as AI-generated. Except for one thing which you mentioned earlier: the post was written before ChatGPT was released.

The other way to combat the false detection problem is to only submit reports if 3 or more of the author's posts are highly likely to be written by AI based on detector results. If its a one-off thing, it probably shouldn't be reported (unless its clearly spam).

[arabspaceship123]

Remember the old good days when people who wanted to learn a language was supposed to study and practice it, to go along trial and error?

It's changed now. AI's got practical use if you're trying to improve language levels. How's it going to help ppl if they're using it to make posts in forums.

My Spanish and my German are brutal. My English is okay. I tend to type too fast and not correct my mistakes. With google translate and AI I sometime will write in the German section.

But I am very clear that I am doing that.

It's helping your German posts but does adding AI disclaimers cause a defence for avoiding a ban? If it isn't supposed to be used you've got to be careful. If posts get identified for using AI you're risking a ban so even if I needed I wouldn't use it. I can't understand why ppl would use AI in forum posts.

[jhonjhon]

The point is, AI detectors aren't that accurate so don't fully rely on it.

You've oversimplified it. Because of these types of grammar/capitalization errors, I would never assume this was an AI-generated post:

I believe you should take a look at bitcoin. As you can see, many bitcoiners have demonstrated that they have made significant profits from the currency. It all depends on whether you are willing to take the risk. Many investors choose Bitcoin because of its rapid growth, but if you'd rather something more secure and low-risk, consider investing in gold. Here's the idea: To decide which of the two investments you should make, start with a small sum of money. For instance, you could invest the same amount in gold and bitcoin, then watch to see which one is more profitable and able to quickly recoup your initial investment.

Just about every AI detector gives false positives at some point, so its a good idea to check multiple detectors. Some are better than others.

For the text in question, here are the results from all the detectors:

Sapling.ai - 100% fake (occasionally produces false positives)
Hivemoderation (the most reliable detector) - 4.7% likely to contain AI Generated Text
Copyleaks - AI content detected (this one no longer provides a % so I'm not sure it should be used anymore)
Writer.com - 28% human-generated content
Writefull.com - 65% likely this comes from GPT-3, GPT-4 or ChatGPT (not a high enough score, threshold for reporting based on a detector should be 70%)

So yes, based on the results of Hivemoderation and Writer.com, this post could be reported as AI-generated. Except for one thing which you mentioned earlier: the post was written before ChatGPT was released.

The other way to combat the false detection problem is to only submit reports if 3 or more of the author's posts are highly likely to be written by AI based on detector results. If its a one-off thing, it probably shouldn't be reported (unless its clearly spam).

Thanks for this kind of clarification. Now, these are the kinds of replies that I find very objective and something that could benefit the forum generally. I don't like it when other posters treat me like a criminal just because I raised a "sensitive" topic.

They immediately jump to the conclusion that when someone talks about AI, they're already using AI to create posts. I think we should be stricter when it comes to something that involves money, like accusations of scams (thanks to those who are so active), rather than these mere conversations. I understand their concerns, they want to keep the forum from being flooded with AI-generated content or AI Diarrhea as Loycve mentioned. However, that wasn't the point of this thread. Personally, I'm not naive; I know that letting an AI handle posting could lead to a ban of the account. I was just looking for a little help from AI with grammar and minor corrections.

[BabyBandit]

If I wanted to read AI diarrhea, I wouldn't need Bitcointalk. I want to read from humans, that's what forums were created for.
What's next? Only AIs posting and only AIs reading? What's the point?

Nobody need Bitcointalk, but I get your point and you have right forum is to meet other people and communicate with other humans not with a robot.
But I  can almost guarantee you that you don't have anything to worry about, I highly doubt that people using AI will take over the forum, at least as so long we got the base-community here that not using AI and the good people that report users that using AI.
All creed to you guys that doing that, it's actually something I really appreciate and most important, it actually make the forum to a much better place! Thanks a lot for taking your time to do this!  

[Etranger]

I'd like to highlight a particular point here to show that your AI checking mechanism is incorrect. There's no valid argument in this case, and it seems like you haven't investigated thoroughly before making your post.

I have, that is why I proved my point not with 1 of your posts, but 3.

The post you mentioned was created on "July 25, 2022, 02:20:35 PM," and based on my research, ChatGPT (the only AI I'm aware of) was launched on November 30, 2022. So, how could a verifier predict that my post is AI-generated when it was created before the AI was introduced?

ChatGPT-4 definitely was not the first language model. OpenAI released GPT-3 in 2020.

The point is, AI detectors aren't that accurate so don't fully rely on it.

We have a whole topic, which consider those detectors quite accurate, because a lot of AI generated posts were tracked thanks to them.

[arabspaceship123]

Take a look you'll always find polarising topics in the forum. There's users from all over the world who've been posting here for years. They shouldn't be criminalised if they've got a certain point of view or because they've raised a sensitive topic that isn't right.

AI isn't a sensitive topic so it shouldn't be treated that way. It's going to get ppl banned so it shouldn't be used.

Thanks for this kind of clarification. Now, these are the kinds of replies that I find very objective and something that could benefit the forum generally. I don't like it when other posters treat me like a criminal just because I raised a "sensitive" topic.

[Lucius]

The post you mentioned was created on "July 25, 2022, 02:20:35 PM," and based on my research, ChatGPT (the only AI I'm aware of) was launched on November 30, 2022. So, how could a verifier predict that my post is AI-generated when it was created before the AI was introduced?

Before you write something, it would be wise to check the historical facts, because what you claim is simply not true. Each of the posts in question could have been created with the help of AI, because at that time they existed and were publicly available.

GPT-1, the model that was introduced in June 2018, was the first iteration of the GPT (generative pre-trained transformer) series and consisted of 117 million parameters. This set the foundational architecture for ChatGPT as we know it today. GPT-1 demonstrated the power of unsupervised learning in language understanding tasks, using books as training data to predict the next word in a sentence.

GPT-2, which was released in February 2019, represented a significant upgrade with 1.5 billion parameters. It showcased a dramatic improvement in text generation capabilities and produced coherent, multi-paragraph text. But due to its potential misuse, GPT-2 wasn't initially released to the public. The model was eventually launched in November 2019 after OpenAI conducted a staged rollout to study and mitigate potential risks.

GPT-3 was a huge leap forward in June 2020. This model was trained on a staggering 175 billion parameters. Its advanced text-generation capabilities led to widespread use in various applications, from drafting emails and writing articles to creating poetry and even generating programming code. It also demonstrated an ability to answer factual questions and translate between languages

The point is, AI detectors aren't that accurate so don't fully rely on it.

That's true (for now), and while I agree that one should always be careful when someone is accused of using AI for their posts, in some cases it's even more obvious that someone is using just such tools - especially if it's beginners, or when the quality of someone's posts improves dramatically overnight.

[jokers10]

Before you write something, it would be wise to check the historical facts, because what you claim is simply not true. Each of the posts in question could have been created with the help of AI, because at that time they existed and were publicly available.

A guy who shows in this topic that his level in English is not really bad, but who still forces the idea that ChatGPT could be used for language improvement. I suspect that we probably have here the face of a farmer who is upset with many tagged accounts and who is trying to make majority doubt about fighting with AI posting.

AI legitimization is what farmers are very interested in. There are many alt accounts which trashed forum with AI posts and they want to do so further. So they are bringing up this topic again and again.

AI posting should not be tolerated. The total majority of forum users never use it and prefer to communicate with real people and not with hidden bots.

[jhonjhon]

The post you mentioned was created on "July 25, 2022, 02:20:35 PM," and based on my research, ChatGPT (the only AI I'm aware of) was launched on November 30, 2022. So, how could a verifier predict that my post is AI-generated when it was created before the AI was introduced?

Before you write something, it would be wise to check the historical facts, because what you claim is simply not true. Each of the posts in question could have been created with the help of AI, because at that time they existed and were publicly available.

GPT-1, the model that was introduced in June 2018, was the first iteration of the GPT (generative pre-trained transformer) series and consisted of 117 million parameters. This set the foundational architecture for ChatGPT as we know it today. GPT-1 demonstrated the power of unsupervised learning in language understanding tasks, using books as training data to predict the next word in a sentence.

GPT-2, which was released in February 2019, represented a significant upgrade with 1.5 billion parameters. It showcased a dramatic improvement in text generation capabilities and produced coherent, multi-paragraph text. But due to its potential misuse, GPT-2 wasn't initially released to the public. The model was eventually launched in November 2019 after OpenAI conducted a staged rollout to study and mitigate potential risks.

GPT-3 was a huge leap forward in June 2020. This model was trained on a staggering 175 billion parameters. Its advanced text-generation capabilities led to widespread use in various applications, from drafting emails and writing articles to creating poetry and even generating programming code. It also demonstrated an ability to answer factual questions and translate between languages

Well, I wasn't aware that it had already been invented in 2018. My lack of knowledge about this shows that I had not experienced using AI tools at that time. My curiosity arose when I saw Nutilda's thread, which was created just this year, so I'm relatively new to AI when it comes to knowledge.

But wasn't AI back then as sophisticated as it is now? I'm just wondering because this AI thing has only become more popular recently, this year.

This is the thread ;
https://bitcointalk.org/index.php?topic=5448111.msg62060187#msg62060187

A guy who shows in this topic that his level in English is not really bad, but who still forces the idea that ChatGPT could be used for language improvement.

I presume the language improvements you mentioned involve me using an AI to work on my posts, as if I'm allowing AI to generate posts for me. If you were to review all my responses, you would likely understand that I'm primarily interested in grammar correction.

To simplify:

Grammarly is used to correct grammar.
AI is used to correct grammar.

Both serve the same function.  FOR-ME


It seems that the posts of other members have gone off-topic. It is no longer related to the discussion. I should have closed this thread when I announced that I had already found the answer I was looking for.

IMO:

Bad: "Write a response to this question asked on a forum: <OP here>"
Okay: "Fix the spelling mistakes on this post I just wrote: <original post here>

[jokers10]

To simplify:

Grammarly is used to correct grammar.
AI is used to correct grammar.

Both serve the same function.  FOR-ME

To simplify, Grammarly is okay for everyone on the forum, never heard any bad words about those who use it. AI bots are often used to cheat and abuse, many don't like any of their usage on the forum. Why to choose the tool which is not okay for many? It is suspicious.