The new AI Rule.

[KingsDen]

I have seen a post on the announcement sub board by admin about AI and its punishment. It is a welcome development, if here turns to bots discussing topics, majority of us will walk away because I do not want to discuss with organized codes, rather with humans with deep sense of reasoning and humour.

Then, there is a challenge even in the other forum. Users who use AI to generate text often deny when caught and there's no 100% effective tools to detect AI. There are tools but I cannot vouch for them. They can be accurate with one post and total crap with another post. Based on this, to what percentage will a users post show that it is AI generated before they will receive the AI badge?

[dragononcrypto]

I assume we'd go for something similar to the plagiarism rule of 80%+. But also don't really understand how accurate these AI detectors are, or how easy/difficult it'd be to reach say 50% without using AI at all. There's always been the guide to "rewrite [...] in your own words" or "rephrase the content" in order to avoid plagiarism, so I also assume this would apply to AI content as well.

[Zed0X]

Anything above 50%. I think the question should be directed more to what AI detection tools should be used instead of the percentage of AI texts posted. The free software are usually limited but the paid versions should be more accurate.

[KingsDen]

But also don't really understand how accurate these AI detectors are

Some are grossly inaccurate and misleading, while some give fair results. But Zed0x has said that paid versions do better job. But how many people will be willing to be subscribing for AI detector just to catch AI users?. Unless the forum will invest in this, which is actually not necessary.

Anything above 50%. I think the question should be directed more to what AI detection tools should be used instead of the percentage of AI texts posted. The free software are usually limited but the paid versions should be more accurate.

Yea, if a random test is performed and a certain tool proves to be the most reliable, the forum can adopt it as standard for AI detection.

[Fivestar4everMVP]

Anything above 50%. I think the question should be directed more to what AI detection tools should be used instead of the percentage of AI texts posted. The free software are usually limited but the paid versions should be more accurate.

You said exactly what i was thinking and was going to say, the question should really be which Ai detection tool that would be effective and well trusted enough to use, that will be efficient and reliable in being able to accurately detect Ai written posts.

A long time ago when the issue of Ai generated posts became a common problem on that other forum, which made a lot of high ranking users to begin paying attention, someone shared a tool which was really effective at that time, but unfortunately, I can't remember the link to the site anymore.

So, instead of us discussing at what percentage level a post should be considered Ai generated, and the poster tagged, I think for now, we should focus more on finding a great tool that works effectively and accurately in detecting Ai generated posts.

[hugeblack]

Using 3 to 5 tools and a percentage of more than 80% may be enough to confirm that a post is written by AI, and the lack of connection of ideas and the change in language between each reply is evidence that the post was done by AI.


In general, it should be taken case by case after agreeing on a certain number of detection tools.

[examplens]

Any use of AI for further discussion should not be acceptable. We are here to hear the opinions of others, and I assume that each of us knows how to use tools like ChatGPT and can talk to the bot himself if he wants to. Those who use AI do so only because their posts look "smarter" and possibly get a place in the signature campaign.

Some translation or grammar tools can make the post seem like AI wrote it, but I believe no tool will give it more than 20% of AI-generated text.

I agree with the previous conclusion if three different tools give the result but over 50% of the text is generated by AI, and the writer did not emphasize this, such posts should be removed.

[KingsDen]

I agree with the previous conclusion if three different tools give the result but over 50% of the text is generated by AI, and the writer did not emphasize this, such posts should be removed.

Agreeing with Hugeblack's 3 tools but the difference is in the allowable percentage. While hugeblack and a vice president has said 80%, you said 50%.
Maybe we should start suggesting good or reliable AI detection tools here. To make the work easier.

[examplens]

Agreeing with Hugeblack's 3 tools but the difference is in the allowable percentage. While hugeblack and a vice president has said 80%, you said 50%.
Maybe we should start suggesting good or reliable AI detection tools here. To make the work easier.

Even 50% is too much, but it gives room for a more free interpretation.
Would this discussion be interesting to you, if some AI program answers instead of me? even 10%...

[ABCbits]

Using 3 to 5 tools and a percentage of more than 80% may be enough to confirm that a post is written by AI, and the lack of connection of ideas and the change in language between each reply is evidence that the post was done by AI.

Your suggestion sounds reasonable. But what to do if the AI/chatbot detection show some statement (e.g. "This text generated by AI" or "This text probably generated by AI") rather than percentage? Should we group those statement along with >80%?

[Jokers]

Some translation or grammar tools can make the post seem like AI wrote it, but I believe no tool will give it more than 20% of AI-generated text.

Most popular ones don't. I tested Grammarly, Google Translate, Yandex Translate, DeepL and some more. The result of their work doesn't give AI detection. The problems were with those who said they used AI bots not to write texts, but to translate or improve them... well, I can say only: don't use AI bots for any purposes on forum posts.

[Dr.Bitcoin_Strange]

I don't just see need why some members will even want to use bot in the first place, if anyone wants to discuss with a bot or get a result from a bot, I believe the person knows how to go and seek the answer they want from a bot, but rather they come here for a human to human discussion and instead of giving out our immense ideas and contribution, some people will choose to use a bot. It's better any one caught in such must be punished.

[bitterguy28]

Then, there is a challenge even in the other forum. Users who use AI to generate text often deny when caught and there's no 100% effective tools to detect AI. There are tools but I cannot vouch for them. They can be accurate with one post and total crap with another post. Based on this, to what percentage will a users post show that it is AI generated before they will receive the AI badge?

in today’s time thankfully even without tools you can still manage to detect ai generated texts just on your own

1. if a user forgot to remove ai addressing the prompt submitted by used for example “absolutely!” or “not enough evidence for your question” you can have suspicions that it is indeed ai
2. even the most intelligent and most fluent person in english is still subjected to some mistakes which ai unusually has none so if everything is way too good and the text is unnecessary long then that’s probably ai

like you said there’s no accurate tool yet but i think media literacy is enough to see some signs that it is indeed ai generated texts

[dragononcrypto]

While hugeblack and a vice president has said 80%, you said 50%.

Just to clarify I'm not a vice president, hair is currently VP. I'm the 2nd President in the context of Freemind being the 5th. Carry on...

[Findingnemo]

Agreeing with Hugeblack's 3 tools but the difference is in the allowable percentage. While hugeblack and a vice president has said 80%, you said 50%.
Maybe we should start suggesting good or reliable AI detection tools here. To make the work easier.

Even 50% is too much, but it gives room for a more free interpretation.
Would this discussion be interesting to you, if some AI program answers instead of me? even 10%...

Let me quote what admin wrote in one of the sticky.

- You can use AI, as long as it's not above 30%.

So I guess anything less than 30% won't be an issue but it's hardly possible to get accuracy of AI usage and also anyone uses AI will obviously get caught because we will see 100% AI contents for most of the contents created using AI tools unless they found a new way to make it undetectable.