Sorry to see that new AI category

[vickiespencer]

2 minutes ago, uk1000 said:

It can't be accurate if newsmike actually wrote the text shown in that check himself.

But @newsmikedid not write the content. He copied and pasted it from @sonam_abro's post.  

[uk1000]

1 hour ago, newsmike said:

Vickie, why would you assume that he stole that, after all, he is an "SEO and eCommerce" expert according to his profile, and has been on platform since 2017 with a grand total of zero sales?  Plus, his profile states that he is also fluent in English, so there's that as well.  

Newsmike wrote the above post - in this thread.

Look at the 2 screenshots I posted above with the checks from copyleaks.com ai detector

(this post https://community.fiverr.com/forums/topic/311490-how-do-i-get-my-first-order/?do=findComment&comment=1952304)

copyleaks' ai detector then said "ai content detected" - which cannot be accurate if newsmike wrote that post of his himself.

which shows that the copy leaks ai detector is not always accurate (and can't properly be used as proof that some content is or isn't AI generated content).

Edited August 19, 2023 by uk1000

[vickiespencer]

@newsmike Who wrote your post as cited by UK above? Are you messing around with Chatgpt? 

[newsmike]

21 minutes ago, vickiespencer said:

@newsmike Who wrote your post as cited by UK above? Are you messing around with Chatgpt? 

Nope.  But without proper supervision and rigid monitoring by a tribunal of international scientists, I cannot comment on what went on in UK's lab tests. I believe his test came from Wuhan. 

Edited August 19, 2023 by newsmike

[uk1000]

4 minutes ago, newsmike said:

I cannot comment on what went on in UK's lab tests

You can try the same link yourself (the detector you used before in this thread) with your same text from your post (the same one I checked) and see if you get different results:

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

If so you can screenshot it and attach the different results here.

[newsmike]

Just now, uk1000 said:

You can try the same link yourself (the detector you used before in this thread) with your same text from your post (the same one I checked) and see if you get different results:

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

If so you can screenshot it and attach the different results here.

[uk1000]

2 hours ago, newsmike said:

Vickie, why would you assume that he stole that, after all, he is an "SEO and eCommerce" expert according to his profile, and has been on platform since 2017 with a grand total of zero sales?  Plus, his profile states that he is also fluent in English, so there's that as well.  

Try checking the text from your quoted post above in https://copyleaks.com/ai-content-detector

and post a screenshot of the results here if they are different to this:

Edited August 19, 2023 by uk1000

[newsmike]

4 minutes ago, uk1000 said:

Try checking the text from your quoted post above in https://copyleaks.com/ai-content-detector

and post a screenshot of the results here if they are different to the one I showed in this post:

https://community.fiverr.com/forums/topic/311490-how-do-i-get-my-first-order/?do=findComment&comment=1952304

 

Actually, I've had my fill of "trick the AI engine."  But feel free to keep going.  Bottom line is that even without that AI checker we know dead certain that the person who posted the content did not write it. 

Edited August 19, 2023 by newsmike

[vickiespencer]

UK seems to like to argue for the sake of arguing, and I wondered why so I asked Google only to find out the following. 

 

[uk1000]

16 minutes ago, vickiespencer said:

UK seems to like to argue for the sake of arguing

It's not arguing for the sake of arguing. If someone is trying to show proof of someone breaking the forum rules by posting AI generated content (without citing) by showing a test result which is inaccurate but which is meant to show as valid proof of rule breaking it's unfair to the forum posters who have their post shown as "breaking the rule" like that. Since the same could be done to any of our posts and some AI content detector will likely say our posts are AI generated (and breaking the rules) whether they are or not.

If a test is flawed it shouldn't be shown as accurate proof of rule breaking.

And while in this thread it might be that a couple of the previous posters used AI, other posters may not do but if we keep showing "AI content generators" detection as if it's 100% accurate then there will be other times when it gets it wrong (like when it showed newsmikes's post as being "AI generated content" with high degree of probability).

If it's an inaccurate test it should either not be used to prove rule breaking or it should say in the post with the screenshot that the test isn't necessarily accurate.

Edited August 19, 2023 by uk1000

[newsmike]

12 minutes ago, uk1000 said:

If someone is trying to show proof of someone breaking the forum rules by posting AI generated content

I'm not trying to catch someone "breaking rules" but merely flagging obvious fakery when I see it. Just like the "English Fluent" nonsense we flag daily, the sad truth is that there may be rules, but there is absolutely no enforcement here. I'm merely pointing it our so the OP knows that we are aware of their chicanery.  Nothing more. I've been at this long enough to not expect anything official to happen. And to expect a few to always come to the defense of the fraudsters. 

Edited August 19, 2023 by newsmike

[uk1000]

5 minutes ago, newsmike said:

I'm not trying to catch someone "breaking rules" but merely flagging obvious fakery when I see it.

But you're showing the screenshot as though it is accurate proof, when we know that ai detectors (including that one) aren't always accurate (as shown above).

If we keep doing the same again, other posts (including human written ones) will probably get the same done to them, may get flagged as AI and removed from the forum, users may get warnings, all because it was helped by inaccurate ai detector screenshots shown as proof.

So if we still want to do that, we could just add a line beneath the screenshot saying "note: AI content generators aren't always accurate" or something like that (in big enough text, and not show that it totally proves that it's AI content (since those detectors can't prove it totally accurately yet)).

Edited August 19, 2023 by uk1000

[newsmike]

Just now, uk1000 said:

when we know that ai detectors

That is not what (WE) know.  It is what you believe. 

[uk1000]

6 minutes ago, newsmike said:

That is not what (WE) know.  It is what you believe. 

I've shown you proof (in the screenshots in this thread).

1) You wrote a post (which you say you did without using AI)

2) I put that text from that post into the ai detector you seem to think is accurate (with no other text, no quotes by anyone else)

3) That ai detector said "ai content detected"

4) That proves that the ai content detector that you used previously is inaccurate, if you wrote the post you said you did. As either it is human written and the AI detector wrong or it contains AI content when you said you wrote it all yourself.

You saying you wrote it all yourself and it showing "ai content detected" (which you can check by pasting the same text), proves that ai detector is not always accurate (if you wrote that text by yourself).

Edited August 19, 2023 by uk1000

[newsmike]

Just now, uk1000 said:

I've shown you proof.

1) You wrote a post (which you say you did without using AI)

2) I put that text from that post into the ai detector you seem to think is accurate

3) That ai detector said "ai content detected"

4) That proves that the ai content detector that you used previously is inaccurate, if you wrote the post you said you did. As either it is human written and the AI detector wrong or it contains AI content when you said you wrote it all yourself.

You saying you wrote it all yourself and it showing "ai content detected" (which you can check by pasting the same text), proves that ai detector is not always accurate.

1. You will not move my opinion on this should you write another 300,000 words on the topic.

2. You may believe what you like, I don't care.  Be happy. 

3. I feel no need to justify what I believe to you. I'm happy.

4. It feels as if @vickiespencer is correct, that you really like to argue for the sake of arguing.

5. It is a lovely Saturday afternoon, and I am going to do the following.

• Make an espresso.
• Listen to Duke Ellington's orchestra.
• Respond to any further discussion of this with only emojis.

 

🙂

[uk1000]

This link will also help prove the point that AI content generators are not always accurate:

https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text

OpenAI (the people behind ChatGPT that lots of AI text is generated from) wrote in the above link:

Quote

As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy.

Quote

Our classifier is not fully reliable. In our evaluations on a “challenge set” of English texts, our classifier correctly identifies 26% of AI-written text (true positives) as “likely AI-written,” while incorrectly labeling human-written text as AI-written 9% of the time (false positives).

That's about OpenAI's detector (different from the other 2 inaccurate detectors that were used in this thread. But it shows that even the maker or Chat GPT knows that the AI detectors are inaccurate (since they removed their own detector (in their own words) "due to its low rate of accuracy").

Edited August 19, 2023 by uk1000

[newsmike]

2 minutes ago, uk1000 said:

This link will also help prove the point that AI content generators are not always accurate:

https://openai.com/blog/new-ai-classifier-for-indicating-ai-written-text

OpenAI (the people behind ChatGPT that lots of AI text is generated from) wrote in the above link:

That's about OpenAI's detector (different from the other 2 inaccurate detectors that were used in this thread. But it shows that even the maker or Chat GPT knows that the AI detectors are inaccurate (since they removed their own detector (in their own words) "due to its low rate of accuracy".

🙃

[rabihumakhan]

Well, it's true that AI detectors cannot/don't always produce accurate results; it's also true that the above post which NewsMike pointed out has been written through AI. 

How can I be  so sure without even checking it myself?

1. AI, specifically chat gpt has a certain type of writing style and it always produces generalized results which may not be even exactly relevant to the question/s asked. 

2. The comment was written in fluent English with no or fewer grammatical mistakes; however, the person isn't fluent in English. 

[uk1000]

12 minutes ago, rabihumakhan said:

Well, it's true that AI detectors cannot/don't always produce accurate results;

Thanks. So we shouldn't be using those inaccurate tests as though they are totally accurate then. We should at least put a disclaimer with them (eg. under the screenshot) because of their inaccuracy, since they aren't valid, accurate proof.

Quote

AI, specifically chat gpt has a certain type of writing style and it always produces generalized results

I don't know if ChatGPT always produces the same style (I assume it depends on what prompt you enter into it and if you can select different language models with it, which one you selected. eg. if you said in the prompt "write this answer in the style of William Shakespeare" then it would probably write it in a different style than the style they are using).

But also ChatGPT isn't the only possible AI generator so it could be different for others (eg. ones trained differently).

Edited August 19, 2023 by uk1000

[maitasun]

2 minutes ago, uk1000 said:

So we shouldn't be using those inaccurate tests as though they are totally accurate then. We should at least put a disclaimer with them (eg. under the screenshot) because of their inaccuracy, since they aren't valid, accurate proof.

Wrong! And please, just stop it UK!

What no one should be using is AI content. Period!

And regarding the disclaimers, it's them, those who use AI, who should put the disclaimer, not us! Period!

I'm so fed up with this [fill in the blank]...

[rabihumakhan]

Lol I this noon downloaded an AI detector on my Android phone. 

I wanted to 'show off' to one of my new clients who needed a few technical write-ups and was very happy about the articles that I wrote for him (it was on another marketplace).  However, as soon as I copied one of my articles and pasted it to the app, it showed 82 percent AI generated results. I fainted but then I personally wrote a few nonsense sentences and it had the same percentage of AI detection. 

 

I swear to the God of moon, sun, and galaxies, I immediately deleted that 'toxic' app. 😉

[uk1000]

9 minutes ago, maitasun said:

Wrong! And please, just stop it UK!

So you want forum users to continue posting screenshots of AI content detectors with forum posts where we know for fact that they are inaccurate, and to continue posting them as though they are accurate (with no disclaimer), which can falsely accuse some of using AI content without citing (and therefore of breaking the forum rules with false proof)?

Edited August 19, 2023 by uk1000

[vickiespencer]

4 minutes ago, maitasun said:

Wrong! And please, just stop it UK!

What no one should be using is AI content. Period!

[maitasun]

Just now, uk1000 said:

So you want us to continue posting screenshots of AI content generators with forum posts where we know for fact that they are inaccurate, and to continue posting them as though they are accurate, which can falsely accuse some of using AI content without citing?

Two can play this game, UK...

So you want to continue letting people who we know for a fact are not even able to spell Fiverr correctly to post AI content as if it had been written by them, although WE ALL KNOW they're not able to do so, and falsely lead others to think that content is theirs?

This is a very sad joke you're defending.

[vickiespencer]

On 8/19/2023 at 4:24 PM, uk1000 said:

So you want forum users to continue posting screenshots of AI content generators