Do you trust AI detector bullshit?

Even when I check human written content with so called AI detectors, they still say it is AI generated content.

which AI detector are you using?

Let us think logically for just 1 minute only.
AI detectors : who creates or codes them ? The same human beings with the same intelligence as the human beings who creates or codes AI content writers ... Correct ?
And quite often I have seen people mention that AI created content needs to be proofread and corrected ... Correct ?
This means that AI created content is never always perfect to publish straightaway ... Correct ?
So, if AI created content is never always perfect, then how can AI detectors always be perfect ? ... Considering that BOTH AI content creators AND AI content detectors are created or coded by the human being coders with the same levels of overall intelligence.
In other words, if AI created content cannot always be perfect, then it is logically so that AI detectors are also not always perfect. Because if human beings have coded a perfect AI detector, then why are human beings so dumb as to be unable to code a perfect AI content writer or AI content creator ?
Or is it that certain human beings are smart enough to fool the world by deliberately coding AI content writers to make a few grammar or factual mistakes, in order to make a demand for AI detection as well ? We will put a bit of "disease" mixed in the food, so that you also buy the cure as well ?
"The game is a lot bigger than you think you know and if you think you know, then I don't think you know" (DMX).

Too many false positives in my opinion. But interested to hear what everyone is currently using as the standard

As someone who is finetuning a non-openai model, I find results totally random and unreliable from these detectors.

These guys can’t even develop tools that can detect article forge content. They are just a bunch of clowns. If it was possible, google would develop one already.

710fla said:

Too many false positives in my opinion. But interested to hear what everyone is currently using as the standard

Click to expand...

Same. I used Originality.ai and it had a ton of false positives. Don't think any of the AI detectors are actually good.

Trust WorkShop said:

Let us think logically for just 1 minute only.
AI detectors: who creates or codes them? The same human beings with the same intelligence as the human beings who creates or code AI content writers ... Correct?
And quite often I have seen people mention that AI-created content needs to be proofread and corrected ... Correct?
This means that AI-created content is never perfect to publish straightaway ... Correct?
So, if AI-created content is never always perfect, then how can AI detectors always be perfect ? ... Considering that BOTH AI content creators AND AI content detectors are created or coded by human being coders with the same levels of overall intelligence.
In other words, if AI-created content cannot always be perfect, then it is logical that AI detectors are also not always perfect. Because if human beings have coded a perfect AI detector, then why are human beings so dumb as to be unable to code a perfect AI content writer or AI content creator?
Or is it that certain human beings are smart enough to fool the world by deliberately coding AI content writers to make a few grammar or factual mistakes, in order to make a demand for AI detection as well? We will put a bit of "disease" mixed in the food so that you also buy the cure as well?
The game is a lot bigger than you think you know and if you think you know, then I don't think you know (DMX).

Click to expand...

Your answer is worth a lot of plusses
I also corrected some grammar with Grammarly

Don Goldstein said:

These guys can’t even develop tools that can detect article forge content. They are just a bunch of clowns. If it was possible, google would develop one already.

Click to expand...

I’ve tested five pieces of AF content with GPT-2 detector. It displays 95-98% fake content.

For what ai allowed on Google so what the point.

If shit get indexed from Ai then understand.

Would be a idiot trying it on with ai and burning money.

Don't know... the AI detector, I use, has given 100% accurate results (OK, I've done about 20-30 tests).
Basic writing might get picked up as AI, I guess.

praetserge said:

Don't know... the AI detector, I use, has given 100% accurate results (OK, I've done about 20-30 tests).
Basic writing might get picked up as AI, I guess.

Click to expand...

what r u using?

kukuruto said:

what r u using?

Click to expand...

openai-openai-detector.hf.space/

Trust WorkShop said:

So, if AI created content is never always perfect, then how can AI detectors always be perfect ? ... Considering that BOTH AI content creators AND AI content detectors are created or coded by the human being coders with the same levels of overall intelligence.

Click to expand...

Trust WorkShop said:

In other words, if AI created content cannot always be perfect, then it is logically so that AI detectors are also not always perfect. Because if human beings have coded a perfect AI detector, then why are human beings so dumb as to be unable to code a perfect AI content writer or AI content creator ?

Click to expand...

Yes both can make mistakes, but these are very wrong analogies, you can't just compare like that, the ways they work are so fundamentally different.
You clearly don't know what you're talking about.

Trust WorkShop said:

Or is it that certain human beings are smart enough to fool the world by deliberately coding AI content writers to make a few grammar or factual mistakes, in order to make a demand for AI detection as well ? We will put a bit of "disease" mixed in the food, so that you also buy the cure as well ?

Click to expand...

Actually, openai is working on a cryptographic signature to detect content generated by its gpt variants,
So detection would be possible (with a certain accuracy) regardless of mistakes in the generated content.
The longer the text is the more accurate the detection will be.

Trust WorkShop said:

"The game is a lot bigger than you think you know and if you think you know, then I don't think you know" (DMX).

Click to expand...

Clearly applies to you.

nuclearrap said:

Even when I check human written content with so called AI detectors, they still say it is AI generated content.

Click to expand...

Yes and no. It depends. If there is a repeating pattern it will be caught. If a software is just guessing, it won’t be. Don’t believe what others are saying. Do your own research if that is something you’re willing to work for.


When I say pattern I mean having repeated phrases from a large dataset.

Just manually proofread it and correct the facts, then no one can accurately predict it's an AI generated content unless OpenAI starts to take the backup of every query and response. Then only OpenAI's detection tool can predict more accurately.

i just copied directly out of chatGBT and pasted in GPT-2 output detector and got these results. i tried a half dozen different subjects and got similar results. nothing is really refined yet for the AI detection as i also pasted something that i wrote and that was the lowest for being real.

Prediction based on 235 tokens
Real: 99.92%
Fake: 0.08%

but i guess at the end of the day what does it matter if it passes plagiarism and the grammar and content is good.

Too many mixed results. Don't trust any of it. Also, stop using AI for content, this strategy will sink your results in the long run

What's there not to trust LOL
It's simple math to detect those. The statistical probability of what word goes after another. Its really funny how people think of Open AI as real ai to much

if the end user is happy that is what matters... i've been doing this since the early 2000s and staying ahead of the curve is what it takes using anything and everything for tools... and exploiting this tool is where it is at for the moment.