OpenAI just admitted it can't identify AI-generated text.

[dvc]

https://www.businessinsider.com/ope...generated-text-bad-for-internet-models-2023-7
From the article

"The company warned that it's impossible to reliably detect all AI-written text. However, OpenAI said good classifiers are important for tackling several problematic situations. Those include false claims that AI-generated text was written by a human, running automated misinformation campaigns, and using AI tools to cheat on homework.

Less than seven months later, the project was scrapped.

"As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy," OpenAI wrote in a recent blog. "We are working to incorporate feedback and are currently researching more effective provenance techniques for text."

 

[DevMask]

It is a matter of common sense though. A language can be expressed in a multitude of styles, formats, and tones, whether by humans or AI.

So any assertions about AI-generated content being treated differently by Google or other platforms are just baseless.

 

[nakamura]

It is a matter of common sense though. A language can be expressed in a multitude of styles, formats, and tones, whether by humans or AI.

So any assertions about AI-generated content being treated differently by Google or other platforms are just baseless.

This.

People need to stop having this black and white thinking. AI generated content can look more human-like than content written by a random person.

 

[MoonStarz]

most of the AI detectors are fake and always provide false positive, I notice that the false positive was always when using a really formal language, but well let's see what will happen with those sites now.

 

[Proxy_prinz]

Hopefully someday this will be taken care of. And they will force you to sign AI content. It's scary to imagine a future with AI-only content

 

[silvex]

It is a matter of common sense though. A language can be expressed in a multitude of styles, formats, and tones, whether by humans or AI.

So any assertions about AI-generated content being treated differently by Google or other platforms are just baseless.

Aboslutely true.

Language models can definetly find patterns, but who am I to say that only AI uses such patterns? I can definetly be an AI writing this comment and Google, or any other who uses such tools, could only assume it's true or not.

Total BS

 

[dvc]

It is a matter of common sense though. A language can be expressed in a multitude of styles, formats, and tones, whether by humans or AI.

So any assertions about AI-generated content being treated differently by Google or other platforms are just baseless.

Good to see people with logic behind thinking.

 

[moonlighsunligh]

This.

People need to stop having this black and white thinking. AI generated content can look more human-like than content written by a random person.

Not really, if you check AI text and human text you will see that AI text always have lower average perplexity.

Perplexity for example can be calculate using this tool - https://huggingface.co/spaces/PirateXX/Sentencewise-Perplexity.

People's writing has high perplexity, AI text has low.

Many AI content detectors use that to detect AI content.

 

[moonlighsunligh]

For example this is content generated by GPT chat. Sentences with low perplexity <20 look incredibly boring. Human text rarely has perplexity lower that 25-30.

 

[nakamura]

Not really, if you check AI text and human text you will see that AI text always have lower average perplexity.

Perplexity for example can be calculate using this tool - https://huggingface.co/spaces/PirateXX/Sentencewise-Perplexity.

People's writing has high perplexity, AI text has low.

Many AI content detectors use that to detect AI content.

I wrote: Stop the black and white thinking.

Just let me remind you of this: Google is also indexing and ranking absolute shitposts from Reddit or forums. I am talking posts with terrible grammar, no punctuations, etc. with little to no perplexity.

The sad truth is the most people are not capable of writing content with high perplexity. This makes AI content better than average human content.

In the end, it doesn't matter what a tool calculates. It depends what Google ranks and indexes. And Google is indexing and ranking crappy content all day every day.

 

[Elvlin]

anyone saying they can identify ai generate text is just BS.
now a day it's just niche product that monetize on people fear.

 

[GenesisOne]

Plenty of people can't identify text too

 

[barber]

"As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy," OpenAI wrote in a recent blog. "We are working to incorporate feedback and are currently researching more effective provenance techniques for text."

that's why the page is always blank

 

[Deleted member 1635563]

https://www.businessinsider.com/ope...generated-text-bad-for-internet-models-2023-7
From the article

"The company warned that it's impossible to reliably detect all AI-written text. However, OpenAI said good classifiers are important for tackling several problematic situations. Those include false claims that AI-generated text was written by a human, running automated misinformation campaigns, and using AI tools to cheat on homework.

Less than seven months later, the project was scrapped.

"As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy," OpenAI wrote in a recent blog. "We are working to incorporate feedback and are currently researching more effective provenance techniques for text."

That is normal, how they will going to do it haha

 

[kayman28]

anyone saying they can identify ai generate text is just BS.
now a day it's just niche product that monetize on people fear.

Tell that to writing clients. I once wrote an article myself, sent it to the client and he came back with “Looks like you used the assistance of GPT”.

Hit the client back with “I also have a GPT zero premium account I use for an AI side project, if I used GPT, I’d have confirmed it wasn’t flagging content before sending it to you”

Kinda annoying when you don’t use AI at all and the client comes at you with the confidence that you did. Because GPT zero or originality came back with some “90% or this article was written by a human”

 

[Elvlin]

Tell that to writing clients. I once wrote an article myself, sent it to the client and he came back with “Looks like you used the assistance of GPT”.

Hit the client back with “I also have a GPT zero premium account I use for an AI side project, if I used GPT, I’d have confirmed it wasn’t flagging content before sending it to you”

Kinda annoying when you don’t use AI at all and the client comes at you with the confidence that you did. Because GPT zero or originality came back with some “90% or this article was written by a human”

hard to say if it's client, there is a reason people say customer is king. picking up and debate with them wil just give bad rep. my suggestion it's all boild down to is the pay vs work worth it or not. since there are also customer that just say or do anything to get thing for free. better skip this kind waste of time.
yeah i have also tested a real human written article that being told it's ai writen

 

[kayman28]

hard to say if it's client, there is a reason people say customer is king. picking up and debate with them wil just give bad rep. my suggestion it's all boild down to is the pay vs work worth it or not. since there are also customer that just say or do anything to get thing for free. better skip this kind waste of time.
yeah i have also tested a real human written article that being told it's ai writen

I think you said it perfectly!

 

[PURSUIT MEDIA]

Can we just all enjoy this AI moment in peace?

 

[xReminisce]

https://www.businessinsider.com/ope...generated-text-bad-for-internet-models-2023-7
From the article

"The company warned that it's impossible to reliably detect all AI-written text. However, OpenAI said good classifiers are important for tackling several problematic situations. Those include false claims that AI-generated text was written by a human, running automated misinformation campaigns, and using AI tools to cheat on homework.

Less than seven months later, the project was scrapped.

"As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy," OpenAI wrote in a recent blog. "We are working to incorporate feedback and are currently researching more effective provenance techniques for text."

As mentioned so many times when anyone claims that google or any AI detector can detect AI content
It's just simply not true.
I could be an AI but none of you would ever know

 

[hulloween]

There were rumours a few months after ChatGPT launched that many foreign language writers were getting flagged by these so called AI detectors. At that point I suspected detection was a lost cause.

We often think about detection in terms of detection good content as AI, but a much bigger issue is the false positives of all the terrible, but non AI content on the web, which might get flagged as AI.