GPT Countermeasures you should know

[PinguSpy]

I don’t think G could care less about a.i
In the end G will always favor A.I content from big site than manual content from low authority site. Save money than to run a system to scan/process either the content is written by human or robot.

 

[Aglet49]

It's all bullshit. Articles I've written myself are sometimes detected as AI (false positive). It is pretty good at identifying AI generated as AI though.

Lots of false positives but no false negatives...

...Until you throw in a couple of spelling/Grammer mistakes in the AI generated text.

Then it detects it as "human" (false negative).

So the whole concept is still a work in progress in my experience.

 

[malayan]

"Hey guys, I know OpenAI has invested millions into building one of the most sophisticated NLG in history, but I threw together a few lines of Python that can somehow decipher their work!"

Five minutes later it's on Business Insider or some other hackneyed publication.

Most of the guys publishing these tools are on some Greta Turnberg-levels of sanctimonious BS. They know they're full of crap, but who's going to stand up and call them out?

 

[Mr Positive]

Google is fine with AI content just verify the facts, and make it sound good by manual edit you’ll should be fine.

 

[JavascriptPufferfish]

Not only repetitions, but also patterns. Example:
123 ?
432 ?
abcd ?

- or -

123?
432?
abcd?

-- Which one is human written? --

i would bet first one since who tf puts spaces between what they write and the "?". but both have patterns (one just looks more artificial), so i am not sure i follow you here.

 

[Aglet49]

i would bet first one since who tf puts spaces between what they write and the "?". but both have patterns (one just looks more artificial), so i am not sure i follow you here.

I see non-English speakers write like that all the time. Maybe in their language it's correct grammar

 

[JavascriptPufferfish]

i would bet first one since who tf puts spaces between what they write and the "?". but both have patterns (one just looks more artificial), so i am not sure i follow you here.

i meant to say first one is bot

I see non-English speakers write like that all the time. Maybe in their language it's correct grammar

really? i see many weird things about how we write, but i have never seen that specific one.

 

[Aglet49]

i meant to say first one is bot

really? i see many weird things about how we write, but i have never seen that specific one.

A quick Google search shows French is one example where a space before "?" is actually correct.

https://www.thoughtco.com/how-to-use-french-punctuation-4086509
I'm sure there are other languages like that too.

 

[nohupt]

As I wrote in another thread. They can only detect repetitions. Adjust your prompt accordingly and you're fine.

yup. i wrote my thing from scratch and thus far it's been working fine for years now. there was one site that lost ranking and i was wondering what exactly caused it, then i started catching repetitions and was like "doh!"

 

[TheMarquis]

i would bet first one since who tf puts spaces between what they write and the "?". but both have patterns (one just looks more artificial), so i am not sure i follow you here.

I did not provide the best example, but this is how I see the idea of an AI pattern (repetitive text, wrong spaces before symbols - question marks, for example).

 

[bartosimpsonio]

I've bought human content so bad a drunk AI could've done better.

I don't think there's a definitive watermark in AI content. But one can tell it's bad content, which gets you penalized either way.

 

[MoonStarz]

It was just time for this to happen, even though they can detect AI content you can adjust that content and make a few modifications even after your review depending in what you are trying to achieve using the AI, you are saving a lot of time just need to go through a human filter.

 

[KuroKami87]

Just a heads up, the AI detectors are bullshit. Have tried a few, including paid options like Originality.ai and hyped ones like GPTZero.

Originality had the most issues with false positives, had human-written content pegged as AI numerous times. Annoying as hell.

GPTZero was easy enough to trick with Quillbot as was a bunch of other AI detectors I found through Google.

 

[IM Dude]

Looks like "Revenge of the nerds" twitter thread.
Don't do anything and just ignore them, nothing is going to happen to your site.

 

[bandititf9]

Somebody mentioned this in another thread, if you copy AI content and make few small typos in your text on purpose, it automatically detects it as "Human written" article.
Because AI cannot have typos, right?

These tools are useless, for real.

 

[kukuruto]

prompt chatGPT
write a code for AI Detectors

 

[Shreecaterers]

We already know that GPT-3 can be used as a source code generator to ... “If we start thinking about countermeasures for potential

 

[SupperMario]

Sooner or later, simply copying directly from ChatGPT will be punished, it's just a matter of time

 

[leeVanCleef]

Unless the AI providers build a search engine of text they generated so that the provenance of sentences can be checked then there will never be a remotely accurate way of determining if text is AI generated.

When the penny drops on this I wonder where the first demands for this platform to be banned will come from. I mean this is going to completely hammer so many things in the real world.

As for google, if a site appears overnight with 40,000 pages of optimised text then I hope to god they dont think they need an AI detector to work out what is going on here.