Quote from: LoyceV on Today at 07:31:29 AM

It probably stores everything it ever created, so in this case they know how it was generated. Now add some spelling mistakes and make the language less complicated, and it'll be harder to detect.

Lol.

You could be right, they probably have a record of every data that goes through any of their tools, and their detector might be using that data for detecting the content along with other metrics that they might have added through algorithms, etc.

Quote from: LoyceV on Today at 07:31:29 AM

Quote

Response generated through ChatGPT:

I'd say it's a generic shitpost. It uses fancy words and most of it is correct, but it doesn't add anything. It sounds like listening to the average politician: using as many words as possible to say nothing the reader doesn't know yet, but you can agree with it because it's (more or less) correct.

Of course, it is a generic shitpost with fancy words and correct grammar, that's what these LLMs are good at. If you ever notice, when an AI model, or an LLM, is fed some text and asked to generate something using that, either an answer or anything, it will basically take that content, twist it, correct the mistakes if there are any, paraphrase every sentence, maybe add a few extra lines but with the same context or meaning, and then present that to you as the response. You can even see this if you look at @nutildah's post that I used, and the AI-generated response to it that you just read.