Mastodawn

L4sBot Jul 28, 2023

OpenAI just admitted it can't identify AI-generated text. That's bad for the internet and it could be really bad for AI models.

https://lemmy.world/post/2293851

OpenAI just admitted it can't identify AI-generated text. That's bad for the internet and it could be really bad for AI models. - Lemmy.world

OpenAI just admitted it can’t identify AI-generated text. That’s bad for the internet and it could be really bad for AI models.::In January, OpenAI launched a system for identifying AI-generated text. This month, the company scrapped it.

Show thread

ChrislyBear Jul 28, 2023

So every accusation of cheating/plagiarism etc. and the resulting bad grades need to be revised because the AI checker incorrectly labelled submissions as “created by AI”? OK.

Show thread

Cethin

That’s not even the biggest issue here. The big thing that comes with this is that they can train their model against the detection tool. This will train the model to be harder to detect (using the same method). It doesn’t necessarily improve the output quality, just the ability to tell that it’s generated.