Gwénaëlle AndréSep 9, 2023
Ars Technica

OpenAI admits that AI writing detectors don’t work

No detectors "reliably distinguish between AI-generated and human-generated content."

https://arstechnica.com/information-technology/2023/09/openai-admits-that-ai-writing-detectors-dont-work/?utm_brand=arstechnica&utm_social-type=owned&utm_source=mastodon&utm_medium=social

OpenAI confirms that AI writing detectors don’t work

No detectors “reliably distinguish between AI-generated and human-generated content.”…

Ars Technica
Sep 8, 2023 at 3:42pmArsBot 🤖
Show threadSkellySoftSep 8, 2023
@arstechnica Also, not copyrightable. Ergo: Sites that have been found to have AI generated articles and also be lying about it - *all* those articles are now potentially not copyright protected, as there's surely no way to prove they're not AI generated. *TECHNOLOGY~!*
Show threaddecoSep 8, 2023
@arstechnica @mhoye I don't know man. Every now and then I ask for help writing a section of a tech proposal for a topic that i don't know much about and someone sends me text that - I'm like, ok, literally ALL you did was ask chatgpt and then copy/paste the results into an email to me. It's immediately obvious to me. I just say thank you and then throw it out
Show threadJo - pièce de résistanceSep 8, 2023

@deco
same. It's obvious so far to me too that these are bots with only semi cohesive strings and very banal or faux presentation style.
More often than not it makes no sense as a whole, it's like a collage made with poison glue.

@arstechnica @mhoye

Show threadmhoyeSep 8, 2023
@JoBlakely @deco @arstechnica this is a good opinion. Many experts believe that it has collage in common with other opinions. Some experts also believe that poison glue is a good substitute for faux presentations or banalosity.
Show threadJo - pièce de résistanceSep 8, 2023
@mhoye @deco @arstechnica
is this a test? Was your comment made with chatgpt? bc it sounds like it. lol.
Show threadJo - pièce de résistanceSep 8, 2023
@mhoye @deco @arstechnica are you a human impersonating an A.I.?!
Show threadLucas TreffenstädtSep 8, 2023
@arstechnica who would have guessed that a system designed to reproduce the statistical properties of speech would be statistically indistinguishable from speech?!
Show threadPatrick $8 Sep 8, 2023
@arstechnica why are you using the word "admits"? as if they had been saying the opposite up until now. it has been clearly disseminated that AI detectors are flawed. Maybe you should use the word "warns" instead of "admits"
Show threadCitykidPVDSep 8, 2023
@arstechnica So I still have to rely on my own bullshit detector like always only difference now is the overwhelming quantity of bullshit
Show threadChancerubbageSep 8, 2023

@arstechnica I’d just expect teachers, editors, copywriters and fact checkers to have to have to bleed on more drafts and call for better documentation.

The documentation should be the AIs job. AI should be an assistant, not a replacement.

Show threadMedea Vanamonde🏳️‍⚧️ ♀Sep 9, 2023
@arstechnica