[ExI] chatgpt can catch itself... sometimes

[spike at rainier66.com]

OpenAI has figured out how to identify AI-generated text, sorta:

 

 

"Our classifier correctly identifies 26% of AI-written text (true positives)
as 'likely AI-written,' while incorrectly labeling human-written text as
AI-written 9% of the time (false positives)," the OpenAI employees wrote.

 

 

Cool, OK so now we have a fun game or challenge: write something, see if you
can fool it into identifying your writing as AI-generated.  Oh the poetic
beauty: a kind of reverse Turing test.

 

spike

 

 

 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 3702 bytes
Desc: not available
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230131/89ba95b7/attachment-0001.bin>