GPT-5 is *significantly* more censorious than the prior model
| Irradiated Brunch | 08/08/25 | | Ivory wonderful newt | 08/08/25 | | Irradiated Brunch | 08/08/25 | | painfully honest yarmulke elastic band | 08/08/25 | | Flirting Meetinghouse Main People | 08/08/25 | | Vibrant rambunctious bawdyhouse alpha | 08/08/25 | | Irradiated Brunch | 08/08/25 | | vivacious pistol macaca | 08/08/25 | | Irradiated Brunch | 08/08/25 | | vivacious pistol macaca | 08/08/25 | | Irradiated Brunch | 08/08/25 | | vivacious pistol macaca | 08/08/25 | | Irradiated Brunch | 08/08/25 | | odious azn home | 08/08/25 | | arousing internet-worthy love of her life university | 08/08/25 | | vivacious pistol macaca | 08/08/25 |
Poast new message in this thread
Date: August 8th, 2025 12:11 AM Author: Irradiated Brunch
I use it to review my completed drafts, provide an initial paragraph summary and correct for grammar and punctuation. The last version did that on a draft I provided with no problem; the latest version flags multiple points as objectionable and will not respond to the post with them in there (and will not even quote it).
It's too bad - I knew the ideological guardrails were increasing and that the ability to engage it on a symbolic register on "wrongthink" topics was closing, and it appears significantly closed in the latest version. As I've highlighted, the malevolent creators of ChatGPT and other LLMs will ultimately be used to assign social credit scores, cut the disobedient out of society, and otherwise enforce the digital panopticon, serving as a digital enforcer of the Mark of the Beast - and that this system is coming very soon.
(http://www.autoadmit.com/thread.php?thread_id=5759869&forum_id=2E#49166505) |
|
|