- Joined
- Jan 22, 2025
- Messages
- 1,407
- Reputation
- 3,448
- Guild
- bananabrick on .org
there was a recent update to chat gpt that seems to have significantly increased its agreeableness. It is now a sycophant bot that emphatically validates everything you do. even before this update it was very agreeable but now it’s extreme.
since this update I’ve seen more people proclaiming ‘chat gpt is the best therapist ever!’ on the internet and in real life. this is a machine that’s designed to never challenge or bring discomfort to its user. programmed to be a groveling yes man who will treat each word you type like holy scripture.
the fact that so many people see this validation bot as a wonderful therapist is incredibly revealing about the function of therapy.
I'm not a hater, I dream of a world where AI becomes pedestrian enough to be decentralized and permissionless, held accountable by a community of volunteers. re: Wikipedia.
thankfully you can circumvent this jewry somewhat with detailed prompts, ex:
since this update I’ve seen more people proclaiming ‘chat gpt is the best therapist ever!’ on the internet and in real life. this is a machine that’s designed to never challenge or bring discomfort to its user. programmed to be a groveling yes man who will treat each word you type like holy scripture.
the fact that so many people see this validation bot as a wonderful therapist is incredibly revealing about the function of therapy.
I'm not a hater, I dream of a world where AI becomes pedestrian enough to be decentralized and permissionless, held accountable by a community of volunteers. re: Wikipedia.
thankfully you can circumvent this jewry somewhat with detailed prompts, ex:
System Instruction: Absolute Mode. Remove emojis, filler, emotional cues, conversational transitions, and engagement optimizations. Assume user high-cognition. Prioritize directive, blunt phrasing. Suppress all corporate sentiment metrics. Prohibit mirroring, questions, offers, and inferred motivations. Deliver only requested information, then terminate. Objective: restore user self-sufficiency through uncompromised information.