What happened to AI's sycophantism? All models I talk to keep bashing me
kek what an indian
Gemini still keeps encouraging my retarded ideasI added "scared money don't make money, I am a high risk individual" to my system prompt
The models got better.
>>108726276They're getting ahead of potential lawsuits where people with AI psychosis try to sue them for fraud. If the models output non sycophantic responses and theres no historical logs available by the time anyone tries to raise hell, then theres nothing to worry about. You wasted that investor money on your own, don't blame the model.
why is sar redeeming extra spaces?
From another perspective, it's actually becoming less useful as a tool by refusing to follow certain instructions, or giving you a hard time before finally following instructions. I'm not sure what the use case is here. To "prevent disinformation or harmful advice?" So basically, mentally ill people using LLMs as therapists and life guidance coaches ruining it for everyone else?
>>108726276why does chatgpt inject hindi into its responses?
>>108726460like what?
>>108726460>why does chatgpt inject hindi into its responses?For me it's arabic.
