(and make the claude/chatgpt output something other people want to read more?)
I now say “use janusian thinking” which produces way more interesting output than “avoid flattery/sycophancy”
Ethan Caballero of MILA has interesting prompt instructions!
[Question] What do you put to your Claude/ChatGPT system prompt to increase diversity/truthseeking/prevent mode collapse?
(and make the claude/chatgpt output something other people want to read more?)
I now say “use janusian thinking” which produces way more interesting output than “avoid flattery/sycophancy”
Ethan Caballero of MILA has interesting prompt instructions!