Several people who saw me use chatgpt asked me if you need to be polite. No, but why would you be rude?
yeah, I mean the pre-training teaches it that humans tend to reply more helpfully to pleasant tone than to hostility, RHLF helps restrain it from being as reactive as a human would be but you'll still get better results by working w the pre-training than against it
More notes on nearby ideas
Reply · Mar 22, 2023@inerati haha yeah I think you gotta use a bot w less RLHF and more creative prompting to get anything fun out of itReply · Oct 16, 2024@lumpenspace lol when ive actually wanted to understand things ive had to say i was the opposite person in the conversation and chatgpt would slant…Artificial intelligenceReply · Sep 5, 2023@chairsign I think it somehow isn't... my brain is tpot-RLHF'd... I just get confused and mute ppl when they're excessively negative & aggroQuote tweet · Mar 9, 2024lol when ChatGPT has had enough of bpd distortions & decides to take countermeasures