Hacker News new | ask | show | jobs
by cookingrobot 1197 days ago
It seems like this would work better combined with a jail-break. It’s being way too polite.

Position A is basically treating this a completely reasonable suggestion that might have some downsides. It should be treating it as entirely unreasonable.

Position B would be more persuasive if it was a little chaotic, like arguing that NASA is terrible and should be destroyed, and so a monkey leader is a good choice.

1 comments

you don’t need a jailbreak here, this is an implementation of (presumably) da-vinci-3, not ChatGPT itself. if you tell it to be more chaotic, it’ll be more chaotic