x
Do AI agents need "ethics in weights"? — LessWrong