x
Omnipotence test for AI safety — LessWrong