20 Critiques of AI Safety That I Found on Twitter — LessWrong