Safe AI from question-answering - History — AI Alignment Forum