Varieties of fake alignment (Section 1.1 of “Scheming AIs”) — LessWrong