Selfishness, preference falsification, and AI alignment — AI Alignment Forum