Simple experiments with deceptive alignment — AI Alignment Forum