Corrigibility and interruptibility for various agents — AI Alignment Forum