x
Layered Reward Modifiers for Transparent and Self-Correcting AI — LessWrong