Value Uncertainty and the Singleton Scenario — LessWrong