Normalcy bias and Base rate neglect: Bias in Evaluating AGI X-Risks — LessWrong