You can't take an empirical approach to existential risk as you don't get the opportunity to learn from your mistakes. You have to prospectively reason about it and plan for it.