I think calling it a trick is unfair. It's not just P(C3), it's really P(C3 | !C1 && !C2), where C1 and C2 are really big events, so the fact they haven't occur

I think calling it a trick is unfair. It's not just P(C3), it's really P(C3 | !C1 && !C2), where C1 and C2 are really big events, so the fact they haven't occurred narrows the search space quite a bit. Remember, he's not just talking about any conciousness simulations but a specific type - ancestor simulations, which allows him to pull more rabbits out of his hat than might normally be able to hide in there.