With very large sample sets, I believe, a normal distribution is achieved. So if you're just talking about the people in the room with you, then I agree, but if you're talking about 300m people, then that doesn't apply.
Edit: didn't see that someone had already reply'd with basically this same idea.
