That's especially true when doing it based on personal experience.
If you manage to do that five days a week for thirty years, you meet 1.4 million people, or 0.4% of the US population.
Obviously in most cases you'd have no idea about how smart most of these people are. You'd probably remember the ones who did especially dumb or especially smart things in the three minutes you interacted with them, which doesn't actually tell you much about how smart they are. It's even worse for that than a puzzle-trivia interview.
I think my point stands, even for a Hawaiian in tourism.
I think trying to generalize about populations on this scale from personal interactions is worse than pointless.
I have plenty of value for a reduced statement, one that acknowledges explicitly that the speaker can only legitimately talk about his experiences, not the US or Japanese populations as a whole.
He said in Japan, people are usually about the same. But in America, people are either really smart or really stupid.