We now know that playing chess doesn't require human-level intelligence as Littman understands it. It may turn out that destroying the world doesn't require human-level intelligence either. A narrow AI could do just fine.

Interesting: this framing moved me more than your previous explanation.