It does seem unlikely that an "expected utility maximizer" reasoning like this would manage to build interstellar spaceships, but that insight doesn't automatically help with building an agent that is immune to this and similar problems.