Yeah, that's worth exploring more. It's just that it's not safe to depend on "real neurons are complicated, and therefore artificial nets of simple units won't have transformative capabilities".