Intelligence has to be defined relative to goals and the knowledge needed to attain them. In any case the argument against the doomsday fear-mongering of existing AI extends to more powerful systems: any system that monomaniacally pursued one goal (such as making paperclips) while being oblivious to every other goal (such as not turning human beings into paperclips) is not artificially intelligent: it’s artificially stupid, and unlike anything a technologically sophisticated society would ever invent and empower. And scenarios in which the systems take over themselves commit the fallacy that intelligence implies a will to power, which comes from confusing two traits that just happened to come bundled in Homo sapiens because we are products of Darwinian natural selection.