Without specific countermeasures, the easiest path to transformative AI likely leads to AI takeover

Abstract

I think that in the coming 15-30 years, the world could plausibly develop “transformative AI”: AI powerful enough to bring us into a new, qualitatively different future, via an explosion in science and technology R&D. This sort of AI could be sufficient to make this the most important century of all time for humanity.