Reading: DeepMind’s 145-page paper on AGI safety may not convince skeptics