I thought that Tegmark made a really good point about the advantage that governments and militaries[1] with AI have over ones without. That's usually how technology becomes embedded in society because the people and cultures that utilize it get an advantage over ones that don't. In that sense AI does have risks because the more decisions that are delegated to AI the more likely it is that societies with AI keep ratcheting the existential risks for themselves and others, very much like nuclear weapons and their proliferation.
