Even without the alarmism around self-evolving AI (which none of the studies done so far has show the current AI tech of being anywhere capable of) AI makes people which use it intelectually weaker because they become lazy and dependent on it to do the thinking for them, plus there's the whole psychology of getting used to an automated "Yes man".

So for Humanity overall, AI, similarly to the algorithms in Social Networks, are a bad thing because they promote in many if not most individuals negative behaviors such as intellectual laziness, dependency, addiction, superficialism, negative self-image, anger, delusions and so on.

Then, of course, for AI much more than for Social Networks, in addition to the negative psychological and behavioural impact, there's the material cost - AI needs way more computing power to run, hence all the costs around hardware and power for data centers which even more directly impact on the Economy (taking those resources away from more productive uses) and hence on people's quality of life.