"...I don't think it will be sunshine and rainbows."
The past 5 years have been an evolving democidal attack on humanity by pathological banksters using full spectrum dominance to comprehensively deceive and destroy as many people as possible. Billions of people have been harmed by malicious medical interventions, tens of millions have been killed, and many of the vectors of iatrogenic harm are just beginning to eventuate. The primary means by which this violence has been effected is deception, and I am confident it isn't over. While media bleat deceptive narratives, governments prepare to diversify the range and lethality of the harms they are imposing.
You are very right, IOW. People that aren't preparing to survive existential threats across every possible vector are preparing to die. Because deception is the primary mechanism being used to prevent us, the intended victims, from successfully surviving, depending on commercially deployed AI is almost begging to be deceived. FOSS AI privately held will enable the utility of such algorithms to be controlled by folks, as long as potential back doors are precluded. Little could be worse than for the software of our production management systems to suddenly turn out to be controlled by our enemies, as Iran experienced when Stuxnet destroyed their centrifuges.
As is always true, deception is the worst threat we face, and understanding the actual situation and nature of the tools we use to mitigate threats is the critical hurdle we must overcome to even stay on the field of battle.
It will be tough for sure. Deception can lead vulnerable people really anywhere and AI can do that at scale.
AI will do a lot of good things too. It already is increasing efficiency in a lot of ways in various industries.
I do think it will make the world a better place but people should be careful the extent to which they interact with it in a social way. People should avoid thinking of it as friend.