What happens when you train a great AI with propaganda or one ideology?

I think it would generalize the inherent logic of said ideology and apply it to everything. The problem with AI is that it still cannot really think. It can apply formalized knowledge to extrapolate a conclusion that seems likely within this ideological logic. This even has a self-reinforcing effect, although even if that were not the case any modern AI is unlikely to be able to look outside its box.

The result would probably be the opposite of what we would like to gain from AI, that it can improve our blind spots. It would just reiterate the premises it has been fed with and look for any and all correlations that could confirm these believes.

Like humanity, AI can't be black and white. It has to have subtleties, idiosyncrasies, situational awareness, context, subtle cues, body language, etc

Baking in these is a hard problem, of course, but worth the challenge if it means we don't get another M$ sponsored Tay Bot[0] on our hands.

[0] https://en.wikipedia.org/wiki/Tay_(bot)