Because you can't do the Nazi Germany thing these days. I mean... disgust aside, it kinda failed. But you can spy on people under "national security" while keeping them feeling happy enough. And that arrangement can last 1000 years.
Most of Nazi Germany is after the fact revision. They were popular around the world in the 1930s - for their plan to deal with the Jews. It is only after they went to war that we decided they were bad for that plan as well. (some people were opposed to the plan all along, but there were plenty who were in favor of it)
> Because you can't do the Nazi Germany thing these days. I mean... disgust aside, it kinda failed.
It failed because Nazi Germany was not militarily superior to combination of the nations that it got upset with it externally, not because of any internal failure of control. While its nice to think that Nazi Germany “failing” somehow disproves the viability of the same broad kind of one-party, massacre-the-opposition totalitarianism, it isn't really justified.
Still not convinced that AI is offering anything new here. Especially when the statistics you'd reach for are often like 100 years old or more. Bayes theorem is older than the united states. I think among lay people there is a lot of conflation between AI and statistics, and also a lack of understanding of the state of that field and how mature it is. Nazi Germany of course heavily used statistical modeling and even contracted with IBM to quantify Jewish populations.