Mainly because a great deal has been improved and contains been kept in order to be utilized with no adversity.
To abuse power, you have to have it. At this point we are in rule. But rule is seldom meant over oneself, but rather over others. Mastery over others inevitably means that those have less mastery over themselves.
The state that every single person can have maximum (not absolute) control over himself is the goal of anarchism. That means first of all freedom from external domination within the rules of the community. It also involves developing self-mastery, which is the second and no less difficult part of the concept.
If humanity could come closer to this level of maturity and eliminate many unnecessary problems with it, the effort would be worth it.
First of all, the term anarchy has to be removed from the permanent denunciation, thanks to Ilija Trojanow.
In schools there is exactly ZERO about anarchy / anarchism, neither in social studies, ethics or history. The most explosive concept for any system of rule is simply not discussed.
The biggest problem with AI police is that nobody is responsible anymore. The police can only say, well, that our AI classified the shoplifter as a terrorist and our robocop terminated him, we can’t do anything about that, we can’t see through.
The second biggest thing: The programmers (whose job I don’t want to have) trained the AI - it’s not programming in the classic sense in which rules and processes can be coded and corrected, but a switching network is “trained” with patterns. Of course, not with patterns that the programmers come up with, but with real case data from the past, which transports all the prejudices and previous habits of the police officers involved 1: 1 into the new system. And then you get a racist imbalance out of a robot. They have to be retrained, ie fed longer with different patterns so that they weight differently. These patterns are not easy to make up, they have to be consistent so that they halfway represent reality and the entire associated context. And of such cases / scenarios it takes very many, slightly different, to bring an AI to “change behavior”. But they have to match the other patterns, otherwise you will get effects that cannot be overlooked.