Probably more like "arbitration by AI" as a step in "reforming" plea bargains. I think the real questions are 1. whether the AI would have to explain its reasoning, and 2. whether you would get to appeal to a regular court.

Probably never. Jury trials are enshrined in the constitution.

And people don't trust AI and don't like being judged by an algorithm.

Not my experience at work. Our new developer hires have now a year and a half under their belt and still don't understand what they do. They are now able to spot style inconsistencies so there's that, but still have no understanding of our WAN, what is software defined and what is not.

Also, our management explained to us two weeks ago how 'Access& Identity' will help us be more productive, in a slide, and read that out loud without batting an eye. I get that management often isn't the most clever, but this is a degree of stupidity that wasn't present two years ago I feel.

Don't get me wrong, AI was great for me, it's a very good tool, but you can't tell me you understand as much of your code as you did a year ago.