Earlier quoted context omitted.
My first gut reaction was $115k/person/year was way too high... But apparently a person in a NYC prison costs $560k/year[1]...That amount of money can definitely do a lot of good to rehabilitate a person. However > we identified people likely to become criminals This tricky, because people will think it is GOOD to marked by the algorithm; key free money! Then these algorithm will likely be used in court as evidence t…
So its a thought experiment and so this AI is very smart, like even better than Microsoft's Tay. And maybe it not just on the person's behavior but also family wealth, living location, history, convictions of family members, and so on. Lots of things that are hard to manipulate or would take generations to do so. If you are convicted of a crime even if you had this free money, then likely you are doing crime for fun…
But more seriously, even if the AI was super smart, it's still limited on it's inputs. The factors you listed are how the current crime AIs determine likely hood the person committed the crime.
This AI is very smart and people trust it. If it says you committed the crime then you must have done it!
> If you are convicted of a crime even if you had this free money, then likely you are doing crime for fun and not out of necessity, so jail is the best place for you.
> Cities may actually decline to convict because jailing people is more expensive, so maybe this would be a counter-effect.
These are exactly my points. Being marked by the AI gives you money, but people won't trust you.
If you're on trail it's this big weight people want to hate you. The jury isn't likely getting the money, why should this person on trail get it.
The judge has an incentive to incarcerate.
Even as a thought experiment, this idea can not stand.