"'I just had to be sure because I’m not a regular cop, I’m part of the FBI, and when I scanned your plate I could see you had suspicious travel patterns,' I think is how he worded it" [1].
This isn't traffic cops, it's federal police. And it isn't suspicious driving patterns, e.g. weaving, it's full-on movement tracking.
It's a completely lost cause. For example, when you agree to Mazdas TOS for the connected services app, you give them permission to sell your driving data to 3rd parties including law enforcement and insurance companies.
I said no so I'm not allowed to have remote start. I'm a bad cog. Punish me.
Some criminal behaviors are somewhat correlated with skin color, age or gender. It would be extremely hard to train an AI to ignore those correlations. But would it even matter? Nobody knows why the AI flagged suspicious behavior, it usually can't tell even if you are able to ask follow up questions. So nobody could prove racial profiling.
I've noticed this myself. Not being pulled over, but being followed.
When I just came to the US, I got my car and started exploring Utah a lot. Just picking a random spot on a map and going there. Beautiful countryside.
There were several occasions when a police vehicle showed up right after I parked, made a circle around me and disappeared.
One time their cruiser parked 30 feet away and they were watching me. I had a slice of pizza with me, so I rolled down a window and let them watch me eating. They drove away after a few minutes.
Also, driving 60 miles one way to a lake for a swim once or twice a week is suspicious, apparently. And I was doing it even when the water was in 40s. One time K9 parked right next to me. Idk if they searched my car or not, but their drone followed me on a trail back and forth for over 5 minutes. And when I came back, they were packing up and leaving.
Never saw them during moonless nights when I was stargazing in the middle of nowhere or sleeping in my car on the side of a road. Fun times.
Eventually your car will autonomously pull over to the side of the road, lock the doors, and instruct you to wait patiently for the arriving officer. Please return your hands to the steering wheel.
If you watch any amount of police bodycam footage on YT having to do with trafficking you will see ALPRs are now another tool they use.
The given reason for being pulled over is something petty like following too close or not signaling at the right time. The real reason is the vehicle was flagged from some pattern determined by the ALPR networks.
I've had my car searched (I finally gave consent to search after over an hour of being detained waiting on a drug dog) because of a popular 'reasonably suspicious' pattern cops use. The loser let me off with a warning for speeding after finding nothing.
why was this FBI guy so forthcoming with this lady about his agency's arguably civil rights abuses? I mean I guess it's good he was but on the other hand it seems a bit weird, given my experience with feds.
We need hard math for statistical construction of probable cause.
Probably a public model, which is independently vetted against actual detentions, needs to show a 99% probability of an actual crime being committed for police to make a stop like this.
Technology like this is stuck in a no win situation.
Human police officers are allowed to take action based on reasonable suspicion.
AI, if it is worse than humans, will be criticized for false positives.
And if it is better than humans, without being able to explain why it will be criticized for that as well even if it is in general better than humans, it will be criticized for the false positives.
But as the saying goes, this is the worst it will ever be. At some point this thing or something like it will have a precision over 95%. At what point will the debate become awkward because the thing is superhumanly efficient at sorting out the baddies?
Ironically the best defense against police abuse of power seems to be doing your own surveillance. Recording interactions with the police and having a dash cam. Many popular dash cams like Viofo record to an SD and don't send recordings to the cloud.
This isn't traffic cops, it's federal police. And it isn't suspicious driving patterns, e.g. weaving, it's full-on movement tracking.
[1] https://www.aclu.org/news/privacy-technology/ai-driven-polic...
I said no so I'm not allowed to have remote start. I'm a bad cog. Punish me.
Some criminal behaviors are somewhat correlated with skin color, age or gender. It would be extremely hard to train an AI to ignore those correlations. But would it even matter? Nobody knows why the AI flagged suspicious behavior, it usually can't tell even if you are able to ask follow up questions. So nobody could prove racial profiling.
When I just came to the US, I got my car and started exploring Utah a lot. Just picking a random spot on a map and going there. Beautiful countryside.
There were several occasions when a police vehicle showed up right after I parked, made a circle around me and disappeared.
One time their cruiser parked 30 feet away and they were watching me. I had a slice of pizza with me, so I rolled down a window and let them watch me eating. They drove away after a few minutes.
Also, driving 60 miles one way to a lake for a swim once or twice a week is suspicious, apparently. And I was doing it even when the water was in 40s. One time K9 parked right next to me. Idk if they searched my car or not, but their drone followed me on a trail back and forth for over 5 minutes. And when I came back, they were packing up and leaving.
Never saw them during moonless nights when I was stargazing in the middle of nowhere or sleeping in my car on the side of a road. Fun times.
The given reason for being pulled over is something petty like following too close or not signaling at the right time. The real reason is the vehicle was flagged from some pattern determined by the ALPR networks.
I've had my car searched (I finally gave consent to search after over an hour of being detained waiting on a drug dog) because of a popular 'reasonably suspicious' pattern cops use. The loser let me off with a warning for speeding after finding nothing.
Throw the book at her!
Probably a public model, which is independently vetted against actual detentions, needs to show a 99% probability of an actual crime being committed for police to make a stop like this.
The opportunities for abuse are simply too high.
Technology like this is stuck in a no win situation.
Human police officers are allowed to take action based on reasonable suspicion.
AI, if it is worse than humans, will be criticized for false positives.
And if it is better than humans, without being able to explain why it will be criticized for that as well even if it is in general better than humans, it will be criticized for the false positives.
But as the saying goes, this is the worst it will ever be. At some point this thing or something like it will have a precision over 95%. At what point will the debate become awkward because the thing is superhumanly efficient at sorting out the baddies?
-Larry Ellison