Are American cities often quite different when it comes to laws and how justice is enacted?
I am curious, because I wonder if this could prove to be a differentiator between cities. Perhaps you will end up with "Law and Order" cities, where technology is used unempathetically to punish harshly and discriminate without penalty. Then you would have less strict cities, where freedoms are less trampled in the pursuit of total control.
The states are highly autonomous and the laws vary most at the state level, but they vary from city to city within each state as well.
You can already see how this plays out: the protests in Portland were tolerated for a very long time, had they happened in almost any other city I doubt the local officials would have tolerated them for as long. (No comment on whether or not that’s a good thing.)
Yes, they vary wildly. For example, in a lot of rural areas the only law enforcement is a county Sheriff. There's often a "good old boy" system where people that know the sheriff can do whatever they want, driving drunk is a typical example.
Everything is potentially a different system from how jail works, bail amounts, maximum penalties, typical plea bargain offers, what laws actually get enforced, etc.
Does anyone live in a rural place with only a county sheriff? I live in rural ND and there is more law enforcement than that. Also, corrupt sheriffs have been run out of town.
Well, yes, there's technically things like State Police, but they aren't physically around much. The corrupt sheriff problem is rampant in Oklahoma for sure. I know that civil asset forfeiture laws made it worse. Since they are elected positions, it's difficult to get rid of a bad apple because they have more money than anyone else to run campaigns.
I used to, and to my knowledge, that place is still like that. If you lived outside the city limits (which is really easy in a city of 3000 people, but a large county overall) then the only police who would respond to calls were county.
Some towns share the cost of policing by letting the sheriff's office handle law enforcement. The county extracts more money from taxes to do this but there is less duplicated effort overall.
> Perhaps you will end up with "Law and Order" cities, where technology is used unempathetically to punish harshly and discriminate without penalty. Then you would have less strict cities, where freedoms are less trampled in the pursuit of total control.
This is already true. People are free to use marijuana/brothels/certain guns/etc in some places and not in others. Some cities have license plate readers on their police cars and roads and red light/speed cameras. Some prosecutors go after certain crimes and not others.
Although, I’m sure in every place there is little transparency so if you know the right people, you can get away with a lot more.
That process has already happened to a large degree. We can contrast New Orleans with cities like Austin and San Jose, which use all of the technologies New Orleans just banned (except facial recognition) and have much lower crime rates to show for it.
It’s a multifactorial process. Low crime, strong policing, and high average income are all correlated; I can’t claim I know exactly how the causation goes. (Gentrification seems like an unlikely explanation for Austin, though, since cost of living is pretty close to New Orleans.)
San Jose's infrastructure isn't as high tech as you seem to assume. Heck, many downtown streets are on fixed pedestrian cycles rather than controlled using magnetic loop detectors, for one example.
When somewhere as famously corrupt as New Orleans bans this stuff there's a part of me that can't help but wonder if it's due to the corruption, not in spite of it.
In other words the people at the top are worried that they'll get caught up in ways that can't be hidden or denied (because it'll be outside contractors running these things most likely) and so they want to make sure it never comes to pass.
I wish I wasn't so cynical about this as I do think the ban is a good thing in general.
That's certainly a concern, though New Orleans is a city with quite a lot of people of color in leadership positions. So the leaders are familiar with how those tools would most likely be used in a disproportionate way against minorities.
This assessment makes sense, many large Southern cities have this same demographic and hence public leadership profile. This is a strong step that may signal other cities (if not already: maybe Atlanta, Charlotte) to follow suit. That I think would be a boon for the future because of the ways such technology can be used against minorities as you point out.
The previous mayor, Mitch Landrieu, was letting Palantir test their systems here. There are ANPRs at most entry points into the city. I personally have zero surprise at Gebru's freaking out after having lived here.
It's sort of assumed that all people of color experience the various injustices and systemic racism we still have in similar ways, and so they'll all collectively "understand" this sort of thing better, as if they're all interchangeable with one another in some sense. It's a bit repugnant.
A similar sentiment is expressed with respect to politics in some circles: as if people of color have more political acumen and agency than others by virtue of the racism they might have experienced.
Assuming knowledge or opinion based on skin color is a very simple example of racist prejudice. Assuming that a person being black means they must have certain experiences and ideas is wrong, even if the assumption isn't negative.
Yes, things like "Asians are good at math" is racist, even if it's a positive stereotype.
That said, the parent poster said that many leadership positions, plural, were filled by people of color. I think it's probably reasonable to say that a community of many political leaders of color are more familiar with systemic racism. These are folks who specifically choose to get involved in politics and leadership, and likely represent fairly large constituencies. It's not unreasonable to assume that they broadly have experience or are able to communicate with their communities that have experience with systems that negatively impact people of color.
So I do not really understand this. (To me having racial categories is racist already, but this seems to be fine/non racist in the US). So would a statement like 'Asians perform significantly higher (by so and so amounts of standard deviation) on Math tests than Caucasians' a racist stereotype while still beiing a fact?
It's not racist to explain why the majority of your quants are Asian. It is racist to hire a quant because they're Asian. In politics, people tend to do the latter.
There's a big difference between, "These are the observed results" and "These people are inherently X". And especially different than, "Person is Y, therefore they are good/bad at X".
The first statement is open ended, it keeps the door open to systemic factors outside of race that might correlate with race. It also doesn't suggest that everyone of a particular group is better/worse at something.
The second statement says that there's something intrinsic to that everyone of that race. It sets up expectations for a group of people. Maybe you are an Asian who isn't good at math? How does that make you feel? How will you be impacted by other people making assumptions about you?
The third statement is applying stereotypes to individuals, which is just never going to be a good idea.
The differences between the first and second statements are subtle, but important.
So then you are saying the statement 'Asians are good at math' is not the short version of 'we observed that asians are better at math then other races' but the short version of 'Asians are inherently better at math' and therefor racist?
The generic "Asians are good at math", in this context, is probably extended as "We observed that certain groups of Asians are good at math, therefore, all Asians are good at math."
And as parent pointed out, it's the leap from observed results to expectations about a racial group (or many groups in this case) that makes it racist.
The US writ-large tried the 'color blindness' approach for a long time. The general effect was that shit kept rolling downhill, but that those in power were not expected to discuss or do anything to improve the situation: the heavily racist outcomes of the war on drugs, racially disparate education funding+outcomes, etc etc etc were off-limits for 'polite' conversation. The current thinking is that you have to acknowledge race to fight systemic racism.
i also think there is a general concern that 'acknowledging race' is what led to racist outcomes being systematized in the first place. the current thinking also seems to struggle to concretely articulate what it's goals actually are, is success defined as proportional representation across interests and industries? is it 'majority minority' in leadership roles? will these mandates be adjusted as outcomes and demographics predictably shift? and maybe most concerning, are all problems considered on equal merits or is there antipathy and an underlying belief in deserved suffering as atonement for intergenerational guilt?
> Yes, things like "Asians are good at math" is racist, even if it's a positive stereotype.
So what would be the proper way to describe an ethnic group that, when examined purely statistically, really has performed well above average in mathamatics?
Do we simple never acknowledge that such differences exist, as some people have signalled?
> So what would be the proper way to describe an ethnic group that, when examined purely statistically, really has performed well above average in mathamatics?
Exactly the way you said it. "Statistically speaking, controlling (or not controlling) for relevant factors foo, bar, and baz, group X outperforms other groups such Y, and Z in tests of skill A". Which isn't the same thing as "X is good at A". "X is good at A" is a meaningless statement. It's devoid of context and prone to overgeneralization. It's not useful for anything.
Sports might be useful. Do you see the difference between,
"The Patriots have statistically scored more yards per game than average." Vs.
"Patriots team members are good at scoring yards." Vs.
"You are on the Patriots, therefore you are good at scoring yards"
In aggregate, the players perform in some positive or negative way, but you can't just apply that statistic to any individual player or even say that the players are unusually good at that thing. Maybe they have a great coaching staff, maybe they are cheating, maybe they have a few star players who push up the statistics.
It's such a common assumption that when you imply that the children of wealthy African immigrants might not have any more insight into the culture and experiences of black US descendants of slaves than the median US white person, you're reflexively accused of being racist. It's racist to believe black people are fungible. We wouldn't expect the child of an Italian diplomat to give us insight into the experience of growing up on Staten Island.
edit: tbh, it's usually a pragmatic belief. It's easier to convince upper-middle class black people that what you want to do is good for black people in general (because their culture largely overlaps yours and they share your values), and wealthy black people and average black people look the same in pictures.
edit2: I hate to come back to this, but just look up Ray Nagin and his response to the hurricane that ultimately washed all of the black people out of New Orleans for an idea of the character of a lot black leadership in the south, and its total disconnection from any working class awareness. Just looked him up now to see what happened to him since and was not surprised.
Does Ray Nagin really support your point? From what I know he was born in New Orleans with working class parents. Also the demographics section of the wiki page for New Orleans doesn't seem ti support your claim that Katrina washed all the black people out of the city. It was 61.9% black in 1990 and 60.2% black in 2010.
I'd agree if I'd made the statement about one specific person. The idea that more diversity of race across a leadership group is more likely to include diverse experiences, to me, is a different type of assumption that would almost always be true.
Hypothetically, a group of people of different races grow up in the same small town in the same economic conditions, go to the same schools etc. Is race still a good indicator of diversity of experiences? Not at all.
Skin color is a poor proxy for diversity of thought or experience. We could instead just treat people as individuals and not draw broad conclusions on skin color.
> Hypothetically, a group of people of different races grow up in the same small town in the same economic conditions, go to the same schools etc. Is race still a good indicator of diversity of experiences? Not at all.
Even more so in that scenario, I'd think.
If you've eliminated all the other variables, race would be the only differentiator of experience in that grouping.
(Of course, the reality of the non-hypothetical world is that people of different races tend to also have different schools, economic conditions, etc.)
To clarify your position, in the hypothetical, you believe two people with the same color skin will have identical experiences because of their race?
I don't think that's the case at all. Again, skin color is a poor signifier for experience, and while it's good to acknowledge that person's life is impacted by their race, thinking that race is the only differentiator is silly.
Individuals are determined by more than their level of melatonin.
> Again, skin color is a poor signifier for experience...
I suspect people of color would tend to disagree. It's likely a black person will have experiences not shared by your average white person, just like Americans may have experiences unique to them, or men, or disabled people, or people with any other characteristic that distinguishes them from others.
> thinking that race is the only differentiator is silly
No one said only.
> Individuals are determined by more than their level of melatonin.
Sure, but your hypothetical was "what if those things were gone?"
This argument is entirely based on emotion. Even if it's true, which I think it is, there is no objective evidence that black leadership in police leads towards better policing for black citizens. Thus, people only apply this logic when it fits their worldview. The same people who celebrate black leadership in police will call these same departments "systemically racist" when something bad happens, which is exactly what happened in Minneapolis.
The comment about people of color in positions of power is automatically going to be countered with comments reflective of racial anxiety, as you've noticed.
You didn't suggest a generic "more diverse leadership team," though, you stated something explicitly about the specific black leaders in NO.
Regardless of that, it actually isn't clear that a more diverse leadership team brings better understanding. They may bring a diversity of understanding that is important to consider--it's not "better", in the sense that without their participation others would not see the potential downside.
Police state activities don't adversely impact just those in minority populations.
Why assume that these technologies would increase corruption rather than reduce it? Outsourcing work to an arms-length tech firm as opposed to potentially corrupt officers could be helpful in this case.
Isn’t reducing crime in one of the most unsafe cities in the country more helpful to POC living there than making moves to dodge accidental algorithmic bias? What are the expected costs and benefits here?
Is your cynicism based on evidence specific to this event?
Because this kind of cynicism is self-fulfilling: if you don't believe anything good ever happens, then nothing ever will. So I feel it's important to actively avoid defaulting to it.
My anecdotal evidence: I just had a conversation with someone who was involved in the push for this to happen and at least that person (and purportedly, the organization they're involved with) is an idealistic programmer, and not a corrupt cop.
I developed a facial recognition access and timekeeping system a few years back, people refused to use it so it never got past the pilot phase.
Be warned.
I think you're just ahead of the time at which this will become acceptable. In the early 2000s when the UK started deploying cameras everywhere, the end of the world was predicted.
I wonder how this all relates to license plate readers. As a frequent cyclist/pedestrian, I've felt for some time that a low, consistently-applied automatic fine would go a long ways toward eliminating traffic behaviours which are mostly low-incident at present because vulnerable road users have trained themselves into hyper-defensiveness and basically know to jump out of the way. To some extent, this is even in the official messaging, with campaigns about how road safety is a "shared responsibility" and so on.
Basically, things like speeding in neighbourhoods, entering a crosswalk before all peds are off the roadway, running red lights and stop signs, right turn on red without coming to a full stop first, etc.
Currently in my city these things are all pretty much completely unenforced— the fines are in the hundreds of dollars are and only applied when someone is actually hit. This breeds a driver mentality of "well, it doesn't matter what I do as long as I don't hit anyone" which in turn makes the roads a less safe place to be for everyone. The fines are where they are partly as a deterrent, but also because the cost of enforcement is so astronomical when it requires multiple personnel on a special operation.
I feel that if the fine was in the $10-30 range, but it was applied consistently on every single offense, drivers would quickly train themselves out of doing these things. (This is the swift-certain-fair philosophy, see: https://nnscommunities.org/strategies/swift-certain-fair/)
Anyhow, I know that plate != driver, and there's a bunch of existing case law around protecting drivers' "right" to get away with endangering themselves and others by disregarding traffic rules. But given how much of a step beyond plate-recognition a facial-recognition system would be, I'm interested to speculate if there might be more of an appetite for this kind of thing down the road.
Do you think such comprehensive enforcement would be limited to traffic violations? While you might not want it, you're advocating for total surveillance, and severe authoritarianism.
I think a lot of it depends on the culture and oversight around it. For example, most of the developed world has gun laws that many Americans consider to be a stepping stone to authoritarianism, and it seems to be working out okay so far. If police weren't already an established institution, it's easy to imagine opposing setting them up with an argument similar to this ("we're really just going to put a bunch of armed people on the street with the authority to detain anyone for any reason? Isn't this authoritarianism?").
In any case, the wheels are already in motion for this to happen, so I suppose we'll find out soon enough where it leads to:
The thing about firearms is that, unless you're actively hunting, you don't need them - until you do.
You don't need them until there's a home invasion. You don't need them until someone tries to rob or kidnap you or a member of your family. You don't need them until you are, like in Venezuela or the Czech Republic historically, in opposition to a dictatorial government. There is a real reason that every single authoritarian government in history has disarmed its populace.
Vigilance against authoritarianism is the same thing, because authoritarianism rarely happens over night. It takes years, sometimes decades. And if you've built up a network of cameras powered by facial recognition on every street corner, you've now built the apparatus for an authoritarian-leaning government to harness.
> You don't need them until there's a home invasion.
I would argue that this also is a uniquely American perspective— for me (an urban Canadian), the thought of starting a firefight with an intruder is horrifying, especially once the possibility of accidentally firing on a child or pet is accounted for. I'm also deeply uncomfortable with the numbers around what "gun in house" means for suicide and domestic disputes (even if I know I'll always be level-headed, what about down the road when my kids are teenagers and have friends over, etc etc).
Obviously, part of the guns-make-us-safer story is that criminals don't know which houses have them and which don't, and as a passive effect that's difficult to really study. But for a lot of us on the outside of this mentality, it's hard to escape the feeling that the safety provided by a gun is kind of like the safety provided by driving an SUV— mostly illusory, and comes at significant costs both to self and others.
> And if you've built up a network of cameras powered by facial recognition on every street corner, you've now built the apparatus for an authoritarian-leaning government to harness.
I would argue that this is a distraction— that level of surveillance already exists, but it's away from public scrutiny and is managed by Facebook and Google.
The clear and current authoritarianism threat in the US is the 106 house Republicans who supported Trump's absurd Texas lawsuit, that the leaders of one of the two major parties are unwilling to clearly speak with one voice to shut this nonsense down. And I'd be happy to be mistaken on this, but my impression is that gun owner clubs are mostly on Trump's side— those who are talking about taking to streets wouldn't be doing so to ensure a peaceful transition of power, they'd be doing so because they believe conspiracy theories about a stolen election.
I'm a rural Canadian, and the idea of not having firearms in the house horrifies me. As does the constant push by urban Canadians to disarm rural folks (like me) who legally own firearms.
It's ultimately your choice, but I've watched a lot of video on documented home invasions, and a few things are clear:
a) Anyone willing to invade your home while armed will have no compunction about seriously injuring or killing any of the occupants;
b) The defensive use of a firearm, while largely not reported, is an amazing deterrent in probable 95% of attempted home invasions where the homeowner has a firearm. Most intruders will nope out as soon as they see a gun, no shots necessary.
c) It happens in Canada far more often than you think, including in cities. Here's a recent case: https://torontosun.com/news/crime/violent-home-invasion-may-.... Even if this case is gang-related, what are the odds that these people make mistakes and enter the wrong home/apartmnet?
d) In rural areas, police are not getting there in any reasonable timeframe. It would take police probably 30 minutes to get to my house.
Ultimately, my view is that the right to self-defence is meaningless unless you have the means with which to actually defend yourself. I have no problem with people who are uncomfortable owning firearms and choose not to, but it's very problematic when those people force their views on others.
It's not legal in Canada to own firearms strictly for the purpose of self-defence (because our gun laws are quite silly once you become familiar with them), but most rural folks hunt or sport shoot so it's a moot point.
Oh, and by the way - meta-studies on the topic indicate that restricting access to firearms doesn't actually prevent suicides, it just causes people to shift methods to hanging, overdose, etc.
Yeah, I should have clarified: I’m not necessarily for disarming; I think the Canadian laws are fine the way they are, and people who want them and are reasonably fit/trained are welcome to have guns. I just never will and am glad my kids don’t visit friends’ houses where guns are a thing. But my point is that the Canadian laws are basically the reforms that are being asked for in the US and are described by opponents as totalitarian.
"Moreover, studies from Canada, New Zealand, and Australia (at least for the first post-NFA years) show that observed reductions in firearm suicides, after the implementation of these laws, were compensated by substitution methods that resulted in no significant changes in overall suicide rates."
Canadian gun control laws are increasingly bordering on totalitarian. Given the general level of ignorance on firearms issues among the majority of Canadians who don't own firearms (and don't know much about them), I expect the laws to get much, much worse than they are even today. Even though I have been thoroughly vetted and done training and been tested, there are about 1,000 arbitrary and stupid ways in which I can be criminally charged for things that are not actually dangerous.
The recent arbitrary ban on "scary" semi-automatic rifles is a perfect example of this point. Because many of these rifles are classed as "restricted", and thus registered along with requiring more onerous licensing requirements, they are trivially easy to confiscate - this allows the government to score a political victory by appearing to be tackling gun violence, while ultimately threatening with criminal punishment Canadians who have attempted to be insanely diligent in following the law. The story that these guns have no purpose other than to kill people is simply a lie, a brazen one at that considering that the Prime Minister himself has owned such guns over the years.
Long guns, particularly legally-acquired restricted long guns, are used in homicides so rarely that it wouldn't be statistically significant. Legally-acquired handguns are also almost never used in crime or homicide.
Guns are also an important part of Canadian culture, and historically were a symbol of self-sovereignty: the ability to put food on the table and protect your family are simply part and parcel to living life on the frontier. Many in the city have apparently forgotten this in a sea of pleasant conveniences, but it is still a way of life for those of us who have chosen a life of self-sufficiency. (I used to live in a big city, btw)
Honestly, given the statistics on the prevalence of firearms vs. their use in violent encounters in this country, the fear many Canadians have of firearms is highly irrational.
I remember the TV series SeaView or such where license plate readers were on highways and other roads and would fine you for speeding, deducting it automatically from your Social Security account.
The issue of course with your suggestion is these types of fines are highly regressive as it effectively allows the well off to break the law at will.
Percentage of wealth or income type fines would never float as you would end up making a whole new process where the police would only target those who they think will have the biggest payoff. Let alone it would require way too much of an invasion in privacy
The issue of course is how does such a system know who is driving without also combining facial recognition?
> ... these types of fines are highly regressive as it effectively allows the well off to break the law at will.
There will probably always be a segment of the population which can afford to buy their way out of trouble, but that's already the current state of affairs.
I don't think it would have to be a requirement for getting this off the ground, but if down the road it was found that a lot of people were treating the fines as pay-to-play, you could have them escalate. It would be easy to do something like the toilet-tank model where you get X number of base-fee offenses per time period, and then the price increases significantly. Lots of parking regimes do this kind of thing to discourage people from treating the fine as a fee.
(And the idea with the "swift" part of the philosophy is that you don't have to wait until the end of the month to get an itemized list of offenses— it texts you or something right away, so you get that immediate feedback about what you did and what the penalty was.)
> The issue of course is how does such a system know who is driving without also combining facial recognition?
Some existing systems have a camera positioned such that it can read the plate and also snap a picture of the driver through the windshield, so your ticket comes with a picture. In other places, the law has been amended so that the the person to whom the vehicle is licensed is liable for automated fines associated with it regardless of who is driving— Toronto started rolling out speed cams under that framework just this past summer (it's fines only, no demerits, so there should be no insurance impact).
> Percentage of wealth or income type fines would never float as you would end up making a whole new process where the police would only target those who they think will have the biggest payoff. Let alone it would require way too much of an invasion in privacy
It's both possible, and already in place in some countries.
Turns out basing it on income isn't much of an invasion of privacy because the government already has that. Even in the US we submit an annual "here's how much I made" document to the feds.
Given how much this behavior can vary from region to region, I think a lot of this behavior can be driven by culture, road design, and how drivers learn to drive. It may be easier to address it through those channels.
It's easy to say "This tech is dangerous and shouldn't be used.... except in this one scenario which coincidentally is a personal pet interest of mine."
Cyclists want to use it against drivers, I'm sure shop owners want to use it against robbers, and doubtlessly paranoid parents want to use it against child nappers. Why should exceptions to a ban be carved out for anybody's pet interest?
I think this line of reasoning is dependent on your perspective. The current reality is that most places in the world have loads of private- and publicly-controlled/owned cameras observing public spaces for various reasons, and these cameras are often called upon to assist with apprehending and punishing wrongdoers.
So is it me asking for an exemption which would help law enforcement protect me and my family as we get around on foot or by bike? Or is it the drivers asking for an ongoing exemption from being asked to actually follow the laws which are supposed to apply to them?
The thing about predictive policing (and a lot of similar AI/Machine Learning/Big Data/InsertSalesBSHere) is that if a computer is doing it, you can at least see what it's doing. When a a bunch of human police officers arbitrarily decide to over-patrol black areas for no other reason than race its harder to catch and correct...
We'll likely just end up with both. We'll end up with most of the facial recognition tech in black neighborhoods, because they police will decide where to put it.
Also, if there is higher prevalence of crime in certain areas, it makes sense to do more policing there. Now it is illegal to detect higher prevalence of crime?
I agree. The issue comes when you get a feedback loop: the other neighbourhood has zero crime because there are no arrests because there are no police there. Or if the high crime neighbourhood is actually only high crime at night so the police sent there during the day have no "real" crime to prevent and spent the whole time giving out tickets. Then residents get upset.
I don't see why it's so hard to vary policing levels based on local demand as well as need. If neighbourhood Xs representatives say they want fewer police, let them have it.
The point is to be transparent and precise with your models and then override them when that's what residents want.
Of course, there is also the issue of towns raising money from ticketing, but that's another point...
The problem with issues like the ones you describe is bad management. Like they may reward their police officers based on the number of arrests or "tickets", not on actual real problems solved.
And the data issue is there whether you allow "AI" to analyze the data or mere mortal humans have to analyze it.
Advantage of AI may be that you can fix mistakes ones and they are fixed forever (like, let's say, an unjust bias against black people), whereas humans tend to fall in the same traps over and over again.
I know I'm gonna get downvoted for this, but gonna say it anyway: it's one thing to use facial recognition to catch people with an outstanding warrant, another to use it for a jaywalking ticketing blitz.
And implicit in your post is that this system can't be designed to function well at all. A logical view would be that if something isn't "perfect", you build around that imperfection and use it for its good parts whilst mitigating and eliminating the bad parts.
No, it's explicit in my post. A facial recognition system cannot be designed to function well. A logical view says that anything short of perfect will absolutely ruin people's lives.
No, it will improve society overall because it will lead to significantly higher accountability and consequences for criminals who are caught. There is no logic or evidence to your claim that “anything short of perfect will absolutely ruin people's lives”. We already live in an imperfect world. It will never be perfect. Facial recognition with humans in the loop will lead to a far higher ratio of bad people facing consequences to good people incorrectly identified. It is a net improvement for society and it is much more reasonable to claim it will “absolutely improve people’s lives”.
If you get falsely arrested (not even convicted), you can lose your job, get deported, spend thousands trying to clear your name, or even just get shot by the police.
AI tech will have false positives.
Technology that allows officers to screen many many people at scale will result in false positive arrests (or police violence) that would not have happened otherwise.
Therefore, this technology will ruin lives. This is clearly logical.
You might be ok with that cost, I'm not. But that's a different line of debate. I believe the idea that it is a net improvement is the idea that is unbelievable and unfounded given the current trends in AI. What gives you confidence?
Again there is a human in the loop to validate a potential match, before officers get dispatched. And if the dispatcher doesn't do it, we can easily just have the police officer responding to validate the match between the surveillance photo and a known photo of the suspect. Humans already have a false positive rate as part of existing policing procedures, and facial recognition doesn’t make the rate worse as long as it retains a human validating the AI’s match.
Also, almost every state has a process to expunge arrests even IF you were convicted (of something minor).
As for why I’m confident about net improvement - it’s because the amount of true positive arrests (and subsequent convictions) will be far greater than false positive arrests. Therefore relative to the status quo it’ll be a net positive of criminals facing consequences, victims being compensated or getting closure, subsequent reduction in crime through deterrence, etc.
what about an incorrect flagging being used as evidence in a case in order to prove that a suspect was in a given location around the time that an offense occurred? do we render the entire data set as inadmissible in court for the risk of such misclassifications?
>anything short of perfect will absolutely ruin people's lives
this is more or less the basis for all thought within our system of justice.
If a person is incorrectly identified as the target of an open warrant, then as soon as he shows proof of ID, he's free to go and the system is corrected to flag him as a false positive, not only never bothering him again for this search, but probably making him less likely to ever be falsely picked up again. If we can't assume that any public system or institution is capable of functioning well or being corrected, we might as well have no law enforcement whatsoever.
What will happen is the police will consider the person dangerous, approach cautiously or aggressively, cuff them, throw them into the back of a cruiser and then maybe check their id.
And you are assuming that all people have ID on them. Or are even here legally.
"Show me your papers" isn't better because a computer told the cops who to target.
Humans cannot scale the same way AI can. AI can easily scan every face at a football game. Humans can't.
AI could easily scale to watch everyone who walks through every intersection downtown. Humans can't.
If AI makes a much smaller percentage of mistakes (which is itself a questionable thesis), it will still make orders of magnitude more mistakes because it will be making orders of magnitude more evaluations.
Also, for better or for worse, people tend to trust that computers are 'logical' and correct. This is especially troubling because a human can explain their mistakes ("the person was wearing similar clothing, had a limp, etc."), while AI generally can't give you a reason that you can easily understand.
> Computers are only a tool. Humans have to make the final decisions. Under those circumstances, what exactly is the problem?
Humans are bad at evaluating computing systems. Have you heard the stories of people driving their cars into lakes because a GPS told them to? Now imagine the line was even fuzzier ("We think this could be our suspect, but we're not sure. The computer positively ID'd them though...")
well, sci-fi and everybody else was warning humans about that since like mid-20th century, and yet the human race is as unprepared as one can be (similarly to the climate change situation), and still puts the major hope into the "putting the genie back into the bottle" strategy which is obviously not going to work... Kind or reminds of the lockdowns as a strategy to fight the covid :)
Even if the AI identifies many people, there still have to be humans checking them out. The limited number of humans may have more time to check each suspect out than they would have without the AI, because the AI takes over some of the work.
Humans can be aware of the limitations of "AI". It's very rare that somebody drives into a lake because of GPS failures.
You and I have different opinions about how humans will interact with complex computer systems.
I believe humans will go, "well, the computer says you are so and so, and it's almost always right, so I'm going to need to take you in."
I've called customer service call centers and been told, "well the computer says...". If you think that the police would operate any differently, I don't know what to say.
Communities like police are poorly trained, minimally invested, and operate without oversight. They are typically overworked and just trying to close as many cases as they can.
Heck, even well educated people fall victim to this. We just had a front page story about administrators failure to deploy covid vaccines to frontline medical staff because "the algorithm made a bad choice".
The call center people use the computer to make excuses.
And you overlook the possibility of human error without computer involvement. The classical "a black person mugged me" comes to mind.
At the end of the day it is a tool. Law enforcement should be done by professionals anyway. There are a million ways to get it wrong. At least with computers, you can standardize and enforce some good practices.
All facial recognition implementations in US policing have a human in the loop to verify the match. It introduces no more error than already exists when we use human judgment to spot and catch suspects by random luck using beat cops. This is better because it narrows down the search for suspects and uses limited staff efficiently, leading to an overall safer society. All this fear mongering about false positives is irrelevant because there is a human in the loop.
> And you are assuming that all people have ID on them. Or are even here legally.
Sorry but almost everyone does have ID on them at all times. It is unreasonable to design our society for those who don’t. As for those here illegally - well perhaps they shouldn’t be here illegally. Again, we shouldn’t design our society that is based on laws around them.
> If a single person is incorrectly identified by a system design to catch warrants, the system has failed beyond redemption.
I disagree. Any large scale and complex system will be imperfect. Your argument is basically “letting perfect be the enemy of good”.
> Sorry but almost everyone does have ID on them at all times.
This is objectively false. 11+% of Americans don't have a government issued photo id. Presumably many more forget their id on any given day. The folks who don't have IDs are disproportionately minorities and the elderly. In some places, as few as 55% off eligible Black people had IDs.
Millions more are here illegally, and yes, we should design our society to accommodate them. "Maybe they shouldn't be here" is both callous and unrealistic. No one should have to live in fear that they'll be spotted on some camera and have their life ruined.
> All this fear mongering about false positives is irrelevant because there is a human in the loop.
Humans in the loop won't fix failures. Last week we were all discussing how hospital administrators blamed "the algorithm" when vaccines were issued to admins rather than front line workers. People constantly say, "well, the computer says..." when the computer is obviously making a mistake. The TSA regularly screws up people's travel because they make mistakes about the no fly list. A human can be in the loop, but in practice the human isn't going to meaningfully question the outcome.
> Millions more are here illegally, and yes, we should design our society to accommodate them. "Maybe they shouldn't be here" is both callous and unrealistic. No one should have to live in fear that they'll be spotted on some camera and have their life ruined.
Those that break the law SHOULD have to live in fear that they'll face consequences. In some cases that may amount to having "their life ruined" and in other cases the consequences may not be that dramatic. But we have a society built on laws, and enforcing the law is required in a fair and just society.
> Last week we were all discussing how hospital administrators blamed "the algorithm" when vaccines were issued to admins rather than front line workers. People constantly say, "well, the computer says..." when the computer is obviously making a mistake.
In my opinion, you are comparing very different situations. The hospital admins have an incentive to get admins vaccinated first, and are not in a position where they're tasked with validating an algorithm's match. Additionally, the algorithm in question did not involve machine learning as a facial recognition algorithm would. It was simply software implementing a very straightforward formula (https://www.technologyreview.com/2020/12/21/1015303/stanford...).
> The TSA regularly screws up people's travel because they make mistakes about the no fly list.
This isn't because of facial recognition or black box machine learning algorithms. It is because the TSA uses very rudimentary heuristics like name matching, to err on the side of safety. I disagree with the unsophisticated approach they use, for what its worth. A complex AI based matching algorithm would do a better job of balancing safety against such mistakes.
> Those that break the law SHOULD have to live in fear that they'll face consequences.
Let's agree to disagree.
> The hospital admins have an incentive
The cops do too. Someone, somewhere is going to be pushing them to try and maximize the use of this very expensive system they just installed. Politically, they'll be very motivated by their superiors to say it's super good.
And if you aren't convinced that hospital admins or TSA or drivers using GPS deferring to computers aren't examples of general human behavior, here's some more:
* A machine learning algorithm detected you were in violation of copyright. Your content has been deleted/moderated/banned.
* A computer graded application process presents your scores to a human, who evaluates your employability, but largely just accepts whether the computer said, "Hire/No Hire".
* Ever been denied for a loan for something the computer says?
* Ever driven a self-driving car that does most of the work for you, but then you crash into people and things because you just generally trust it to do the right thing and stop paying attention?
Machine Learning is a powerful tool, but it is only as good as the humans who make it and the data they supply to train it. And the closer it gets to good, the more humans will believe it is actually flawless. Humans in the loop absolutely will not fix the flaws with a system like this, as time and time again humans have shown a propensity towards doing the easy thing -- trust the computer.
> es. Last week we were all discussing how hospital administrators blamed "the algorithm" when vaccines were issued to admins rather than front line workers. People constantly say, "well, the computer says..." when the computer is obviously making a mistake.
More often, they say it to defect blame to the computer when it is precisely implementing the intended policy.
> The TSA regularly screws up people's travel because they make mistakes about the no fly list.
No, because the no fly list implementation is designed to do exactly what it does, with (among other features) extremely loose matching criteria applied. “Mistakes” are a cover for policy.
And it's a completely different thing to use it to solve murder, rape and assault. The problem is that they all get bundled together and collectively dismissed under the guise of "privacy" and "potential for abuse".
Expect law abiding citizens migration to other places. The same way people migrate now from California because of tax issues. I don't understand what is the problem with computers doing facial recognition, at the end of the day it lands on some cop desk who can just look and see if there is real match, the same way he would have done if he had to browse through a folder manually. We did it for fingerprints, why not for faces? I don't get it.
I also don't understand what is wrong with "predictive" policing, whenever there is an area ridden with crime the locals complain why there is no more police there. This is exactly what predictive policing is about. People just don't want to deal with fact that there is more crime in black neighbourhoods. All those people who complain about it here will never go and live in those neighbourhoods because they do the same predictive algorithm in their minds but just based on prejudices rather than facts, they will not even live in a black neighbourhood even if there was no crime there, so now they complain about computers?
Can't help but wonder if this restriction applies to Sidney Tores' private, off-duty, but uniformed police force that runs rampant over the French Quarter, Marigny, lower Treme and parts of the CBD.
I'd expect not, but this city has trained it's citizens to keep an extra eye out for the corruption.
105 comments
[ 3.8 ms ] story [ 165 ms ] threadI am curious, because I wonder if this could prove to be a differentiator between cities. Perhaps you will end up with "Law and Order" cities, where technology is used unempathetically to punish harshly and discriminate without penalty. Then you would have less strict cities, where freedoms are less trampled in the pursuit of total control.
The states are highly autonomous and the laws vary most at the state level, but they vary from city to city within each state as well.
You can already see how this plays out: the protests in Portland were tolerated for a very long time, had they happened in almost any other city I doubt the local officials would have tolerated them for as long. (No comment on whether or not that’s a good thing.)
Everything is potentially a different system from how jail works, bail amounts, maximum penalties, typical plea bargain offers, what laws actually get enforced, etc.
This is already true. People are free to use marijuana/brothels/certain guns/etc in some places and not in others. Some cities have license plate readers on their police cars and roads and red light/speed cameras. Some prosecutors go after certain crimes and not others.
Although, I’m sure in every place there is little transparency so if you know the right people, you can get away with a lot more.
In other words the people at the top are worried that they'll get caught up in ways that can't be hidden or denied (because it'll be outside contractors running these things most likely) and so they want to make sure it never comes to pass.
I wish I wasn't so cynical about this as I do think the ban is a good thing in general.
There was also a lot of highly publicized examples of the police corruption and abuse during and after Katrina. There's a DOJ report here: https://www.justice.gov/sites/default/files/crt/legacy/2011/...
I'm hoping those things were also part of why it was easier to pass these sort of preventative laws.
A similar sentiment is expressed with respect to politics in some circles: as if people of color have more political acumen and agency than others by virtue of the racism they might have experienced.
Wow, really? Suggesting that a more diverse leadership team would better understand the potential downside of a police state is now distasteful?
That said, the parent poster said that many leadership positions, plural, were filled by people of color. I think it's probably reasonable to say that a community of many political leaders of color are more familiar with systemic racism. These are folks who specifically choose to get involved in politics and leadership, and likely represent fairly large constituencies. It's not unreasonable to assume that they broadly have experience or are able to communicate with their communities that have experience with systems that negatively impact people of color.
There's a big difference between, "These are the observed results" and "These people are inherently X". And especially different than, "Person is Y, therefore they are good/bad at X".
The first statement is open ended, it keeps the door open to systemic factors outside of race that might correlate with race. It also doesn't suggest that everyone of a particular group is better/worse at something.
The second statement says that there's something intrinsic to that everyone of that race. It sets up expectations for a group of people. Maybe you are an Asian who isn't good at math? How does that make you feel? How will you be impacted by other people making assumptions about you?
The third statement is applying stereotypes to individuals, which is just never going to be a good idea.
The differences between the first and second statements are subtle, but important.
And as parent pointed out, it's the leap from observed results to expectations about a racial group (or many groups in this case) that makes it racist.
So what would be the proper way to describe an ethnic group that, when examined purely statistically, really has performed well above average in mathamatics?
Do we simple never acknowledge that such differences exist, as some people have signalled?
Exactly the way you said it. "Statistically speaking, controlling (or not controlling) for relevant factors foo, bar, and baz, group X outperforms other groups such Y, and Z in tests of skill A". Which isn't the same thing as "X is good at A". "X is good at A" is a meaningless statement. It's devoid of context and prone to overgeneralization. It's not useful for anything.
Or even statistics like that to show what cloud hosting company overall has the best uptime?
I guess your saying that for any group, statistics are "not useful for anything"
I'm genuinely somewhat dumbfounded by this apparent difference.
"The Patriots have statistically scored more yards per game than average." Vs.
"Patriots team members are good at scoring yards." Vs.
"You are on the Patriots, therefore you are good at scoring yards"
In aggregate, the players perform in some positive or negative way, but you can't just apply that statistic to any individual player or even say that the players are unusually good at that thing. Maybe they have a great coaching staff, maybe they are cheating, maybe they have a few star players who push up the statistics.
And even more than that...I am seeing how old assumptions I've made don't hold up the way I've always thought they did.
edit: tbh, it's usually a pragmatic belief. It's easier to convince upper-middle class black people that what you want to do is good for black people in general (because their culture largely overlaps yours and they share your values), and wealthy black people and average black people look the same in pictures.
edit2: I hate to come back to this, but just look up Ray Nagin and his response to the hurricane that ultimately washed all of the black people out of New Orleans for an idea of the character of a lot black leadership in the south, and its total disconnection from any working class awareness. Just looked him up now to see what happened to him since and was not surprised.
Skin color is a poor proxy for diversity of thought or experience. We could instead just treat people as individuals and not draw broad conclusions on skin color.
Even more so in that scenario, I'd think.
If you've eliminated all the other variables, race would be the only differentiator of experience in that grouping.
(Of course, the reality of the non-hypothetical world is that people of different races tend to also have different schools, economic conditions, etc.)
I don't think that's the case at all. Again, skin color is a poor signifier for experience, and while it's good to acknowledge that person's life is impacted by their race, thinking that race is the only differentiator is silly.
Individuals are determined by more than their level of melatonin.
I suspect people of color would tend to disagree. It's likely a black person will have experiences not shared by your average white person, just like Americans may have experiences unique to them, or men, or disabled people, or people with any other characteristic that distinguishes them from others.
> thinking that race is the only differentiator is silly
No one said only.
> Individuals are determined by more than their level of melatonin.
Sure, but your hypothetical was "what if those things were gone?"
Regardless of that, it actually isn't clear that a more diverse leadership team brings better understanding. They may bring a diversity of understanding that is important to consider--it's not "better", in the sense that without their participation others would not see the potential downside.
Police state activities don't adversely impact just those in minority populations.
The police force is 50+% black, fwiw. Not saying that there isn't systemic racism, but there's proper racial representation on the force.
(Gender equality us another story, but I expect that they have a lot of trouble recruiting women for the job)
Isn’t reducing crime in one of the most unsafe cities in the country more helpful to POC living there than making moves to dodge accidental algorithmic bias? What are the expected costs and benefits here?
Because this kind of cynicism is self-fulfilling: if you don't believe anything good ever happens, then nothing ever will. So I feel it's important to actively avoid defaulting to it.
My anecdotal evidence: I just had a conversation with someone who was involved in the push for this to happen and at least that person (and purportedly, the organization they're involved with) is an idealistic programmer, and not a corrupt cop.
Instead of banning...regulate it.
Tech does not fix getting what we pay for.
Cheap police = thugs
They, we all deserve much better.
Basically, things like speeding in neighbourhoods, entering a crosswalk before all peds are off the roadway, running red lights and stop signs, right turn on red without coming to a full stop first, etc.
Currently in my city these things are all pretty much completely unenforced— the fines are in the hundreds of dollars are and only applied when someone is actually hit. This breeds a driver mentality of "well, it doesn't matter what I do as long as I don't hit anyone" which in turn makes the roads a less safe place to be for everyone. The fines are where they are partly as a deterrent, but also because the cost of enforcement is so astronomical when it requires multiple personnel on a special operation.
I feel that if the fine was in the $10-30 range, but it was applied consistently on every single offense, drivers would quickly train themselves out of doing these things. (This is the swift-certain-fair philosophy, see: https://nnscommunities.org/strategies/swift-certain-fair/)
Anyhow, I know that plate != driver, and there's a bunch of existing case law around protecting drivers' "right" to get away with endangering themselves and others by disregarding traffic rules. But given how much of a step beyond plate-recognition a facial-recognition system would be, I'm interested to speculate if there might be more of an appetite for this kind of thing down the road.
In any case, the wheels are already in motion for this to happen, so I suppose we'll find out soon enough where it leads to:
https://en.wikipedia.org/wiki/Traffic_enforcement_camera#Mul...
You don't need them until there's a home invasion. You don't need them until someone tries to rob or kidnap you or a member of your family. You don't need them until you are, like in Venezuela or the Czech Republic historically, in opposition to a dictatorial government. There is a real reason that every single authoritarian government in history has disarmed its populace.
Vigilance against authoritarianism is the same thing, because authoritarianism rarely happens over night. It takes years, sometimes decades. And if you've built up a network of cameras powered by facial recognition on every street corner, you've now built the apparatus for an authoritarian-leaning government to harness.
I would argue that this also is a uniquely American perspective— for me (an urban Canadian), the thought of starting a firefight with an intruder is horrifying, especially once the possibility of accidentally firing on a child or pet is accounted for. I'm also deeply uncomfortable with the numbers around what "gun in house" means for suicide and domestic disputes (even if I know I'll always be level-headed, what about down the road when my kids are teenagers and have friends over, etc etc).
Obviously, part of the guns-make-us-safer story is that criminals don't know which houses have them and which don't, and as a passive effect that's difficult to really study. But for a lot of us on the outside of this mentality, it's hard to escape the feeling that the safety provided by a gun is kind of like the safety provided by driving an SUV— mostly illusory, and comes at significant costs both to self and others.
> And if you've built up a network of cameras powered by facial recognition on every street corner, you've now built the apparatus for an authoritarian-leaning government to harness.
I would argue that this is a distraction— that level of surveillance already exists, but it's away from public scrutiny and is managed by Facebook and Google.
The clear and current authoritarianism threat in the US is the 106 house Republicans who supported Trump's absurd Texas lawsuit, that the leaders of one of the two major parties are unwilling to clearly speak with one voice to shut this nonsense down. And I'd be happy to be mistaken on this, but my impression is that gun owner clubs are mostly on Trump's side— those who are talking about taking to streets wouldn't be doing so to ensure a peaceful transition of power, they'd be doing so because they believe conspiracy theories about a stolen election.
It's ultimately your choice, but I've watched a lot of video on documented home invasions, and a few things are clear: a) Anyone willing to invade your home while armed will have no compunction about seriously injuring or killing any of the occupants; b) The defensive use of a firearm, while largely not reported, is an amazing deterrent in probable 95% of attempted home invasions where the homeowner has a firearm. Most intruders will nope out as soon as they see a gun, no shots necessary. c) It happens in Canada far more often than you think, including in cities. Here's a recent case: https://torontosun.com/news/crime/violent-home-invasion-may-.... Even if this case is gang-related, what are the odds that these people make mistakes and enter the wrong home/apartmnet? d) In rural areas, police are not getting there in any reasonable timeframe. It would take police probably 30 minutes to get to my house.
Ultimately, my view is that the right to self-defence is meaningless unless you have the means with which to actually defend yourself. I have no problem with people who are uncomfortable owning firearms and choose not to, but it's very problematic when those people force their views on others.
It's not legal in Canada to own firearms strictly for the purpose of self-defence (because our gun laws are quite silly once you become familiar with them), but most rural folks hunt or sport shoot so it's a moot point.
Oh, and by the way - meta-studies on the topic indicate that restricting access to firearms doesn't actually prevent suicides, it just causes people to shift methods to hanging, overdose, etc.
Regarding suicide, do you have a source for this claim? Everything I’ve read on this says quite the opposite, for example: https://www.hsph.harvard.edu/magazine/magazine_article/guns-...
This is from a giant meta-study that is often cited by advocates for more gun control (incorrectly, in my view, given the conclusions of the study): https://epirev.oxfordjournals.org/content/38/1/140.full.pdf+...
Canadian gun control laws are increasingly bordering on totalitarian. Given the general level of ignorance on firearms issues among the majority of Canadians who don't own firearms (and don't know much about them), I expect the laws to get much, much worse than they are even today. Even though I have been thoroughly vetted and done training and been tested, there are about 1,000 arbitrary and stupid ways in which I can be criminally charged for things that are not actually dangerous.
The recent arbitrary ban on "scary" semi-automatic rifles is a perfect example of this point. Because many of these rifles are classed as "restricted", and thus registered along with requiring more onerous licensing requirements, they are trivially easy to confiscate - this allows the government to score a political victory by appearing to be tackling gun violence, while ultimately threatening with criminal punishment Canadians who have attempted to be insanely diligent in following the law. The story that these guns have no purpose other than to kill people is simply a lie, a brazen one at that considering that the Prime Minister himself has owned such guns over the years.
Long guns, particularly legally-acquired restricted long guns, are used in homicides so rarely that it wouldn't be statistically significant. Legally-acquired handguns are also almost never used in crime or homicide.
Guns are also an important part of Canadian culture, and historically were a symbol of self-sovereignty: the ability to put food on the table and protect your family are simply part and parcel to living life on the frontier. Many in the city have apparently forgotten this in a sea of pleasant conveniences, but it is still a way of life for those of us who have chosen a life of self-sufficiency. (I used to live in a big city, btw)
Honestly, given the statistics on the prevalence of firearms vs. their use in violent encounters in this country, the fear many Canadians have of firearms is highly irrational.
The issue of course with your suggestion is these types of fines are highly regressive as it effectively allows the well off to break the law at will.
Percentage of wealth or income type fines would never float as you would end up making a whole new process where the police would only target those who they think will have the biggest payoff. Let alone it would require way too much of an invasion in privacy
The issue of course is how does such a system know who is driving without also combining facial recognition?
There will probably always be a segment of the population which can afford to buy their way out of trouble, but that's already the current state of affairs.
I don't think it would have to be a requirement for getting this off the ground, but if down the road it was found that a lot of people were treating the fines as pay-to-play, you could have them escalate. It would be easy to do something like the toilet-tank model where you get X number of base-fee offenses per time period, and then the price increases significantly. Lots of parking regimes do this kind of thing to discourage people from treating the fine as a fee.
(And the idea with the "swift" part of the philosophy is that you don't have to wait until the end of the month to get an itemized list of offenses— it texts you or something right away, so you get that immediate feedback about what you did and what the penalty was.)
> The issue of course is how does such a system know who is driving without also combining facial recognition?
Some existing systems have a camera positioned such that it can read the plate and also snap a picture of the driver through the windshield, so your ticket comes with a picture. In other places, the law has been amended so that the the person to whom the vehicle is licensed is liable for automated fines associated with it regardless of who is driving— Toronto started rolling out speed cams under that framework just this past summer (it's fines only, no demerits, so there should be no insurance impact).
It's both possible, and already in place in some countries.
https://www.theatlantic.com/business/archive/2015/03/finland...
Turns out basing it on income isn't much of an invasion of privacy because the government already has that. Even in the US we submit an annual "here's how much I made" document to the feds.
Cyclists want to use it against drivers, I'm sure shop owners want to use it against robbers, and doubtlessly paranoid parents want to use it against child nappers. Why should exceptions to a ban be carved out for anybody's pet interest?
So is it me asking for an exemption which would help law enforcement protect me and my family as we get around on foot or by bike? Or is it the drivers asking for an ongoing exemption from being asked to actually follow the laws which are supposed to apply to them?
This is stupid.
I don't see why it's so hard to vary policing levels based on local demand as well as need. If neighbourhood Xs representatives say they want fewer police, let them have it.
The point is to be transparent and precise with your models and then override them when that's what residents want.
Of course, there is also the issue of towns raising money from ticketing, but that's another point...
I don't see the feedback loop for robberies and murder etc. Maybe for speeding and cannabis possesion.
And the data issue is there whether you allow "AI" to analyze the data or mere mortal humans have to analyze it.
Advantage of AI may be that you can fix mistakes ones and they are fixed forever (like, let's say, an unjust bias against black people), whereas humans tend to fall in the same traps over and over again.
If a single person is incorrectly identified by a system design to catch warrants, the system has failed beyond redemption.
No system can guarantee such perfect recognition, so the only reasonable thing to do is never deploy them.
If you get falsely arrested (not even convicted), you can lose your job, get deported, spend thousands trying to clear your name, or even just get shot by the police.
AI tech will have false positives.
Technology that allows officers to screen many many people at scale will result in false positive arrests (or police violence) that would not have happened otherwise.
Therefore, this technology will ruin lives. This is clearly logical.
You might be ok with that cost, I'm not. But that's a different line of debate. I believe the idea that it is a net improvement is the idea that is unbelievable and unfounded given the current trends in AI. What gives you confidence?
Also, almost every state has a process to expunge arrests even IF you were convicted (of something minor).
As for why I’m confident about net improvement - it’s because the amount of true positive arrests (and subsequent convictions) will be far greater than false positive arrests. Therefore relative to the status quo it’ll be a net positive of criminals facing consequences, victims being compensated or getting closure, subsequent reduction in crime through deterrence, etc.
>anything short of perfect will absolutely ruin people's lives
this is more or less the basis for all thought within our system of justice.
What will happen is the police will consider the person dangerous, approach cautiously or aggressively, cuff them, throw them into the back of a cruiser and then maybe check their id.
And you are assuming that all people have ID on them. Or are even here legally.
"Show me your papers" isn't better because a computer told the cops who to target.
I notice that all the articles warning about AI supervision never mention the error rate of humans. Humans misidentify people, too.
AI could easily scale to watch everyone who walks through every intersection downtown. Humans can't.
If AI makes a much smaller percentage of mistakes (which is itself a questionable thesis), it will still make orders of magnitude more mistakes because it will be making orders of magnitude more evaluations.
Also, for better or for worse, people tend to trust that computers are 'logical' and correct. This is especially troubling because a human can explain their mistakes ("the person was wearing similar clothing, had a limp, etc."), while AI generally can't give you a reason that you can easily understand.
> Computers are only a tool. Humans have to make the final decisions. Under those circumstances, what exactly is the problem?
Humans are bad at evaluating computing systems. Have you heard the stories of people driving their cars into lakes because a GPS told them to? Now imagine the line was even fuzzier ("We think this could be our suspect, but we're not sure. The computer positively ID'd them though...")
Humans can be aware of the limitations of "AI". It's very rare that somebody drives into a lake because of GPS failures.
I believe humans will go, "well, the computer says you are so and so, and it's almost always right, so I'm going to need to take you in."
I've called customer service call centers and been told, "well the computer says...". If you think that the police would operate any differently, I don't know what to say.
Communities like police are poorly trained, minimally invested, and operate without oversight. They are typically overworked and just trying to close as many cases as they can.
Heck, even well educated people fall victim to this. We just had a front page story about administrators failure to deploy covid vaccines to frontline medical staff because "the algorithm made a bad choice".
And you overlook the possibility of human error without computer involvement. The classical "a black person mugged me" comes to mind.
At the end of the day it is a tool. Law enforcement should be done by professionals anyway. There are a million ways to get it wrong. At least with computers, you can standardize and enforce some good practices.
> And you are assuming that all people have ID on them. Or are even here legally.
Sorry but almost everyone does have ID on them at all times. It is unreasonable to design our society for those who don’t. As for those here illegally - well perhaps they shouldn’t be here illegally. Again, we shouldn’t design our society that is based on laws around them.
> If a single person is incorrectly identified by a system design to catch warrants, the system has failed beyond redemption.
I disagree. Any large scale and complex system will be imperfect. Your argument is basically “letting perfect be the enemy of good”.
This is objectively false. 11+% of Americans don't have a government issued photo id. Presumably many more forget their id on any given day. The folks who don't have IDs are disproportionately minorities and the elderly. In some places, as few as 55% off eligible Black people had IDs.
Millions more are here illegally, and yes, we should design our society to accommodate them. "Maybe they shouldn't be here" is both callous and unrealistic. No one should have to live in fear that they'll be spotted on some camera and have their life ruined.
> All this fear mongering about false positives is irrelevant because there is a human in the loop.
Humans in the loop won't fix failures. Last week we were all discussing how hospital administrators blamed "the algorithm" when vaccines were issued to admins rather than front line workers. People constantly say, "well, the computer says..." when the computer is obviously making a mistake. The TSA regularly screws up people's travel because they make mistakes about the no fly list. A human can be in the loop, but in practice the human isn't going to meaningfully question the outcome.
Those that break the law SHOULD have to live in fear that they'll face consequences. In some cases that may amount to having "their life ruined" and in other cases the consequences may not be that dramatic. But we have a society built on laws, and enforcing the law is required in a fair and just society.
> Last week we were all discussing how hospital administrators blamed "the algorithm" when vaccines were issued to admins rather than front line workers. People constantly say, "well, the computer says..." when the computer is obviously making a mistake.
In my opinion, you are comparing very different situations. The hospital admins have an incentive to get admins vaccinated first, and are not in a position where they're tasked with validating an algorithm's match. Additionally, the algorithm in question did not involve machine learning as a facial recognition algorithm would. It was simply software implementing a very straightforward formula (https://www.technologyreview.com/2020/12/21/1015303/stanford...).
> The TSA regularly screws up people's travel because they make mistakes about the no fly list.
This isn't because of facial recognition or black box machine learning algorithms. It is because the TSA uses very rudimentary heuristics like name matching, to err on the side of safety. I disagree with the unsophisticated approach they use, for what its worth. A complex AI based matching algorithm would do a better job of balancing safety against such mistakes.
Let's agree to disagree.
> The hospital admins have an incentive
The cops do too. Someone, somewhere is going to be pushing them to try and maximize the use of this very expensive system they just installed. Politically, they'll be very motivated by their superiors to say it's super good.
And if you aren't convinced that hospital admins or TSA or drivers using GPS deferring to computers aren't examples of general human behavior, here's some more:
* A machine learning algorithm detected you were in violation of copyright. Your content has been deleted/moderated/banned.
* A computer graded application process presents your scores to a human, who evaluates your employability, but largely just accepts whether the computer said, "Hire/No Hire".
* Ever been denied for a loan for something the computer says?
* Ever driven a self-driving car that does most of the work for you, but then you crash into people and things because you just generally trust it to do the right thing and stop paying attention?
* A police officer arrested an innocent man, and several people interrogated him based on a bad facial recognition algorithm. (https://www.nytimes.com/2020/06/24/technology/facial-recogni...)
Machine Learning is a powerful tool, but it is only as good as the humans who make it and the data they supply to train it. And the closer it gets to good, the more humans will believe it is actually flawless. Humans in the loop absolutely will not fix the flaws with a system like this, as time and time again humans have shown a propensity towards doing the easy thing -- trust the computer.
More often, they say it to defect blame to the computer when it is precisely implementing the intended policy.
> The TSA regularly screws up people's travel because they make mistakes about the no fly list.
No, because the no fly list implementation is designed to do exactly what it does, with (among other features) extremely loose matching criteria applied. “Mistakes” are a cover for policy.
I also don't understand what is wrong with "predictive" policing, whenever there is an area ridden with crime the locals complain why there is no more police there. This is exactly what predictive policing is about. People just don't want to deal with fact that there is more crime in black neighbourhoods. All those people who complain about it here will never go and live in those neighbourhoods because they do the same predictive algorithm in their minds but just based on prejudices rather than facts, they will not even live in a black neighbourhood even if there was no crime there, so now they complain about computers?
I'd expect not, but this city has trained it's citizens to keep an extra eye out for the corruption.