If the basis of banning facial recognition technology is its poor accuracy, will facial recognition technology be unbanned if it is 99% accurate for everyone?
A good analogy would be a chef tolerating fecal matter in their food. Yes, there is always some small chance of that happening, but no one accepts food from s chef that explicitly treats toilet water and claims 99% of the fecal matter is gone and only one in 100 people will get sick from it.
Also there is the idea that such an infastructure would scare criminals into not doing incriminating things.
To put it differently, explicitly accepting injustice is incompatible with mine and I believe society's understanding of justice. This is the same reason why ML shouldn't be used to replace human judges regardless of improvements in outcome. The authority of judges,cops and legislators is legitimate because the people have a contract with them where the people's view of justice is implemented as law in return the people subject themselves to these laws. Facial recogntion and ML judges violate that contract and as a result the outcome is illegitimate,100% of convicts become wrongly convicted because as a people the views on justice of their society was nor enforced by a member of their society that is elected or appointed by the people but by a blackbox logic that only looks at the outcome as a metric. A defense attorney can claim eyewitness testimony is inaccurate for example and cross examine witnesses but that attorney cannot interrogate a ML presenting as evidence a blurry picture with a 99% accuracy match.
Even if the criminals were afraid of the repercussion of their actions, and were aware of the system to catch them and be convinced it will work and would lead to their capture (that's already a LOT of if), this would only work on premeditated crimes, which is a minority of the ones commited in the street.
A lot of crimes are not though in advance, but done on the moment. People fighting because they are angry. Somebody stealing because they saw an opportunity.
We hear a lot about recurring pick pockets, or bank heists. But those are not the majority.
Most bad things in the street happen without being planned, and without consideration for the consequences.
Planned crimes are white colar crimes, and they don't get caught by cameras.
I still think it makes a big difference. I spent time in China and honestly I felt safer in areas where terrorists were a concern than I do in downtown SF.
This isn't the right characterisation.
Facial recognition at 95% can be very useful in helping to find people, initiate investigations. It doesn't have to provide grounds for arrest, let alone evidence for a trial.
Of course, there are problems even there, depending on a variety of issues.
Your license plate is being recorded in a variety of places, and for serious crimes, they will definitely use that information to 'look you up' FYI and face-matching isn't that much different.
As for arrest and prosecution, obviously that's another can of worms, but even then, there is a threshold at which we will effectively call it 'evidence'. 99.999% ID + was in the vicinity, plus no alibi, plus motive and prior record? That's a case.
If you think about it - we consider 'video evidence' to be fairly conclusive ... well I doubt humans are actually 99.999% accurate in their ability to matches faces to video! So it adds up.
" When it comes to justice, mistakes are tolerable so long as adequate compensation exists"
There doesn't need to be any compensation if there wasn't any malfeasance or incompetence in the system. To the extent our public servants are all acting above bar, using efficient, fair and lawful methods, then 'compensation' is not really the issue.
"It's better to let actual guilty criminals get away than explicitly and systemically accept even one innocent person being punished incorrectly,"
"accepting any amount of injustice invalidates that legitimacy and authority."
This is not true - by your very own logic. First - a justice system that 'lets people off who are 99.999% chance of being guilty' - implies quite a lot of injustice. Moreover, that in some instances, innocents will go to jail is tragic but absolutely inevitable. Not in 100 years will we have fixed this problem.
Along with some of the racial injustice you mention, you know what is 'unjust'? The amount of crime, particularly murder that goes unpunished in the US due to low case clearances, particularly in violent communities. That means aggressive killers on the block go free, gloat in their crimes, and do it again. Can you imagine having your sibling or child killed, and essentially knowing 'who did it' - but having them go free? This is common.
Facial recognition is not quite ready for prime-time, and it's invasive in ways that license plates are not, so it makes sense to ban it for the time being - however, there are thresholds at which it would make sense. Facial recognition in airports, near high value targets and such systems for alerting police to 'high value, dangerous targets' might make sense. But for everything else it's just not needed.
Even getting asked to come down to the station to give a statement based on an incorrect facial recognition is not something I would want to deal with for the sake of "the greater good".
95% seems awfully low accuracy, and even without an arrest this will cause a lot of headaches for many honest citizens.
> No - there doesn't need to be any compensation if there wasn't any malfeasance or incompetence in the system. To the extent our public servants are all acting above bar, using efficient, fair and lawful methods, then 'compensation' is not really the issue.
Sounds like a utopia, definitely not the world we live in.
Exactly
>We are not willing to give up liberties regardless of the benefits of the technology.
Finally town with smart people. I wish I could live there.
> As a people, we do not trust the government or the ruling class to stop at the good uses of this technology. This is why technocracy is evil, it only looks at outcomes, disregarding the will and consent of the governed.
That's the way to go. Face recognition is even worse than Traffic cams. I don't even think it's the matter of referendum. It should be considered a human right. I do not wish to see some majority could strip anyone from basic human rights.
Second - "We are not willing to give up liberties regardless of the benefits of the technology."
This makes no sense. It's like a line from a Rambo film. Communities trade liberties for benefits all the time.
Suppose police could clear 100% of murder crimes if we all consented to having our fingerprints taken. We we then do that? Probably. So long as there were no other ill effects.
This arbitrary notion of 'freedom' completely abnegates the fact that people live in communities with other people. Everything is a tradeoff. Absolutely everything.
Go ahead and use your chainsaw at 1 am in the city and see how 'free' you are to do such a simple thing as trim a hedge when you please. Or do it naked in the front yard.
Second - both these comments have shades of fear of the unknown.
This is the same kind of uproar people had about 'finger prints', 'DNA' and possibly 'seat belts'.
Both DNA and fingerprinting have been used expansively and widely for law enforcement, we generally accept their use, and are better to understand the parameters of them.
Using AI on a case by case basis in fact - is considerably less invasive.
Your fingerprint and DNA are semi-private bits of information.
You face is de-facto public.
Using 'AI' is no more 'technological' than any thing else, frankly, the term 'AI' is misleading because there is nothing special about it.
We could very well frame it as 'database search'.
---> The police took video footage from the crime scene and identified three possible suspects to the database one of the assailants was found to have the gun used to commit the crime.
It's ridiculous to assert there's something wrong here, let alone some kind of fundamentally new way of policing.
As for 'arbitrary surveillance' - that's another thing entirely. We don't have camera speed traps in a lot of places, that's fine, that's one application of technology.
I suggest we don't need broad surveillance either, but that has little to do with AI.
Crime is a real thing, cases unsolved and criminals unpunished definitely represent 'injustice' and so we need to use the tools we have to make the world a better place.
Let's take this simple one as example of stupid laws that suppose "to make world a better place "
Can you justify rationally and logically why naked person in the front yard is a problem? I mean, we have born this way aren't we? And yet having such biases you advocating for technology amplifier to enforce laws??
And you think you are making the world a better place ?
If you ready to prosecute someone for being naked, which means actually simply being himself and not sharing your views about nakedness, I prefer that you would be prohibited to have access from any surveillance completely.
>We could very well frame it as 'database search'.
And in that case I do not wish you to have any access to any database what so ever , and the 'data base' itself should be destroyed. Because 'AI' is a different level of magnitude in surveillance, because one thing is when you go some people notice you while some don't and another when something watching your every step. This by itself is intimidating. People would not feel free and relaxed they would fake it or smash the cameras at some point.
We already have too much surveillance to the level it's annoying to go outside. I do not feel comfortable any more. I do not like somebody watching me with AI precision. I was questioned by police just for taking the walk not in a 'proper' hours and all I wanted is just to take the air and think about deep topics without interruption. The world is already going crazy with too much control and some people seems do not realise that at all.
Will it ever be? Should it ever be? You have no perfect laws or rules and never will to begin with, so the idea of following rules with ideal tech-supported precision is wrong in it's core. Pursuing laws perfectly would kill any freedom and it's a perfect way to a totalitarian state. Searches were limited for a reason and thus tracking should be prohibited too, even if you have tech ability.
Mr. Turing himself was a victim of idiotic laws, I just wish he could get away with his "crime" of making love. If you think today it's different, it's not.
Too much laws and you have zero progress or evolution because progress comes from reviewing known and accepted concepts. You can look in history how many idiotic laws were in place, add to this tech precision of your dreams and the free world is finished. Would you like that? I don't, for sure.
This makes absolutely no sense.
We already use fingerprinting, DNA, photographic evidence - and in some cases worst of all '1st hand witnesses' as evidence.
Pursuing techniques and possibly technologies that more effectively help us create just out comes is exactly what we want to do.
Your example of 'Turing' is another thing entirely: what is deemed legal and not-legal is a completely separate thing from what we construe as evidence and not evidence.
"Too much laws and you have zero progress or evolution because progress comes from reviewing known and accepted concepts"
What does 'too much laws' mean, and what does it have to do with anything?
It means, that you have laws that you should not have at all.
It means that you never can make absolutely wise perfect laws. And it has to do EVERYTHING with this topic.
If you use perfect means to follow imperfect laws you would end up with stupidity amplified.
If you can't grasp this concept, look at the example of 'Turing' it is exactly about this pure dumb stupidity amplified.
We have now all kinds of idiotic laws and if to insist on them with too much of precision using technical amplifier their stupidity will be revealed too vividly and this would lead to a strong opposition and let us hope it would be not armed opposition, but I would not bet on it.
'95%' isn't meaningful, do you mean a 5% false positive rate, or 5% false negative rate?
A false positive rate as high as 5% would wrongly flag thousands of people.
How would the use of a technology make the person acting on the results of the information provided not accountable? There is no agency in those systems, so accountability can not be deferred to them.
"A good analogy would be a chef tolerating fecal matter in their food. Yes, there is always some small chance of that happening, but no one accepts food from s chef that explicitly treats toilet water and claims 99% of the fecal matter is gone and only one in 100 people will get sick from it. "
Actually that is exactly how the food industry operates.
https://www.technologyreview.com/2019/01/21/137783/algorithm...
Note that I am not arguing for nor against the specific use of the facial recognition systems by city officials.
Even worse, if you have a shipment that is 20% fecal matter, you can use it as long as you dilute it with other batches to the legal proportion.
if you sample enough, nearly any real world classifier will have a nontrivial false positive rate.
How would this work for the acceptance of self driving cars?
There have already been some other comments pointing out that this is an unreasonable standard, but they aren't really thinking big enough - the court-based system of justice isn't even that accurate. "Beyond a reasonable doubt" isn't going to be a standard of evidence even reaching 99% of evidence. Out of every 100 people in jail for murder, some of them just didn't do it. 100 people is a lot.
But if we have a reliable percentage that won't create too many false positives that can alert polices if a suspect is seen on a camera so that they would start the investigation, what is wrong with that?
I predict a lot of cities and countries will follow suit and start banning facial recognition technology altogether. If not then we're screwed. I already avoid traveling to the UK for the insane privacy invasion of their CCTV system. I get that there are benefits for crime investigation but it's far from worth it, the whole concept feels incredibly surreal (and wrong) to me.
Edit: To further your point, even if fingerprints were 100% accurate, there's still the off chance that someone planted your fingerprints. And someone could be wearing a you-mask, so you cannot rely on fingerprints or facial recognition for waterproof evidence. Which is why proper trials require multiple sources of evidence (I hope - IANAL).
The question is if we can do this without the people in charge abusing or not. If we can guarantee that it will be only being used for catching people with arrest warrant then I would have no problem with increasing the efficiency of cops
> The question is if we can do this without the people in charge abusing or not.
I think this is the main question and goes hand in hand with whether governments should be able to decrypt the internet.
Please let me move around freely, meet the people I want to meet, without having me added to some database of people with suspect contacts. I am fine with granting "criminals" the same privileges (as I posted in a comment yesterday[0] I am definitely a criminal, given the definition of the word).
You start with cameras everywhere for facial recognition, then you add microphones... it's hard to encrypt real-life discussion without inventing a new language, which could be easily decrypted anyways.
No one is presenting the outcome of a facial recognition algorithm's certainty in prosecutions. This is fundamentally incorrect
But what you are describing is really a problem with your "justice" system. A system where most people are coerced not to use their right to trial ("accept the plea deal or your sentence will be much worse", with many innocent people taking the plea deal) is simply incapable of delivering justice.
Think 99.99% and even then
99% accuracy in a city setting means (more or less) that you'll get the wrong person 1 in 100 times.
If you're scanning a crowd with this, you can imagine what can happen
I am not keen of my state knowing my whereabouts either and think camera deployment would create more problems than it solves. Countries where it has been deployed don't have impressive advantages to show for and all the privacy disadvantages. So why should we even consider it? What problem should it solve?
My opinion, if you are still scared of terrorists a psychiatrist might be more useful than a camera.
It used to be that law enforcement relied on fingerprints, eye-witnesses or a video tape backed by a judicial and legal system that - more or less - severely limited how and when that information could be collected and used e.g. within the context of discrete on-going criminal investigations.
This balance has been shifting entirely over the few short years.
Now, video camera's, fingerprinting and so on are collected pre-emptively; on literally everyone. Undeniably, it's easier to move forward on an investigation if you already have a lot of the data at hand. The argument put forward by proponents is that speeding up investigations saves lives. But that comes with a ton of hard problems.
First, searching accurately through a vast mountain of diffuse information is a hard problem. Law enforcement has outsourced that part to private companies, which opens up a can of worms in terms of confidentiality and privacy.
Second, searching through such a database becomes a black box as far as law enforcement is concerned: just upload a picture or a video fragment and it will easily yield any result. The fallacy being that convenience lulls LE into assuming that the results are accurate enough at face value. No need for verification or critical thinking (Much like you'd accept the results yielded by Google Search at face value).
Third, it's a convenient way to offshore responsibility towards the public. Law enforcement didn't make the wrong assertion if the wrong person ends up getting prosecuted and convicted: the data was just "not good enough". The fallacy here is that data gets treated as a commodity, which it is anything but.
Fourth, there's a difference between what's morally right and what's legal. The latter changes depending on how defines public governance. A database may be a great idea, until control is ceded to someone who uses that data against the public.
There's actually a historic precedent here: the 1943 bombing of the Amsterdam Civil Registry office. Following the 1940 German invasion of the Netherlands, all Dutch Jews had to carry a mandatory identity card and there whereabouts were recorded centrally at the Civil Registry. This was easy as, prior to the war, people's religious denomination had been recorded. The Dutch Resistance understood the importance and the downsides of a centralised record containing private information falling into the wrong hands. And so, they ended up attacking and demolishing the Civil Registry office containing that record. Albeit half-successful.
https://en.wikipedia.org/wiki/1943_bombing_of_the_Amsterdam_...
Obviously, I don't advocate attacking datacenters - that would be an effort in futility anyway - but your assertion that investing in mental health support would be more helpful rings very much true.
That's not great, but people perceive it as "fingerprints are unique".
I’ll take it.
There are narrow use cases where i think facial recognition in law enforcement would be a good thing, but it is ultimately too much power and too easily abused to trust legal systems to adhere to right usage. Banning is the correct action when misuse is as bad and correct use as complex as they are.
Sure, if your police/justice system is not competent you have a problem, but you had that in the first place.