Buzzfeednews does ok sometimes but I'd have to disagree with this. Even as a staunch privacy advocate facial recognition could do wonderful things to benefit society, is not the real problem overzealous governments and wideranging police powers lauded over innocent citizens?
Maybe there really needs to be a discussion about how intrusive we are going to allow them to be and the dangers that poses. It's a silly kneejerk response to say "we should only allow human beings to stare at cctv cameras, not an algorithm"
The cat's out the bag, the technology isn't the problem, tech is amoral, it's the people using it we need to solve.
There are countless things that our brains can do that are illegal to do with computers because our brains are so bad at it. Recording a phone call without notifying is illegal even though our brains remember it. Posting someone else's nudes without consent is illegal even though you can describe them from memory legally.
My eyes can't instantly recognise hundreds of strangers on the street and sync that data up with the global adtech database.
But my point would still stand even if I couldn't provide an example either. The burden of proving that a technology has no significant benefits should fall on the people trying to ban it.
You can simply equip those people with a tracker. Then it also works if they leave the stadium. Put them in their shoes and you dont need their cooperation either.
If they dont have any just put the persons picture on the stadium screens.
> The threat that facial recognition poses to human society and basic liberty far outweighs any potential benefits. It’s on a very short list of technologies — like nuclear and biological weapons — that are simply too dangerous to exist, and that we would have chosen not to develop had we had the foresight.
> is not the real problem overzealous governments and wideranging police intelligence agency powers lauded over innocent citizens?
Yes, they're the real problem. That doesn't mean it's futile to take away a powerful tool for them to use.
> tech is amoral
Tech is not amoral. It doesn't exist in a vacuum -- it has a context of who created it and what human actions it facilitates.
For example, ransomware is amoral. Sure, the tools used to build it are neutral, but that's true of anything if you zoom out far enough (e.g. steel is amoral, but CIA drones aren't).
> it's the people using it we need to solve
You can't "solve" people. There will always be selfish, stupid, and shortsighted people in the world. All you can do is make sure that they don't have concentrated, unilateral power over others.
Technology gives people that kind of power, and we can at least make sure that law enforcement in liberal countries doesn't misuse it.
Guns are also amoral but still manage to leave a bad taste in my mouth.
I think the reality is, this technology will be used against the public in negative ways. It's easier to get something banned than it is to change the nature of people.
I don't know where I stand on this to be honest. You are right that the real issue is "overzealous governments and wideranging police powers lauded over innocent citizens" but I don't know what the fix is.
The fix is laws that limit the power of governments/police. If governments are not willing to pass these laws, they probably won't be willing to pass laws preventing them using facial recognition to spy on citizens either.
You may cite me on this -
1. first of all to ban the idea you will have to be specific enough in describing the idea in a law that can be referenced so that when people think they are encountering the idea they may reference the law to be sure. Any one who can reference that law will be exposed to the idea.
2. ideas are not static expressions, perhaps we can think of them as nodes in a directed graph of all human thought, if you want delete the node you should direct all the nodes with arrows pointing at that node or someone is going to notice hey there are a lot of arrows pointing over here at this node that is gone and come up with the idea all over again by what is implied by the arrows.
3. if we think of ideas instead of how memetics would describe them we have a similar problem to number 2, that idea evolved from other ideas, and it is evolving into other ideas as we speak. Perhaps at some point we will have the technology to get rid of a memetic infection but not yet. And considering how well we've done on getting rid of measels I doubt it.
4. Probably the closest we can come to killing off an idea is the destructions of heresies in the middle ages, for example the Adamite heresy https://en.wikipedia.org/wiki/Adamites however the ideas still exist as evidenced by the Wikipedia page, just nobody cares enough to embody the idea.
So, instead of killing facial recogniton if a similar thing were done with facial recognition that was done with the ideas of the Adamites we should kill a lot of facial recognition researchers and people building facial recognition tools, and then hope that nobody new will care to do any facial recognition stuff (if you get brutal enough that just might be).
Making the idea unpopular might work, but I doubt it, because the utility and power the idea promises would be too great a temptation. Do you need a citation on the too great a temptation statement?
what does the ATF's gun database have to do with banning an idea?
The database may be used to ban guns, but probably not every gun in existence, and as such it will not do anything about banning the idea that you can shoot metal projectiles out of a small handheld device with a cylinder for shooting the projectile s out of (so as to provide aiming capabilities and speed) in a way that these small handheld devices can be used as effective people killing tools.
A similar assertion was made earlier today on Reddit. A few examples of technologies and concepts which have been banned or restricted, to greater or lesser effect:
wait - so you mean you can go to that site and read what the idea was that was banned!?! please refute my point #1 given this detail - an Idea which can be looked up online has not been banned, it has been disseminated.
This original subthread started off as a response to statement you can't ban an idea and someone wanting a citation on that.
I never said you can't ban a technology or an implementation of an idea, I said that it is logically impossible to ban an idea (point 1 - because you have to have an accurate enough description of the idea for people to reference when they encounter it in order to enforce the ban).
To reiterate an earlier example in as concise a way as possible - you cannot ban the idea of a gun but you can ban guns - of course people will still make guns but at much reduced numbers but the idea of gun will probably not be especially affected by the ban.
Yet in response I've gotten a lot of examples how people can ban implementations of ideas.
I'm trying to decide what makes facial recognition more dangerous than the tracking beacon we carry in our pockets that is famously being used by every bad actor on the planet to perform population surveillance for fun and profit.
Just riffing, I think the core differences are probably:
1) Your phone has some security, so only your telecom or manufacturer can spy on you easily and consistently if you take some precautions.
2) You can leave your phone at home, switch it off, or put it in airplane mode.
3) There are wiretapping laws already on the books that probably make it difficult to use certain evidence in court.
Whereas with facial recognition:
1) The recognizer is controlled by random property managers, both public and private. This makes it easy for bad actors to create systems of surveillance where they share with each other.
2) You can change your phone, but you can't change your face except in bad movies.
3) Fewer laws on the books to make abusing this data easier.
4) It's only semi-accurate, which is a plus, because a random detection gives you probable cause to fuck with people so long as you claim high accuracy on a carefully concocted test sample. The semi-random terror is a feature not a bug.
5) You can find people that are off grid that the community is hiding from the state or private security (undocumented immigrants, leftist activists, etc).
I think the comparison with smartphones misses the point,
>You can leave your phone at home, switch it off, or put it in airplane mode.
Makes it something entirely different. The core characteristics of facial recognition are
1) No need for a cooperative, consenting subject
2) Not alerting the subject being identified and no protection against being identified without noticing
3) Unchangeable identification characteristic
We already have reliable easy to use systems for characteristic 1 and 3, fingerprints, and I see how there might be a need for such systems at specific locations. For example at a police station or border checkpoints.
Then we have facial recognition which is a lot less accurate and its only additional benefit is point 2. Point 2 however has no positive use cases, its a purely totalitarian instrument aimed at its scale-ability. For not only being used in specific places but everywhere. You cant even make the argument which makes nuclear weapons a worthwhile technology to have, the possible usage for war against a foreign aggressor. Facial recognition is only beneficial to keep a population suppressed.
Good analysis. I hadn't thought of the lack of consent angle (though information leakage / hacking from phones renders them able to do similar things).
For what it's worth, the power dynamics might be different here. During cryptowars, we were trying to give people a power to hide from the state while the state already had the power to hide from the people. With this technology, while fun and gimmicky examples (like phone unlock, AR) come to mind for small scale uses, only dark and dangerous uses immediately come to mind which benefit the ruling class.
> With this technology, while fun and gimmicky examples (like phone unlock, AR) come to mind for small scale uses, only dark and dangerous uses immediately come to mind which benefit the ruling class.
That may say more about you than about the technology. You could use it to find the lost kid in a crowded stadium. Or find lost people with dementia anywhere.
And you are being incredible optimistic. No one is arguing that facial recognition can't be beneficial. We are discussing whether those pros outweigh the cons and honestly, with most government track record... I cannot say with a shred of confidence this technology will not be abused, both privately & publicly, without heavy, thoughtful and tech-savy regulations.
I will say with full confidence that this technology most certainly will be abused, regardless of whether it is regulated, and that the most flagrant abuses will be committed by organizations that are practically above the law. No regulation will stop the Chinese government, and there is not and never will be the political will to push through regulations powerful enough to constrain the NSA.
However, the regulations will be powerful enough stop any beneficial uses of facial recognition because say, the CDC, will follow the regulations to the letter and not spend the GDP of a South American nation in fighting it tooth and claw.
Does anyone believe that a ban would actually be effective? A talented teenager can already train a facial recognition model. In a few years, you'll be able to drop the "talented".
> Only a full ban — a federal ban, covering the use of facial recognition by government agencies, in public places, and in public contracts with private entities — can prevent our nightmares from becoming reality.
Talented teenagers are irrelevant, this is a call for a ban of using that technology by government entities.
Effective limitations on application, storage, transfer, admissibility (in criminal or civil courts), as well as substantial enforcement rights and noncompliance penalties, could reduce the risks to noise levels.
The major utility of such technolgies comes from centralised and intercorrelated records, sold for business use. Disrupt aggregation, transfer, correlation, and the business model, and you've effectively killed much the basis for establishing such tracking.
Banning FR-enabled monitoring cameras or surveillance systems (most deployments are not DYI), again, with bars on compiled databases (image collection does little without matching), and again, the effectve commercial basis is killed.
Specific targeted LEO or national intelligence aplications and deployments may persist, but the former would be of little use if barred by courts' evidentiary rules.
National intelligence use without commercial databases, and with bars on, e.g., sharing of state- and local-level identity documents or photos, again becomes much less useful for mass control, but instead reduces to fairly targeted investigation.
But one of the easiest to track everyone everywhere. What makes it so dangerous is its scaleability. The large error margin isnt a problem if your goal isnt finding single terrorists but create a surveillance state.
Often it's the impression, threat, or procedural presumed validity of effectiveness that matters far more than actuality.
The history of debunked methods used in government regulation, court and criminal evidence, or business practices is high. Simply posting "this area under visual surveillance" changes behaviours, and animal studies show dramatic differences in behaviour and detectable stress indicators under conditions of surveillance or predator presence. Even alpha predators react strongly negatively to the presence of human voices:
Some more specific dangers other than "the rich will get richer" would have been useful to mention.
As with most AI scares, people forget that AIs only replace the even more faulty human brain. Bias in AI can be debugged and worked around. Bias in human brains is much harder. You get no insights into why your bank doesn't give you a credit. Is it because of your skin color? Because they didn't like your nose?
The scary scenario of an AI judging people's attitudes and ruling them criminals only because they had a bad morning: the scenario of a human mob believing somebody to be guilty is much more scary, and it happens on a regular basis.
It would of course be bad to leave the judging to an AI alone. Of course the whole justice system would still be in place. That is unrelated to AI. The surveillance camera is not calling a killer squad on an unsuspecting grumpy person on public transport. It could merely draw the attention of a human to the scene, who could then judge on needed actions.
The example of the illegal immigrants also seems weird. Isn't it a good thing to be able to expose illegal behavior? I sympathize with the plight of immigrants, but rather than tolerating illegal behavior, it seems better to improve the laws to make good, desirable behavior legal.
I doubt HN is the best place to have discussions about design of societal/civil norms, culture and laws. Surely, there are people who are more expert than us who have well-researched, historically informed opinion on human behavior and intricacies of changing the balance of privacy vs centralized powers.
One thing is for sure. All of the technological advantages is creating asymmetry in information and hence power balance. This cannot be without serious ramifications to the society.
It's a less perfect forum than I'd hope (original submitter, BTW), and often fails to meet my expectations on important tech- and startup-adjacent questions. But it remains better than most online fora, and the questions should at least be raised or presented (if not adequately discussed or answered) here.
There's also the fact that this issue and its regulation or prohibition may be a concern to numerous startups, existing tech companies, employees, investors, etc., among HN's readership and community.
This particular submission is doing relatively well, so far. Toes crossed.
>> All I'm wondering is how this would play out in a dystopian future where I'm recognised by everyone wherever I go.
>>
>> And in some cases they might even have additional information maybe in a probabilistic fashion. Like what kind of clothes I prefer or the kinds of food I like. And that would be the good version.
>>
>> The worse scenario would be that some malicious actor puts all this together to play on others' fears.
>>
>> The worst case, of course, being big brother!
Thinking about this again though, I'm wondering whether this kind of transparency will actually start favouring common citizens. Imagine being able to recognise every bad (and good) cop, every corrupt politician, every criminal wherever they go. Would the social pressure to commit immoral or unethical acts increase many fold?
I do think that facial recognition has it's place in society, after all the cat is already out the bag. But I believe it should be reserved ONLY for border integrity and very high-level national security incidents. For example, I do not have any issue with facial recognition being implemented at all points of entry with it's data being purged after a randomized time with a set undisclosed minimum time frame. It is entirely within a country's rights to know who is entering and leaving their country in my opinion. This would strike a middle ground in both security and privacy and it's not like countries don't already have this information.
Private & for-profit use in any capacity should be banned.
Basically, the technology isn't bad. Humans are bad. So we should close off any and all avenues of abuse from a human agent. An AI system should be in place to create, access, store and retrieve the data with no human middle-man. The request to retrieve the data should follow the same procedures as for a warrant, but should require a high-level, non-partisan court order and should only return whether the subject was recognized and those specific frame of footage. The procedure should have sufficient checks and balances so no one can arbitrarily access this data. Every. Single. Request should be audited before being processed by multiple parties.
Stopping governments that abuse this technology should be encouraged. But that is naive thinking. It has definitely already begun in China for example. All we can do is keep it out of the hands of private corporations.
60 comments
[ 2.8 ms ] story [ 126 ms ] threadMaybe there really needs to be a discussion about how intrusive we are going to allow them to be and the dangers that poses. It's a silly kneejerk response to say "we should only allow human beings to stare at cctv cameras, not an algorithm"
The cat's out the bag, the technology isn't the problem, tech is amoral, it's the people using it we need to solve.
My eyes can't instantly recognise hundreds of strangers on the street and sync that data up with the global adtech database.
If they dont have any just put the persons picture on the stadium screens.
> The threat that facial recognition poses to human society and basic liberty far outweighs any potential benefits. It’s on a very short list of technologies — like nuclear and biological weapons — that are simply too dangerous to exist, and that we would have chosen not to develop had we had the foresight.
False equivalence aside, that's a fundamental misunderstanding of the nature of civilisation there
Oh wait, hold on, now we can see what that person is going to do next Tuesday. Let’s stop him doing that thing.
Same problem, different hypothetical technology.
Yes, they're the real problem. That doesn't mean it's futile to take away a powerful tool for them to use.
> tech is amoral
Tech is not amoral. It doesn't exist in a vacuum -- it has a context of who created it and what human actions it facilitates.
For example, ransomware is amoral. Sure, the tools used to build it are neutral, but that's true of anything if you zoom out far enough (e.g. steel is amoral, but CIA drones aren't).
> it's the people using it we need to solve
You can't "solve" people. There will always be selfish, stupid, and shortsighted people in the world. All you can do is make sure that they don't have concentrated, unilateral power over others.
Technology gives people that kind of power, and we can at least make sure that law enforcement in liberal countries doesn't misuse it.
> For example, ransomware is amoral.
Errrrr. Foot. Shot. In the.
I think the reality is, this technology will be used against the public in negative ways. It's easier to get something banned than it is to change the nature of people.
I don't know where I stand on this to be honest. You are right that the real issue is "overzealous governments and wideranging police powers lauded over innocent citizens" but I don't know what the fix is.
2. ideas are not static expressions, perhaps we can think of them as nodes in a directed graph of all human thought, if you want delete the node you should direct all the nodes with arrows pointing at that node or someone is going to notice hey there are a lot of arrows pointing over here at this node that is gone and come up with the idea all over again by what is implied by the arrows.
3. if we think of ideas instead of how memetics would describe them we have a similar problem to number 2, that idea evolved from other ideas, and it is evolving into other ideas as we speak. Perhaps at some point we will have the technology to get rid of a memetic infection but not yet. And considering how well we've done on getting rid of measels I doubt it.
4. Probably the closest we can come to killing off an idea is the destructions of heresies in the middle ages, for example the Adamite heresy https://en.wikipedia.org/wiki/Adamites however the ideas still exist as evidenced by the Wikipedia page, just nobody cares enough to embody the idea.
So, instead of killing facial recogniton if a similar thing were done with facial recognition that was done with the ideas of the Adamites we should kill a lot of facial recognition researchers and people building facial recognition tools, and then hope that nobody new will care to do any facial recognition stuff (if you get brutal enough that just might be).
Making the idea unpopular might work, but I doubt it, because the utility and power the idea promises would be too great a temptation. Do you need a citation on the too great a temptation statement?
The database may be used to ban guns, but probably not every gun in existence, and as such it will not do anything about banning the idea that you can shoot metal projectiles out of a small handheld device with a cylinder for shooting the projectile s out of (so as to provide aiming capabilities and speed) in a way that these small handheld devices can be used as effective people killing tools.
By the the Firearm Owners Protection Act (1986) a/k/a FOPA, 18 U.S.C. 926
https://www.law.cornell.edu/uscode/text/18/926
https://en.wikipedia.org/wiki/Firearm_Owners_Protection_Act
Which is why ATFs National Tracing Center remains paper-based:
https://www.latimes.com/nation/la-na-atf-guns-trace-center-2...
https://old.reddit.com/r/privacy/comments/cewumq/opinion_don...
I never said you can't ban a technology or an implementation of an idea, I said that it is logically impossible to ban an idea (point 1 - because you have to have an accurate enough description of the idea for people to reference when they encounter it in order to enforce the ban).
To reiterate an earlier example in as concise a way as possible - you cannot ban the idea of a gun but you can ban guns - of course people will still make guns but at much reduced numbers but the idea of gun will probably not be especially affected by the ban.
Yet in response I've gotten a lot of examples how people can ban implementations of ideas.
Banning facial recognition wouldn't prevent you from doing it, it would prevent anyone that doesn't run a crime syndicate from using or selling it.
https://globalnews.ca/news/4355444/chinook-mall-calgary-faci...
Edit: consider supporting the Canadian civil liberties association! They speak out about these sorts of things when others don’t
Just riffing, I think the core differences are probably:
1) Your phone has some security, so only your telecom or manufacturer can spy on you easily and consistently if you take some precautions.
2) You can leave your phone at home, switch it off, or put it in airplane mode.
3) There are wiretapping laws already on the books that probably make it difficult to use certain evidence in court.
Whereas with facial recognition:
1) The recognizer is controlled by random property managers, both public and private. This makes it easy for bad actors to create systems of surveillance where they share with each other.
2) You can change your phone, but you can't change your face except in bad movies.
3) Fewer laws on the books to make abusing this data easier.
4) It's only semi-accurate, which is a plus, because a random detection gives you probable cause to fuck with people so long as you claim high accuracy on a carefully concocted test sample. The semi-random terror is a feature not a bug.
5) You can find people that are off grid that the community is hiding from the state or private security (undocumented immigrants, leftist activists, etc).
So much potential here.
>You can leave your phone at home, switch it off, or put it in airplane mode.
Makes it something entirely different. The core characteristics of facial recognition are
1) No need for a cooperative, consenting subject
2) Not alerting the subject being identified and no protection against being identified without noticing
3) Unchangeable identification characteristic
We already have reliable easy to use systems for characteristic 1 and 3, fingerprints, and I see how there might be a need for such systems at specific locations. For example at a police station or border checkpoints.
Then we have facial recognition which is a lot less accurate and its only additional benefit is point 2. Point 2 however has no positive use cases, its a purely totalitarian instrument aimed at its scale-ability. For not only being used in specific places but everywhere. You cant even make the argument which makes nuclear weapons a worthwhile technology to have, the possible usage for war against a foreign aggressor. Facial recognition is only beneficial to keep a population suppressed.
What is the author's knowledge of the technology behind facial recognition?
What is the author's position on the snowball effect of banning algorithms?
Through what means does the author suggest we ban facial recognition?
https://evangreer.org/
That may say more about you than about the technology. You could use it to find the lost kid in a crowded stadium. Or find lost people with dementia anywhere.
However, the regulations will be powerful enough stop any beneficial uses of facial recognition because say, the CDC, will follow the regulations to the letter and not spend the GDP of a South American nation in fighting it tooth and claw.
> Only a full ban — a federal ban, covering the use of facial recognition by government agencies, in public places, and in public contracts with private entities — can prevent our nightmares from becoming reality.
Talented teenagers are irrelevant, this is a call for a ban of using that technology by government entities.
A talented teenager could also build a neutron source in 1990s [1]. That fact doesn't translate to the ineffectiveness of nuclear safety laws.
[1] https://en.wikipedia.org/wiki/David_Hahn
The major utility of such technolgies comes from centralised and intercorrelated records, sold for business use. Disrupt aggregation, transfer, correlation, and the business model, and you've effectively killed much the basis for establishing such tracking.
Banning FR-enabled monitoring cameras or surveillance systems (most deployments are not DYI), again, with bars on compiled databases (image collection does little without matching), and again, the effectve commercial basis is killed.
Specific targeted LEO or national intelligence aplications and deployments may persist, but the former would be of little use if barred by courts' evidentiary rules.
National intelligence use without commercial databases, and with bars on, e.g., sharing of state- and local-level identity documents or photos, again becomes much less useful for mass control, but instead reduces to fairly targeted investigation.
The history of debunked methods used in government regulation, court and criminal evidence, or business practices is high. Simply posting "this area under visual surveillance" changes behaviours, and animal studies show dramatic differences in behaviour and detectable stress indicators under conditions of surveillance or predator presence. Even alpha predators react strongly negatively to the presence of human voices:
https://www.theatlantic.com/science/archive/2019/07/humans-p...
Vigilance stress is almost certainly deeply innate. Humans are just as much affected:
https://www.healthline.com/health/hypervigilance
https://www.healthcentral.com/article/hypervigilance-in-anxi...
http://fellowtraveller.games/games/orwell/
As with most AI scares, people forget that AIs only replace the even more faulty human brain. Bias in AI can be debugged and worked around. Bias in human brains is much harder. You get no insights into why your bank doesn't give you a credit. Is it because of your skin color? Because they didn't like your nose?
The scary scenario of an AI judging people's attitudes and ruling them criminals only because they had a bad morning: the scenario of a human mob believing somebody to be guilty is much more scary, and it happens on a regular basis.
It would of course be bad to leave the judging to an AI alone. Of course the whole justice system would still be in place. That is unrelated to AI. The surveillance camera is not calling a killer squad on an unsuspecting grumpy person on public transport. It could merely draw the attention of a human to the scene, who could then judge on needed actions.
The example of the illegal immigrants also seems weird. Isn't it a good thing to be able to expose illegal behavior? I sympathize with the plight of immigrants, but rather than tolerating illegal behavior, it seems better to improve the laws to make good, desirable behavior legal.
One thing is for sure. All of the technological advantages is creating asymmetry in information and hence power balance. This cannot be without serious ramifications to the society.
There's also the fact that this issue and its regulation or prohibition may be a concern to numerous startups, existing tech companies, employees, investors, etc., among HN's readership and community.
This particular submission is doing relatively well, so far. Toes crossed.
>> All I'm wondering is how this would play out in a dystopian future where I'm recognised by everyone wherever I go. >> >> And in some cases they might even have additional information maybe in a probabilistic fashion. Like what kind of clothes I prefer or the kinds of food I like. And that would be the good version. >> >> The worse scenario would be that some malicious actor puts all this together to play on others' fears. >> >> The worst case, of course, being big brother!
Thinking about this again though, I'm wondering whether this kind of transparency will actually start favouring common citizens. Imagine being able to recognise every bad (and good) cop, every corrupt politician, every criminal wherever they go. Would the social pressure to commit immoral or unethical acts increase many fold?
I wonder...
[0]: https://news.ycombinator.com/item?id=20470887
Private & for-profit use in any capacity should be banned.
Basically, the technology isn't bad. Humans are bad. So we should close off any and all avenues of abuse from a human agent. An AI system should be in place to create, access, store and retrieve the data with no human middle-man. The request to retrieve the data should follow the same procedures as for a warrant, but should require a high-level, non-partisan court order and should only return whether the subject was recognized and those specific frame of footage. The procedure should have sufficient checks and balances so no one can arbitrarily access this data. Every. Single. Request should be audited before being processed by multiple parties.
Stopping governments that abuse this technology should be encouraged. But that is naive thinking. It has definitely already begun in China for example. All we can do is keep it out of the hands of private corporations.