Always easier when you can avoid the law and just buy it off the shelf. It’s fine to do this, we say, because it’s not being done by the government - but if they’re allowed to turn around and buy it we’re much worse off.
That's why it doesn't make sense to ban governments from doing things while still allowing private companies. Either it is illegal to surveil the public for everyone, or the government can always do it indirectly with the same effect.
I don't think the deal described here is even that egregious. It's basically a labeled data scrape. Any entity capable of training these LLMs are able to do this.
The difference is that a government can take personal liberty away from people in the most direct way. A private company can't decide to lock somebody away in prison or send them to death row. (Hopefully anyway.) So we put a higher standard on government.
That said, I do believe there ought to be more restrictions on private use of these technologies.
Yeah but these companies are operating hand in glove with govt such that there's no discernible difference between the current system and government just doing it themselves. Ban it outright.
Or we could just restrict funding for the governments/agencies/projects which are doing bad things (power of the purse, line item veto, sunset clauses, etc)...without hampering legitimate private enterprise. Instead we renew the NDAA every year.
(downvotes, keep em coming. I thought defunding ICE was more popular)
I would much rather have a democratically elected and constitutionally constrained government than private enterprise with limitless power. It would also be helpful if the “government is bad” people would stop electing the people who seek to sabotage the government.
This is why we should shun the people that build this stuff. If you take a paycheck to enable fascism, you're a bad person and should be unwelcome in polite society.
There are certain people who believe that average citizens can be held responsible for the actions of their government, to the point that they are valid military targets.
Well, if that's true then employees of the companies that build the tools for all this to happen can also be held responsible, no?
I'm actually an optimist and believe there will come a time whena whole lot of people will deny ever working for Palantir, for Clearview on this and so on.
What you, as a software engineer, help build has an impact on the world. These things couldn't exist if people didn't create and maintain them. I really hope people who work at these companies consider what they're helping to accomplish.
I never worked at a company that could broadly be considered unethical, I don't think. But it was always a bit disheartening how many little obviously unethical decisions (e.g., advertised monthly plans with a small print "annual contract" and cancellation fee) almost every other employee would just go along with implementing, no pushback whatsoever. I don't know what it is, but your average employee seemingly sees themselves as wholly separate from the work they're paid to do.
I have friends who are otherwise extremely progressive people, who I think are genuinely good people, who worked for Palantir for many years. The cognitive dissonance they must've dealt with...
Don't we already have facial recognition technology that isn't based on AI? why is throwing AI into the mix suddenly a reasonable product? Liability wavers?
I think the facial rec systems you're thinking of will recognize faces, but not ID them. They need you to label a face, and then it recognizes that face with a name from there on. Clearview is different in that you can provide it an unknown face and it returns a name. Whether it's just some ML based AI vs an LLM, it's still under the AI umbrella technically.
There are plenty of facial rec systems. Thinking of systems like in iOS Photos, or any of the other similar photo library systems. I think pretty much everyone would be freaked out if they started IDing people in your local libraries.
Note that there is no difference in the model or in the training. The only thing needed to convert ios photos into one that IDs people is access to a database mapping name to image. The IDing part is done after the "AI" part, it's just a dot product.
After the literal first one which just measured distance between nose and mouth and stuff like that from the 1960s, everything else has been based on AI.
If my memory serves me, we had a PCA and LDA based one in the 90s and then the 2000s we had a lot of hand-woven adaboosts and (non AI)SIFTs. This is where 3D sensors proved useful, and is the basis for all scifi potrayals of facial recognition(a surface depth map drawn on the face).
In the 2010s, when deep learning became feasible, facial recognition as well as all other AI started using an end to end neural network. This is what is used to this day. It is the first iteration pretty much to work flawlessly regardless of lighting, angle and what not. [1]
Note about the terms AI, ML, Signal processing:
In any given era:
- whatever data-fitting/function approximation method is the latest one is typically called AI.
- the previous generation one is called ML
- the really old now boring ones are called signal processing
Sometimes the calling-it-ML stage is skipped.
[1] All data fitting methods are only as good as the data. Most of these were trained on caucasian people initially so many of them were not as good for other people. These days the ones deployed by Google photos and stuff of course works for other races as well, but many models don't.
And this right here is why Clearview (and others) should have been torn apart back when they first appeared on stage.
I 'member people who warned about something like this having the potential to be abused for/by the government, we were ridiculed at best, and look where we are now, a couple of years later.
I think anything short of fully obscuring your face (a-la ICE-agent/stormtrooper) will be merely a mitigation and not 100% successful. I recall articles talking about face recognition being used "successfully" on people wearing surgical masks in China. In the US they ask you to remove face masks in places where face recognition is used (at the border, TSA checkpoints), but would be unsurprised if that isn't strictly needed in most cases (but asking people to remove it preemptively ends up being faster for throughput).
99.9% of people walk around with an electronic device that identifies them. If a particular person doesn’t, it should be trivial to filter out all the people that it couldn’t have been, leaving only a small list of possible people.
We need a Constitutional amendment that guarantees a complete right to anonymity at every level: financial, vehicular, travel, etc. This means the government must not take any steps to identify a person or link databases identifying people until there has been a documented crime where the person is a suspect.
Only if an anonymous person or their property is caught in a criminal act may the respective identity be investigated. This should be sufficient to ensure justice. Moreover, the evidence corresponding to the criminal act must be subject to a post-hoc judicial review for the justifiability of the conducted investigation.
Unfortunately for us, the day we stopped updating the Constitution is the day it all started going downhill.
Maybe. Anonymity is where bad actors play. Better to have better disclosure and de-anonymization in some cases. If some live in fear (e.g. of cartels), go after the cartels harder than they go after you.
That will be wildly unpopular with both parties and most importantly their constituents. I doubt even the libertarian party should they get the president, house and senate could pull it off
Note that the Amendment would apply only to the government, not to private interests. Even so, i could be unpopular among advertisers and data resellers, e.g. Clearview, who sell to the government. I guess these are what qualify as constituents these days. The people themselves have long been forgotten as being constituents.
(ProPublica, 404media, APM Marketplace, Associated Press, Vox, Block Club Chicago, Climate Town, Tampa Bay Times, etc get my journalism dollars as well)
Having AI assisted law enforcement will be a big force of making the law applied evenly. Law enforcement has limited resources so being able to give them a force multiplier will help clean up a lot of issues that were thought to be impossible to enforce before.
This is exactly, precisely the opposite of what the impact will be.
For example:
- every technology has false positives. False positives here will mean 4th amendment violations and will add an undue burden on people who share physical characteristics with those in the training data. (This is the updated "fits the description."
- this technology will predictably be used to enable dragnets in particular areas. Those areas will not necessarily be chosen on any rational basis.
- this is all predictable because we have watched the War on Drugs for 3 generations. We have all seen how it was a tactical militaristic problem in cities and became a health concern/addiction issues problem when enforced in rural areas. There is approximately zero chance this technology becomes the first use of law enforcement that applies laws evenly.
Not only is this incredibly naive, it misses that whole "consent of the governed" thing. I don't want AI involved in policing. They are bad enough and have so little accountability without "computer says so" to fall back on, That's all AI will do, make a bad situation worse.
The targets for the AI are still set by humans, the data the AI was trained on is still created by humans. Involving a computer in the system doesn't magically make it less biased.
are you sure it won't enabled targeted enforcement for people law enforcement finds irritating, more than evenly applied law? It's still people setting the priorities and exercising discretion about charging.
do you think the records of the vast number of police departments and agencies would be combinable with the separate court records, as well as the facial recognition access data source (if it exists?)
Then cops should be taken out of the core law enforcement agentic loop. There could be a new role of people who the AI dispatches instead to do law enforcement work in the real world.
I think you fundamentally misunderstand what the role of the police is. They protect property, the owning class, and the status quo. Laws are just a tool for them to do that. Equal justice for all is not a goal for them, and AI will not provide more of it.
Why do you write so many low-effort, disingenuous, inflammatory comments? They're "not even wrong", yet they just suck energy right out of productive discussion as people inevitably respond to one part of your broken framing, and then they're off to the races arguing about nonsense.
The main problem with the law not being applied evenly is structural - how do you get the people tasked with enforcing the law to enforce the law against their own ingroup? "AI" and the surveillance society will not solve this, rather they are making it ten times worse.
I want to share my opinion even if I know that it may not be a popular one on HN. I am not trying to maximize my reputation by always posting what I believe will get the most upvotes, but instead I prioritize sharing my opinion.
>people inevitably respond to one part of your broken framing, and then they're off to the races arguing about nonsense.
I agree that this unproductive. When people have two very different viewpoints it is hard for that gap to be bridged. I don't want to lay out my entire world view and argument from fist principals because it would take too much time and I doubt anyone would read it. Call it low effort if you want, but at least discussions don't turn into a collection of a single belief.
>how do you get the people tasked with enforcing the law to enforce the law against their own ingroup?
Ultimately law enforcement is responsible to the people so if the people don't want it then it will be hard to change. In regards to avoiding ingroup preference it would be worth coming up with ways of auditing cases that are not being looked into and having AI try to find patterns in what is causing it. The summaries of these patterns could be made public to allow voters and other officals to react to such information and apply needed changes to the system.
Always easier when you can avoid the law and just buy it off the shelf. It’s fine to do this, we say, because it’s not being done by the government - but if they’re allowed to turn around and buy it we’re much worse off.
That's why it doesn't make sense to ban governments from doing things while still allowing private companies. Either it is illegal to surveil the public for everyone, or the government can always do it indirectly with the same effect.
I don't think the deal described here is even that egregious. It's basically a labeled data scrape. Any entity capable of training these LLMs are able to do this.
The difference is that a government can take personal liberty away from people in the most direct way. A private company can't decide to lock somebody away in prison or send them to death row. (Hopefully anyway.) So we put a higher standard on government.
That said, I do believe there ought to be more restrictions on private use of these technologies.
The separation between private and the government is purely theatrics - a mere administrative shell.
Yeah but these companies are operating hand in glove with govt such that there's no discernible difference between the current system and government just doing it themselves. Ban it outright.
Or we could just restrict funding for the governments/agencies/projects which are doing bad things (power of the purse, line item veto, sunset clauses, etc)...without hampering legitimate private enterprise. Instead we renew the NDAA every year.
(downvotes, keep em coming. I thought defunding ICE was more popular)
I would much rather have a democratically elected and constitutionally constrained government than private enterprise with limitless power. It would also be helpful if the “government is bad” people would stop electing the people who seek to sabotage the government.
This is why we should shun the people that build this stuff. If you take a paycheck to enable fascism, you're a bad person and should be unwelcome in polite society.
local laws forbidding facial recognition tech have never been wiser
https://archive.is/3qywb
225k USD per year sells us cheaply!
There are certain people who believe that average citizens can be held responsible for the actions of their government, to the point that they are valid military targets.
Well, if that's true then employees of the companies that build the tools for all this to happen can also be held responsible, no?
I'm actually an optimist and believe there will come a time whena whole lot of people will deny ever working for Palantir, for Clearview on this and so on.
What you, as a software engineer, help build has an impact on the world. These things couldn't exist if people didn't create and maintain them. I really hope people who work at these companies consider what they're helping to accomplish.
I never worked at a company that could broadly be considered unethical, I don't think. But it was always a bit disheartening how many little obviously unethical decisions (e.g., advertised monthly plans with a small print "annual contract" and cancellation fee) almost every other employee would just go along with implementing, no pushback whatsoever. I don't know what it is, but your average employee seemingly sees themselves as wholly separate from the work they're paid to do.
I have friends who are otherwise extremely progressive people, who I think are genuinely good people, who worked for Palantir for many years. The cognitive dissonance they must've dealt with...
> I don't know what it is, but your average employee seemingly sees themselves as wholly separate from the work they're paid to do.
Hannah Arendt coined the term “the banality of evil”. Many people think they are just following orders without reflecting on their actions.
Don't we already have facial recognition technology that isn't based on AI? why is throwing AI into the mix suddenly a reasonable product? Liability wavers?
I think the facial rec systems you're thinking of will recognize faces, but not ID them. They need you to label a face, and then it recognizes that face with a name from there on. Clearview is different in that you can provide it an unknown face and it returns a name. Whether it's just some ML based AI vs an LLM, it's still under the AI umbrella technically.
Uh no? Facial recognition to names has been the bread and butter of facial recognition since the beginning. It’s literally the point.
There are plenty of facial rec systems. Thinking of systems like in iOS Photos, or any of the other similar photo library systems. I think pretty much everyone would be freaked out if they started IDing people in your local libraries.
Note that there is no difference in the model or in the training. The only thing needed to convert ios photos into one that IDs people is access to a database mapping name to image. The IDing part is done after the "AI" part, it's just a dot product.
Huh? What relevance does that have with the discussion?
After the literal first one which just measured distance between nose and mouth and stuff like that from the 1960s, everything else has been based on AI.
If my memory serves me, we had a PCA and LDA based one in the 90s and then the 2000s we had a lot of hand-woven adaboosts and (non AI)SIFTs. This is where 3D sensors proved useful, and is the basis for all scifi potrayals of facial recognition(a surface depth map drawn on the face).
In the 2010s, when deep learning became feasible, facial recognition as well as all other AI started using an end to end neural network. This is what is used to this day. It is the first iteration pretty much to work flawlessly regardless of lighting, angle and what not. [1]
Note about the terms AI, ML, Signal processing:
In any given era:
- whatever data-fitting/function approximation method is the latest one is typically called AI.
- the previous generation one is called ML
- the really old now boring ones are called signal processing
Sometimes the calling-it-ML stage is skipped.
[1] All data fitting methods are only as good as the data. Most of these were trained on caucasian people initially so many of them were not as good for other people. These days the ones deployed by Google photos and stuff of course works for other races as well, but many models don't.
And this right here is why Clearview (and others) should have been torn apart back when they first appeared on stage.
I 'member people who warned about something like this having the potential to be abused for/by the government, we were ridiculed at best, and look where we are now, a couple of years later.
"This cannot happen here" should be classified as a logical fallacy.
As stated in many of the comments in my code where some else branch claims this shouldn't be happening
Skynet. "You only postponed it. Judgment Day is inevitable."
Wear a face mask in public. Got it.
I think anything short of fully obscuring your face (a-la ICE-agent/stormtrooper) will be merely a mitigation and not 100% successful. I recall articles talking about face recognition being used "successfully" on people wearing surgical masks in China. In the US they ask you to remove face masks in places where face recognition is used (at the border, TSA checkpoints), but would be unsurprised if that isn't strictly needed in most cases (but asking people to remove it preemptively ends up being faster for throughput).
Probably room to add little cheek pads or other shape-shifters under the mask.
You have to change how you walk and sounds as well
99.9% of people walk around with an electronic device that identifies them. If a particular person doesn’t, it should be trivial to filter out all the people that it couldn’t have been, leaving only a small list of possible people.
Aren't we back to where this is illegal again, unless you're an ICE agent.
"Hey man, doctor's orders. Gotta wear it to get allergy relief. And no, can't ask about it... HIPAA stuff."
It is not a good idea to lie to an employee of the USA.
https://www.law.cornell.edu/uscode/text/18/1001
Sadly, I'm sure that will go over "not well" with ICE agents who will happily assault you for carrying a phone...
I disagree with the shooting too, but this is such a massive oversimplification of the event.
"I'll show you mine if you show me yours"
We need a Constitutional amendment that guarantees a complete right to anonymity at every level: financial, vehicular, travel, etc. This means the government must not take any steps to identify a person or link databases identifying people until there has been a documented crime where the person is a suspect.
Only if an anonymous person or their property is caught in a criminal act may the respective identity be investigated. This should be sufficient to ensure justice. Moreover, the evidence corresponding to the criminal act must be subject to a post-hoc judicial review for the justifiability of the conducted investigation.
Unfortunately for us, the day we stopped updating the Constitution is the day it all started going downhill.
Maybe. Anonymity is where bad actors play. Better to have better disclosure and de-anonymization in some cases. If some live in fear (e.g. of cartels), go after the cartels harder than they go after you.
> Anonymity is where bad actors play
No, that is a myth promulgated by control freaks and power seekers. As of late, deanonymization is where the bad actors play.
That will be wildly unpopular with both parties and most importantly their constituents. I doubt even the libertarian party should they get the president, house and senate could pull it off
Note that the Amendment would apply only to the government, not to private interests. Even so, i could be unpopular among advertisers and data resellers, e.g. Clearview, who sell to the government. I guess these are what qualify as constituents these days. The people themselves have long been forgotten as being constituents.
"You’ve read your last free article."
I don't think I've read a Wired article since 2002...
Wired still seems to write some good pieces.
I subscribe to keep the reporting going. Journalism costs money.
Most Americans don’t pay for news and don’t think they need to - https://news.ycombinator.com/item?id=46982633 - February 2026
(ProPublica, 404media, APM Marketplace, Associated Press, Vox, Block Club Chicago, Climate Town, Tampa Bay Times, etc get my journalism dollars as well)
are you using a vpn or something like that that might look like “you” have read wired articles?
Having AI assisted law enforcement will be a big force of making the law applied evenly. Law enforcement has limited resources so being able to give them a force multiplier will help clean up a lot of issues that were thought to be impossible to enforce before.
This is exactly, precisely the opposite of what the impact will be.
For example:
- every technology has false positives. False positives here will mean 4th amendment violations and will add an undue burden on people who share physical characteristics with those in the training data. (This is the updated "fits the description."
- this technology will predictably be used to enable dragnets in particular areas. Those areas will not necessarily be chosen on any rational basis.
- this is all predictable because we have watched the War on Drugs for 3 generations. We have all seen how it was a tactical militaristic problem in cities and became a health concern/addiction issues problem when enforced in rural areas. There is approximately zero chance this technology becomes the first use of law enforcement that applies laws evenly.
Same could be said about the computer systems that have been developed in the last 20 years. But that hasn’t happened…
Not only is this incredibly naive, it misses that whole "consent of the governed" thing. I don't want AI involved in policing. They are bad enough and have so little accountability without "computer says so" to fall back on, That's all AI will do, make a bad situation worse.
The targets for the AI are still set by humans, the data the AI was trained on is still created by humans. Involving a computer in the system doesn't magically make it less biased.
That is true for now, but eventually it should be possible for it to be more autonomous without needing humans to set its target.
are you sure it won't enabled targeted enforcement for people law enforcement finds irritating, more than evenly applied law? It's still people setting the priorities and exercising discretion about charging.
It should be easier to audit since you would have a list of who broke the law, but action had not been taken yet.
do you think the records of the vast number of police departments and agencies would be combinable with the separate court records, as well as the facial recognition access data source (if it exists?)
I think that is pretty unlikely
I wonder how many laws and sentencing guidelines etc are formulated with an implicit assumption that most of the time, people aren't caught.
In my estimation all of the criminal ones and at least half of the civil ones.
I think it will reveal unfair laws and as a society we will have to rebalance things that had such an assumption in place.
Meanwhile all AI face recognition software works poorely on non-caucasians.
With this administration, I think that is a feature not a bug
None of the destruction of your rights has lead to improvement in clearance rates.
Crimes aren't solved, despite having a literal panopticon. This view is just false.
Cops are choosing to not do their job. Giving them free access to all private information hasn't fixed that.
Then cops should be taken out of the core law enforcement agentic loop. There could be a new role of people who the AI dispatches instead to do law enforcement work in the real world.
I think you fundamentally misunderstand what the role of the police is. They protect property, the owning class, and the status quo. Laws are just a tool for them to do that. Equal justice for all is not a goal for them, and AI will not provide more of it.
Why do you write so many low-effort, disingenuous, inflammatory comments? They're "not even wrong", yet they just suck energy right out of productive discussion as people inevitably respond to one part of your broken framing, and then they're off to the races arguing about nonsense.
The main problem with the law not being applied evenly is structural - how do you get the people tasked with enforcing the law to enforce the law against their own ingroup? "AI" and the surveillance society will not solve this, rather they are making it ten times worse.
I want to share my opinion even if I know that it may not be a popular one on HN. I am not trying to maximize my reputation by always posting what I believe will get the most upvotes, but instead I prioritize sharing my opinion.
>people inevitably respond to one part of your broken framing, and then they're off to the races arguing about nonsense.
I agree that this unproductive. When people have two very different viewpoints it is hard for that gap to be bridged. I don't want to lay out my entire world view and argument from fist principals because it would take too much time and I doubt anyone would read it. Call it low effort if you want, but at least discussions don't turn into a collection of a single belief.
>how do you get the people tasked with enforcing the law to enforce the law against their own ingroup?
Ultimately law enforcement is responsible to the people so if the people don't want it then it will be hard to change. In regards to avoiding ingroup preference it would be worth coming up with ways of auditing cases that are not being looked into and having AI try to find patterns in what is causing it. The summaries of these patterns could be made public to allow voters and other officals to react to such information and apply needed changes to the system.
LE has been getting increasingly advanced technology over the years. The only thing that’s increased is their ability to repress and oppress.
Go lick boots elsewhere.