The right not to be subjected to AI profiling based on publicly available data
link.springer.com
link.springer.com
Sure, in the 1970s the cops could follow you around any time you were out in public. But they needed 6+ cops to provide round-the-clock surveillance, with an annual cost in the high six figures.
So the average citizen's privacy was protected not by law or high-minded ideals about privacy, but by simple numbers. There weren't enough cops to follow even 0.1% of the population.
But in the modern age of smartphones and license plate recognition and credit reference agencies? I think there's a strong argument to be made that we need new laws and new rights to reflect the new reality, where tracking people around the clock is several orders of magnitude cheaper.
Take something like speeding. We have speed limits and fines set based on the idea that cops won't catch most speeders and that cops can give leeway based on the situation (driving 10 mph over the speed limit on an empty straight highway is different than driving 10 mph over the speed limit weaving in heavy traffic).
We set a relatively low max speed and a high penalty, because you are only going to be pulled over a fraction of the time.
Then we roll out speed cameras which can catch people EVERY time they speed. It doesn't make sense to have the limits as low and the fines as high when every single person that goes over the limit can be fined. We have to tune the laws for perfect enforcement.
This is true for a lot of laws... most are designed based on the enforcement capabilities of the time. We need to adjust for technology.
Hikvision, in China, has a comprehensive system for this.[1]
Hikvision has other monitoring tools. Here's their Behavior Analysis Server data sheet. It's a 1U server for monitoring people.[2]
Hikvision Behavior Analysis Server
Based on the latest intelligent algorithm of deep learning, Behavior Analysis Server with a high-density GPU architecture supports the detection of behavior events in the perimeter, street, densely populated areas, indoor and other places, and triggers alarms in a timely manner, which can effectively improve the security of various places.
Behavior Analysis Server can detect the specific behaviors of individuals and groups in the perimeter, street, densely populated areas, indoor and other places, and provide professional video intelligent analysis applications:
* Perimeter Protection -- Real-time detection and alarming of events such as line crossing, area intrusion, region entrance/exiting, loitering, parking, unattended baggage or object removal.
* Trend Analysis -- Real-time detection and alarming of people density, real-time people counting statistics and people counting statistics.
* Street Behavior Detection -- Real-time detection and alarming of events such as fast moving, physical conflict, people gathering or falling down.
* Indoor Behavior Detection -- Real-time detection and alarming of events such as getting up, key person getting up, climbing, absence or sleep on duty, abnormal number of people, overstaying, sudden change of sound intensity, regional overstaying, physical conflict, standing up, sitting, or people counting.
[1] https://www.youtube.com/watch?v=Uh8DBYuDZyo
[2] https://www.hikvision.com/content/dam/hikvision/products/S00...
For me, I live in a quiet neighborhood. The residential street outside my apartment has speed limit signs of 20mph. People are trying to walk their dogs and kids. 9 out of 10 cars go 40-50mph. Part of that is road design. The road is new, straight, barrier and speed bump free. Worse, if you try to obey the law and go 20 the person behind you will get road rage and then illegally pass on the left (going into the opposite lane) or pass on the right and swerve around to get to the left turn lane.
That said, a street a block over is 25mph with speed bump and at least 1 of 3 cars races to each bump. And 1 of 20 just flies over the bumps.
This is a street with apartments and condos one one side and a park and elementary school on the other.
There might be some roads that don't need enforcement (somewhere in Nevada) but even in outside of the city, say LA to Vegas or SF to LA there's enough traffic that speeders cause accidents.
There's also a great video I stumbled across yesterday
https://www.youtube.com/watch?v=i3D7XYQExt0
Summary: Kinetic energy = 1/2 * mv^2
So if have a car going 70mph and they slam on their breaks to avoid an accident. When they stop they'll have expended 4900 units of kinetic energy. Another car, same weight, going 100mph slams on the breaks. When their car has expended 4900 units of kinetic energy they still have another 5100 units to get rid before their car will stop because 70^2 = 4900, and 100^2 = 10000. In other words, after they've applied as much stopping force as the 70mph car required to come to stop they're still going over 70mph
The point being, speeding issues scale exponentially, not linearly.
Look at the level of road fatalities [1] in the US - comparable with developing world countries. I vividly remember travelling on the interstate in the US in the 2000s, and seeing a couple of burnt out recent wrecks (most likely fatal) by the highway side every hour or so. That's incredibly shocking to someone from the EU. For contrast, I've driven by the site of a serious accident perhaps twice in my life here in Ireland - which has a road fatality rate 1/4 that of the US.
[1] https://en.wikipedia.org/wiki/List_of_countries_by_traffic-r...
Being a driver or passenger on a roadway in India is much more risky, but people in India are far less likely to die in this scenario because they spend so much less time in those situations than Americans.
Suppose drivers are perfect though. Things happen. You can't account for every kid, dog, ... running out in front of you. You don't know when there will be gravel on the road greatly increasing your braking distance. Top speeds of 15-25 are reasonable in a residential area in the sense that anything more greatly increases the chance and severity of pedestrian injury.
I don't think there is any truth to this at all.
> We set a relatively low max speed
We made the mistake of mixing vehicle and pedestrian infrastructure. Our speed choices have way more to do with this than with imagined enforcement outcomes.
> because you are only going to be pulled over a fraction of the time.
Many cites engage in "zero tolerance traffic enforcement" programs. This is where they patrol a single stretch of road and stop every single person who is even 1mph above the limit.
> most are designed based on the enforcement capabilities of the time.
Most are "designed" (a.k.a rapidly created and pushed into existence) in reaction to disasters that occurred and people broadly feel could have been prevented if there was a law curtailing the behavior that led up to the accident.
We didn't make speeding laws based upon "enforcement capabilities" we made them in response to "wasteful deaths."
Who is "we"? Vehicles and pedestrians mixing is something that predates recorded history.
> who is even 1mph above the limit.
Neither the radar gun, not the speedometer on a car are accurate enough for that.
The collective "we."
> Neither the radar gun, not the speedometer on a car are accurate enough for that.
You have a radar gun. You clock a car. The gun says 30mph. You clock the next car. The gun says 31mph. In the latter case you write a ticket.
You're free to argue the vagaries of measurement systems in court. Good luck.
This is incredibly common [1].
[1] https://www.findlaw.com/traffic/traffic-stops/can-i-challeng...
Won't need much luck, the first time this shows up in court it will be thrown out and officers told to stop writing false tickets.
There's a reason cops don't ticket anything less than 10mpg over - the tools they use are not just that accurate. It's a fake number.
>Many cites engage in "zero tolerance traffic enforcement" programs. This is where they patrol a single stretch of road and stop every single person who is even 1mph above the limit.
Searching this all I find is Virginia where over 80 is an automatic reckless driving, but the highest speed limit in the state is actually 70. I've never heard of anyone being pulled over for 1mph over the limit.
FWIW, that's the opposite of how citation requests work. Support your claims
I'm in the US and speeding is a leading cause of death. Driving deaths are increasing so I don't know if having less or no cameras is a good thing.
In my state, everyone speeds 10-15mph above the posted limits on highways. But the people that cause everyone around them to maneuver aren't the speeders, it is the people driving the posted limits and creating a bottleneck on the highway. And the number of rear end collisions caused by distracted drivers looking at screens instead of paying attention to their driving is the type of driving incident I see most often.
Pedestrian fatalities are at an all time high in the US and this trend is not present in other countries that also have smartphones.
Talk to anyone on a motorcycle, especially in states where they're allowed to lane split. Almost everyone is on their phones. Almost all the time.
People will not put down their devices with the current no consequence state we find ourselves now.
5-10 years should do it.
Rates per capita are much higher in the US, but the average American drives more than the average European so even if American drivers and European drives were equally good drivers we'd expect a higher death rate per capita in the US.
It is generally more useful to look at rates per vehicle per kilometer. By that the US is still higher than most of Europe, but not all. The Czech Republic at 9.8 is higher than the US at 8.3. Second in Europe is Belgium which is about 13% less than the US, followed by Slovenia at 16% less than the US. The rest of Europe ranges from about 30% below US to 64% below US.
This is honestly the biggest issue. So, so, so many drivers are so utterly shit at driving, and because our infrastructure is completely 100% car-centered, they HAVE to be permitted to drive unless their infractions add up to a degree where it becomes untenable to let them continue. And even then, due to the same pressures, they will probably still be driving because in many places in the States, there is simply no public transit whatsoever. They'll just then be driving without a license, and be subject to an extra fine on top of the fortune they already owe.
I work remotely but make a drive down to my employer for various reasons very regularly, usually once a month or so, and it takes me about three hours, and never, ever am I able to make that trip without seeing dozens and dozens of boneheaded, brain-dead maneuvers out of people. Traffic weaving, left-lane camping, people merging onto highways doing 40 mph, people who don't understand roundabouts, people making illegal U turns, crossing several lanes so as to not miss an exit. The state of driving in the US is an utter disgrace. So many drivers have absolutely no business behind the wheel ever again.
i.e. "We used to believe lies about how people drive, but thanks to the presence of more concrete evidence we are disproving those false assumptions."
One of the things Alphabet discovered early on in the Waymo experiment that was an eye-opener to the whole industry is that auto accidents were probably underestimate by a factor of three. When they started rolling out vehicles on the road, the best numbers available for accidents-per-mile were insurance reports and NHTSA incident records. Having vehicles with cameras on the road continuously revealed that there were 300% more accidents than those numbers suggested because humans are bumping into each other due to mis-estimates at stoplights all the time, but nobody wants their insurance rates to go up so they just don't report those incidents.
What if ubiquitous mass surveillance is good actually because it forces us to come to grips with realities we'd rather pretend are otherwise?
Base rate matters though; for example if literally everyone is speeding because speed limits are too low then every accident will involve someone speeding.
is that really the case though, or is it people are so self involved that they feel they are too important to have to move that slowly? Just because everyone else is speeding does not automatically mean that faster speed is safe. It could also just be that people are assholes and they do what they want.
The tongue-in-cheek way it was explained to me once was "the highway isn't 70 because you need that. It's 70 because the trucker driving sleep-deprived in a light rain who doesn't know his left tire is about to blow needs that."
A few decades ago the NYS Thruway caused a bit of political controversy because they started issuing speeding tickets if you traveled between two tollbooths faster than would be possible going the speed limit.
Then it still isn't good. The downsides outweigh the upsides.
That's actually only true per population. If you measure per miles driven the US does better then Europe (although Europe is large, and numbers vary in different countries).
US road death rate: 12.9 per 100k inhabitants, 8.3 per billion vehicle-km
France road death rate: 5 per 100k inhabitants, 5.8 per billion vehicle-km
UK : 2.9 per 100k inhabitants, 3.8 per billion vehicle-km
Sweeden: 2.2 per 100k inhabitants, 3.3 per billion vehicle-km
etc
The numbers aren't close, US roads and drivers are just much more dangerous
There are also some differences in data methodology US vs European countries, and when checked the rates from a different source (it was a while ago, I'll have to try to dig it up) the US came out better by comparison.
I think the difference had to do with what counted as a KM traveled.
Collision energy and thus damage increases with the square of speed (or ~speed^4 for head-on) so there is still an interest in controlling speed.
Most drivers (especially those over 35yo) will auto-regulate their speed to the optimal (safety vs throughput) for the road design. The problem is the ones who don't. Speed limits are set lower than this optimal speed, partly to make it easier to stop and charge drivers that can't auto-regulate well. Most of the time you will be ignored for going 5-10mph over. If you are over that, it is seen as deliberate defiance and "you are asking to be pulled over".
Automatic enforcement turns this de facto road law on it's head however.
During the 1970's oil crisis, highway speeds were capped at 55 mph nationwide. It took several decades for this to reverse and only after safety studies showed that differential speeds (those obeying and those going the optimal natural speed for the road) is a significant contributing factor in crashes. Unfortunately, speeds limits are still often below optimal because of an assumption that every driver will always go at least 5mph over the limit (which is incorrect).
On 70 mph interstates away from urban/commuter traffic (where time pressures often affect driving), It's not unusual to see some cars going 5mph below the limit. That is a sign that these Interstate segments have the optimal natural speed.
Also the safety of speed for a given environment should include pedestrians. Many advocates for urban areas rightfully push for 25 mph limits for exactly that reason. If you want to successfully convert in town urban roads that are wide and have high speeds to 25mph there are two good options: speed calming measures or speed cameras.
https://www.chp.ca.gov/CommunityOutreachAndMediaRelationsSit...
They aren't enforced (nor does it feel like any traffic laws are enforced anymore)
Hmm... seems like they're enforced here? Perhaps we need to up the fines though.
The energy adds, not multiplies. Multiplying would change the units
https://injuryfacts.nsc.org/motor-vehicle/motor-vehicle-safe...
I think you are already doing subconciously what i am talking about; you are thinking of speeding, but I doubt you are thinking of someone driving 50 mph in a 45mph zone. That is not the type of speeding that kills, but would be the type that could be caught with perfect enforcement.
I know you're pulling this up as an example of a small infringement, but there are studies that quantify the fatality rate as a function of velocity. The numbers you picked are right in the steepest part of the increase. Using equation 2.3 from [0] (with conversions from mph to kph), there's a 64% chance of fatality at 45mph, but a 83% chance of fatality at 50mph.
[0] https://nacto.org/docs/usdg/relationship_between_speed_risk_...
Speeding/speeds would be reduced if they simply designed roads to be harder to drive above the speed limit on, like being narrower and less straight. They don't do this.
We adjust for technology 100% of the time when technology has made enforcement harder.
We adjust for technology 0% of the time when technology has made enforcement easier.
Surely this is only reasonable for large, out of town roads. Within cities, you definitely still want low speed, high fine.
Obviously this is an oversimplification, Switzerland and Spain are 120 km/h (75 mph), Germany often has no upper limit, a few places are 140 or higher. But in general it seems like most of Europe sets the limit at the speed Americans actually drive.
In towns it’s 30 and there’s plenty of cameras, for speed, red lights, bus lanes etc.
The world doesn’t end.
The cameras are established to clock you at position X1 at time Y1. Then the next cameras a handful or more miles down the road clock you at position X2 at time Y2. You get a ticket if (X2 - X1) / (Y2 - Y1) > Limit + e.
You can speed all you'd like between those cameras, but unless you're exiting before the next set, you'll have to pull over and wait the amount of time necessary to bring you back to the speed limit for that area, achieving zero reduction in total trip time.
This also doesn't really touch on the perverse incentives when implementing automated fines. Want to make a stop light intersection safer? Increase the yellow light time. It's been proven to work, over and over and over again. Sometimes we see cities shorten yellow lights, increasing accident risk in an effort to get more revenue.
https://www.koaa.com/news/news5-investigates/news-5-investig...
We know we can better control driver speed through road design. That's been effectively demonstrated through studies. Yet our "solution" to speeding is to make the roads as straight and wide and clear as possible and then give you a fine for using them as they have clearly been designed.
Oh it makes sense alright...for police departments mandating a ticket quota and receiving direct revenue from speed camera fines.
I was not saying there is anything special about being behind the wheel that makes it ok to break a law. I used that example because it fits with the 'low enforcement rate but a high penalty to make up for the low enforcement rate' combo.
How about an example like playing poker for money with friends? Illegal in most places where gambling is illegal, but people do it all the time. The laws might make sense if we think casinos are bad for society and want to prevent them, but think occasional gambling amongst friends is fine.
The current laws don't carve out home poker games because there was never a need to; there was no way to enforce the law against small friend groups gambling. If there suddenly became a way, we would need to re-write the laws to permit home games.
One other example for me could be something like drinking alcohol in a public place. It's never enforced if people are just enjoying a beer quietly during a picnic and bother no one. However if sound equipment mounted in trees could detect the opening of a can and write a ticket, I would feel the law obviously would need to be amended.
We haven't redesigned our cities to be non-car centric, with good public transportation and street design that favors lower speeds. But we've got a bunch of people running around wanting to severely punish anyone breaking the speed limits (because "fuck cars" means holding individual people responsible for being born in a car-centric society). The result will be that one morning they'll wake up and have to deal with how much they've punitively hurt the working poor in this country. I guess that'll be okay though because that "in this house" rainbow flag in their window means that they care.
But not breaking the law when driving is something people are in full control of themselves. I don't buy the premise that it's a regressive tax. Yes, some laws disproportionately hit certain demographics, but not speeding is not one of those.
> I guess that'll be okay though because that "in this house" rainbow flag in their window means that they care
I'm not sure what you mean by that? Are you conflating lgbt stuff into a comment about cars and technological advancements in enforcement, or am I misunderstanding what you mean?
The poorest may not, but the working poor are likely to be highly dependent on some cheap, old vehicle.
> Yes, some laws disproportionately hit certain demographics, but not speeding is not one of those.
WFH software developers are probably going to be hit a lot less than someone who needs to drive from where the cheap rents are to the job site every day.
And that was a comment on performative liberalism.
And the "if the penalty for a crime is a fine, then that law only exists for the lower class" aphorism applies.
This is the kind of thing that's only repeated by slumming upper-class children, the same ones who live in Brooklyn, have leftist podcasts, and think it's actively good when you see someone smoking crack on the subway because it's "cool".
Actual working class people don't like it when other working class people speed near them or commit crimes!
(It's also a very American statement, because in other countries the upper class is not the people with the most money, it's the people with the most tradition and social status. Or they own a lot of land but are cash poor.)
Survey 100 working class Americans and ask if they think another working class American should receive a ticket for driving 65 in a 55 on an Interstate.
I’d be shocked if even 5% of them wanted that outcome.
50 in a 25? Sure. 100 in a 55? Sure. But I doubt they want perfect enforcement, which is the topic being discussed here.
In fact, the reason most national highways have a speedlimit of 50mph, even out in the boonies in Kansas where everything is flat and straight, is because of the fuel crisis of the 70s.
When I was a child, I was curious why highways were 50mph but the interstates had a speedlimit of 65-70mph, so I went and found out instead of assuming it was a disingenous revenue grab.
Most interstates were generally built after the fuel crisis, and they modified the national speed limit in 1988, setting the speed to 65mph, again for efficiency reasons. It was repealed in 1995. Perhaps 65mph is what was considered safe for the road and vehicular technology at the time, and no one has had the political or municipal capital to do a new study ever since.
You want that speed limit changed, contact your reps, I guess.
Wouldn't this be an argument for a higher EV speed limit?
The efficiency loss at speed still applies to EVs due to friction and laws of momentum.
A large enough change in scale manifests as a change in kind.
As a topical example, SpaceX is trying to reduce launch costs by 20x with Starship, which just had its 4th test flight this morning. Some don't see the point: there's not nearly enough demand to launch thousands of tons into orbit per year. But 20x is a lot cheaper, so they're banking on induced demand expanding the kinds of projects that send things to space.
It's certainly a gamble that there will be a million people willing to spend $200k to do this, which… well, I think it's not entirely impossible, as the idea was something I liked until Musk's own personality put me off the idea of being stuck on a planet with only his sycophants for company.
Which is a bit of a shame, but still, I'd give it 50-50 odds of being a product-market fit just for that.
Elon is undoubtedly great at getting folks to commit money and talent to his companies, and talking about plausible-ish things like becoming a multiplanetary species is one way he accomplishes that.
"Dying on Mars, just not on impact" is Musk's final bucket list item. Almost everything else he does is to enable that vision, either directly by creating the tech, or indirectly because he knows this is expensive.
He may well fail, nobody's ever done this and we don't know how many surprises there will be.
But the ship working well enough for $200k tickets is plausible.
(The idea that banks will give people loans for that, not so much: without multi-planetary trade, nothing that happens on Mars can repay a debt on Earth, and I don't see Mars as having any special economic benefits to make such trade worthwhile).
This is just marketing hype.
That money is a thing he needs for the very expensive bucket list item.
They won't make a satellite but they will demand things which middlemen can solve with a satellite.
> You have to find the customers, it is not like you make a product and everyone changes their desires to conform.
Once smartphones were invented, people found within themselves the desire for a smartphone.
Although there's a lot of debate about this one: https://en.wikipedia.org/wiki/Say's_law
So-called "induced demand" is just a dumbing down of the more fundamental notion of supply and demand, for people who aren't comfortable thinking about math, calculus, dynamic equilibria, etc. If you read the wikipedia article on it, you'll see they constantly describe it in terms of supply and demand. The term was originally "defined" in 1999 in a paper that was not written by economists. It's not an economics term.
in terms of transportation planning, a better way to think about it is, "misery distributes itself throughout the system".
With things like Project2025 being embraced, it's horrifying what the future may hold if checks and balances aren't in place.
[1] https://merics.org/en/comment/chinas-social-credit-score-unt...
nobody thought there was a problem when you stuck up a wanted posted and asked a community 'do you know who this is'? but if you stick that poster into a computer and ask, all of a sudden people complain that it's a privacy issue.
i really struggle to see how they're different and the only thing i've been able to some up with is that people feel the only counterweight to abuses and biases of policing is to build in random inefficiencies
Given how much crime currently goes uninvestigated because the backlog is so high, is automating some of the coincidence-sniffing a bad thing?
If in the past you’d have to hire more local policemen, they’d need more police vans made, they’d need more uniforms made, more wear on police vans means more work for the mechanics, etc etc etc, now you just pay Protector Co. once for your new surveillance software. Now you haven’t dissuaded people from crime with good plentiful local jobs (as a gross oversimplification, but hopefully you can see this would still apply to a more realistically complex economy)
If there was, like, a club devoted to organizing the sort of informal observations of who was where and trying to track everybody the old fashioned way, I’d think everyone involved was a real creep there, too. Although, of course, this would be a bit impractical.
And the paparazzi weren’t some beloved people, and they mostly targeted public figures!
There’s something to be said for the inefficiency I guess. But, I think it is also to a large extent a proxy for the idea that normal people only went on the hunt for somebody for a good reason.
1. Wanted posters were only hung up based on human witness reports and for a very limited number of suspects. With face recognition, you can hunt automatically for hundreds who happened to be around a specific place at the wrong time.
2. Reach, and therefore likelihood of false-positive face matches.
3. Investigators wouldn’t blindly believe a random community member reporting a match, they would check if the reported suspect really matches the description and the circumstances. With face recognition and AI, people tend to just assume that the computer is correct.
When you look at a punishment which is set up as a “deterrent” or you hear about someone “throwing the book at you” it’s because of all the perceived inefficiency. They caught “you” this time, and they’re accounting for all the times “you got away with it” (statistically).
If you go from 10% efficiency to 100% efficiency without changing the punishment, observant people will, I believe correctly, take issue with the fact the punishment has now become “10x more drastic” when taken as a whole society.
There is also something about this inefficiency that makes us more free. Everyone bends the rules to some extent in an area they feel is worth the risk. Maybe you changed an outlet without consulting a licensed electrician / inspector.
To that end, the inefficiencies are built into our economies. Would people be safer with every electrical change being measured? Surely. Could the economy support every electrical change being measured? As in, can the average American even afford to pay an electrician every time they want to change a fixture? I think the answer is realistically no. If everyone has enough free cash they don’t mind spending it on an electrician, 100% efficiency and compliance might be reasonable and good.
You’ll see this phenomenon while traveling, there are things like traffic violations, sketchy wiring, eyesores, etc that you’d say “that would never fly in $HOME_COUNTRY” and it’s commonplace there. They may even still be illegal there. But the standards of enforcement change to meet what are in principle the ideals of the people, but in practice what’s possible for the economies to support. You might go back in 10 years and find their economy is stronger, the enforcement has increased, but people are not unhappy with the change and life goes on. The people who did the sketchy wiring have probably had time and an influx of cash from the stronger economy to pursue better training, and are now the ones doing good to-code work.
The way technology throws this out of balance is to give enforcement the ability to expand without incurring the traditional economic penalties for doing so. Those economic penalties would’ve grown the economy in other areas. If you follow the line of thinking, it leads to a significant increase in the wealth gap. Of course, there’s enough evidence to support that this is exactly what’s been unfolding. My thought is, if this natural economic balancer is reduced, we should seek to add another.
This is why every government serious on surveillance would turn its citizens into snitches
Now it's easily hidden. Somebody can monitor you just by accessing information systems without the people collecting the data even knowing. You can now do the surveillance alone, no pesky accomplices needed.
The personality summary was concerning as it was both deeply accurate in some respects and deeply inaccurate in others. Most concerning, it said I was "risk adverse" and "struggles to make decisions with incomplete data".
With 35 years at startups and independent contracting, risk tolerance and ability to make decisions with incomplete data are kind of in my wheelhouse.
Worse, if this profile was being shown to potential employers, it could (would?) be a deal breaker. It's kind of like being judged by your MBTI results.
Similar with social media advertising profiles. I've seen what FB and Twitter thought of me, and it included interests in various spectator sports, which were wrong both as a general statement about my personality, and specifically those sports don't even have any societal significance outside the USA.
Also several languages that I don't speak.
And they showed me ads for both dick pills and breast surgery; and others for a lawyer who specialised in renouncing a citizenship I've never had (for tax purposes) when moving to a country that I left; and also an announcement by the government of a country I don't live in about a ban to a specific breed of dog I've never heard of, when I don't own any dog anyway and never have.
And people complain about LLMs making things up :P
I had to do that test a bunch of times coming out of university, so I am sure that is happening.
It's better for it to aggregate the information and then provide the resources to be able to verify whether any deduction is well supported or not.
I guess with a few more iterations you could have another agent verify whether a deduction is well justified, but that will also have some significant percentage of errors too.
Story time: while I still worked at Facebook, there was a company wide project for data attribution to comply with opting out of personalization (I believe to comply with an EU directive but don't quote me on that). The idea was to identify the source of any data by esentially tagging it on a granular level. This affected all the offline data procesing and ML processes but also code (online and offline) where various tools and systems were being built to analyze sites of data usage to detect use of personalized data and respect opt outs where applicable.
I made two predictions at the early stages of this:
1. The tools and systems would tell us "all data is used for everything" and
2. Creating new non-personalized data pipelines and add things to them would be far easier and faster than trying to remove personalzied data from existing pieplines. Then, things like ad performance become an optimizatino problem with a clear benchmark (eg new unpersonalized ad serving pipeline vs the old personalized pipeline).
A lot of work did, I believe, basically confirm (1). Untangling that seems, at least to me, to be a Sisyphean task. I don't know where this project ended up since I left while it's ongoing but I stand by (2).
My point is that (IMHO) opt-out just doesn't work for this kind of thing. If we really care about data privacy and authorized use of data at some point we will need to take the oposite approach and enumerate what data we're allowed to use.
Almighty Dollar™, god...
Reminds me of the tendency in program analysis to just spit out top (e.g. all possible values) because the state space overwhelms automated reasoning, and the only way back is annotating/rewriting code to help the analyzer forward.
But, as someone who doesn't think/deal with data much, this is a surprise to me. It makes sense though. Does this mean our data is forever tainted, as were?
Since you've been proximate to this work: do you think there's any hope in flipping the script such that possessing/processing this data is a liability rather than an asset? That's one of the few ways I can see to align incentives between highly monied-interests and the targets of data collection. (I'm doubt this will happen in the US in my lifetime, barring a major, 9/11-level event where abuse of this data is directly tied to the catastrophe.)
I believe this project was in response to the EU's Digital Services Act ("DSA"). Now IANAL but it always struck me as ambiguous as to what constitutes "personalization". Like, can your data be used to train an ML system? What if identity is scrubbed from that data? What if it's aggregated in some way so there's no individual user activity at all?
It also raises questions about every aspect of product behavior. Let's say you see a video post on your news feed. You get shown a snippet of the comments. Usually this is from friends but maybe it's just a comment that we thought might interest you. Is that personalization? What about a shared link where a snippet is shown from the article?
Also, the DSA seems to apply to on-site contextual personalization. What about off-site? Is that covered? Is there a separate law for that? Can I use derived demographics from a display ad pixel but not what pages you've liked and what groups you participate in on site?
Can I use your friends' activity to customize your experience?
The list goes on.
I'm a fan of clear legislation. If the goal is, for example, to allow opt out of personalized ads based on contextual on-site behaviour, we should enumerate what applications (eg feed ads) are covered and what information you're allowed to use if someone opts out.
The problem is that if you don't narrowly target AI you bump into "well this is what adtech does" and getting legislation that hurts the profits of one of the US's golden calfs is a nonstarter.
I have a camera that can see the street which meets these requirements, and would definitely go through extra steps to inform passers-by that I've taken steps to keep data about them out of the cloud. I just don't know what those steps ought to be.
This is reason to rush, not to decide somebody else will do a better work.
>> The people who nominally you want to be enforcing this "right" are also some of the people who most want to AI profile you.
If your goal is specifically to have someone enforce a rule against themselves, that's not an important issue, because it won't work.
If you change the statement of your goal so that you're not requiring the enforcement power to be vested in the target of the rule, then the issue might become more important.
The problem is that we(as humans) label the output of these algorithms to give them meaning.
For example if you build a resume grading AI based off historic data(as some companies have done) it's quickly realized that what's being outputted by the algorithm is typically not the same as what people label it. In this case the output is "how likely would this candidate been to have been hired using previous methods", but it's being used to try and answer "how good of a candidate is this".
IMO it's these discrepancies between what these algorithms are doing(which in more complex cases we can't know) and what we label the output that causes these issues. If we acknowledge that historic data can't be used to accurately predict future outcomes(especially a subset of historic data), we wouldn't even try to do things like profile candidates.
More data is needed before thinking you can determine the formula.
More importantly, it is trivial to understand the reasoning behind assuming a higher probability of someone repaying debt if they have a history of successfully repaying debts.
The same as anything else in life.
There are a lot of very specific assumptions in there about the mechanism.
How do you know which dates' balances are relevant to the calculation? If your credit score is 800, then you probably don't have a "credit utilization" of zero, as it's widely believed that the optimal % is more than zero (not "too low" but also not "too high", whatever that means).
The optimal ratio is very close to 0[1]:
+-----------------------+---------------------+
| Range | Average Utilization |
+-----------------------+---------------------+
| 300-579 (Poor) | 82.1% |
| 580-669 (Fair) | 56.1% |
| 670-739 (Good) | 35.2% |
| 740-799 (Very good) | 14.7% |
| 800-850 (Exceptional) | 6.5% |
+-----------------------+---------------------+
Also, as a bonus, the utilization component doesn't have history. If you're about to take out a loan, you can just borrow/pay back as much debt as you want the previous month for an ideal utilization.
[1]: https://www.experian.com/blogs/ask-experian/credit-education...
> Now things get a little trickier. Paying your balances in full each month isn't the same as maintaining 0% utilization. Here's why.
> Credit scoring systems calculate utilization using balance information that card issuers report monthly to the national credit bureaus (Experian, TransUnion and Equifax). Each issuer reports balance information on its own schedule, and many report to different bureaus on different days of the month. Each credit bureau also has its own timetable for revising your credit report once it has received a card issuer's update.
> For these reasons, if you use your credit cards at all, your utilization can vary from day to day at any one credit bureau—and it will differ from one credit bureau to another, even though all of their records are accurate.
> Here's a simple example:
> Let's say you use a credit card with a $5,000 credit limit and zero balance to make a $500 purchase on the 10th of the month. You then pay that balance in full on the 20th, before the charge even appears on your statement. If the card issuer reports your balance information to Experian on the 15th, then credit scores based on Experian data will reflect 10% utilization for that card on that month. Meanwhile, another credit bureau that gets updated on, say, the 25th will reflect 0% utilization for that card.
> Factor in multiple cards and balances, and you can see that your utilization on any given day is something of a moving target, and so are credit scores based on it. (The normal differences between credit scores based on data at different credit bureaus is one reason many lenders use more than one credit score when processing loan or credit applications.)
> Put another way, the only way to be sure you have 0% utilization all the time is to refrain from using your credit cards at all...
They also won't knock you if you are a few days late. I have been late a handful of times before and never took a hit. They will also reimburse the late fee if you call, but obviously they won't do this often.
If you're trying to say that you should avoid credit altogether, that's a terrible idea... Unless you plan on not getting a mortgage and paying for your home in cash. In which case your personal financial situation might not be representative of the average / median Joe.
If you build a robust budget and spend within that, you will have an 800 score. Do that. Do it intentionally.
Credit agencies evaluate risk, so even if you can stomach that $1200/mo car payment on your $50k/yr job. You are gonna take a hit for making a stupid financial decision, because even if you make those payments, you are almost certainly teetering on the edge to do it. (never mind that your score was probably already bad given the thought process to take on that loan).
Banks don't want to loan to people who are lucky and always somehow pull it out. They want to loan to people who are smart and reliable. Credit agencies try to filter for that.
I don't think that's how credit scores work, at all.
Also, it's not actually true that everyone uses credit scores. More recently companies use custom risk models because they see this as a competitive advantage. (Especially the afterpay companies like Klarna don't use them.)
Please show me the calculation methodology of any major credit scoring agency.
Rough guidelines aren't cutting it.
So, F this line. It's only true if it also includes "...and intentionally aquire and actually use a couple of credit cards that you don't want and don't need, for no other reason than to go through the motions." And even then it still doesn't result in an exactly great credit score if you always pay in full and never carry a balance. Even after years. 10 so far for me since the above. It's ok, but not great.
So yeah, F this line.
On the other side, you get benefits and money back for using your credit card. So people are in a situation where they really should use a card for everything. It also makes it easier to get good credit as long as your treat it like cash.
Paying off cards in full each month is the absolute best thing to do for your credit score (and saving money by not paying interest).
I think it would depend on your credit utilization %. If your credit utilization is too low, it may hurt your score. Of course if your credit utilization % is too high then it's an (even worse) issue. But who really knows? It's not like we have the formula.
Of course, who really knows? And the fact that we can't really know is the real problem IMO.
I'm saying that credit scores being a black box is inherently not a good idea. There can never be any accountability, because there is no visibility. In general, it's impossible to verify the correctness of a credit score. It's also impossible to discuss the fairness of the scoring methodology, because all we can do is speculate and guess.
For something so impactful to an individual, you'd think there would be more government oversight / regulation... And more transparency.
If you've never been a victim to this then you don't understand how pernicious this system is. Once one creditor does this, then all the other creditors automatically operate on that information. They don't double check, they don't notice the strangeness of the report, they don't care. Your credit changed, so your limits changed, so they want their money back NOW.
It's a nightmare.
We can pass all the laws we want trying to put this genie back in the bottle. It may work, some, it might keep the law-abiding government agencies at bay (are there any)?
But anybody wanting to make a buck off of us, they won't hesitate to use it. There's money in it. You refuse to do it, well, you lose in the marketplace. Social Darwinism. Or, economic.
We have to come up with some social rules to control how we feel about it, how we are subjected to it. That could help. Like, you don't mention what you saw on a streaming microcam that drifted through your neighbor's bedroom window, because that is gauche. Some fiction of privacy. Like not mentioning what you hear through a bathroom door. It's just not polite.
That's the best scenario we can hope for, I imagine.
A. We don't fully understand this and our responses are incorrect. Let's build an AI system to help us.
B. We'll give it data from online which is limited, temporal, and to the same degree as humans; incorrect.
C. Then we'll build an algorithm based on the generalized structure of the human brain, which, is regularly wrong.
D. We don't fully understand this and our responses are incorrect.
Not just you, I agree completely.
Lime Torrents still exists.
Annas Archive still exists.
Where do we draw the lines? Do we all end up walking around cloaks and masks in public and if you suss out who the faceless individual is next to you in the crowd it's your fault you figured it out?
The funny part is that deleting/hiding some piece of your personal data often requires keeping a record of it so they know what to delete/hide.
In fairness, so is the right to deletion.
Once your data is in some of these systems, it's sold on to other systems immediately. (Most of the "sell side" market is set up on the basis of subscriptions). So, for instance, your data gets written to some merchant system or whatever, and it's slurped out in the pull that happens 15 minutes later by data subscriber 531962. Unless you get your request to delete your data to the merchant customer support within that 15 minute time period, your data has already been pumped out into the ether.
Make it so you have to give affirmative consent to opt out of right to deletion? Great idea!
Of course, most of these companies slip the opt out language into the Terms of Service and Privacy Statement.
You, uh, you did read them before you clicked 'OK', right?
(BTW, you likely suspect that it doesn't take 15 minutes for these automated processes to slurp up new data. And, yeah, your suspicion is correct, on most of these systems, it takes far less than 15 minutes before one of their subscribers pulls new data. So. Yeah. Fun Times!)
I point this out because I was liking the argument until this point: He's off by a couple orders of magnitude, according to Wikipedia. Not enough for me to read it as absurd hyperbole, like calling Adam Smith a communist, but enough for me to stop reading the essay, look up his claim, and then feel like he's either misinformed or lying.