Police in England installed an AI camera system along a major road. It caught almost 300 drivers in its first 3 days.::An AI camera system installed along a major road in England caught 300 offenses in its first 3 days.There were 180 seat belt offenses and 117 mobile phone
Photos flagged by the AI are then sent to a person for review.
If an offense was correctly identified, the driver is then sent either a notice of warning or intended prosecution, depending on the severity of the offense.
The AI just “identifying” offenses is the easy part. It would be interesting to know whether the AI indeed correctly identified 300 offenses or if the person reviewing the AI’s images acted on 300 offenses. That’s potentially a huge difference and would have been the relevant part of the news.
The system we use in NL is called “monocam”. A few years ago it caught 95% of all offenders.
This means that AI had at most 5% false negatives.
I wonder if they have improved the system in the mean time.
Nobody cares about false negatives. As long as the number isn’t something so massive that the system is completely useless false negatives in an automatic system are not a problem.
What are the false positives? Every single false positive is a gross injustice. If you can’t come up with a number for that, then you haven’t even evaluated your system.
The system works with AI signaling phone usage by driving.
Then a human will verify the photo.
AI is used to respect people’s privacy.
The combination of the AI detection+human review leads to a 5% false negative rate, and most probably 0% false positive.
This means that the AI missed at most 5% positives, but probably less because of the human reviewer not being 100% sure there was an offense.
Look, I’m not saying it’s a bad system. Maybe it’s great. “Most probably 0%” is meaningless though. If all you’ve got is gut feelings about it, then you don’t know anything about it. Humans make mistakes in the best of circumstances, and they get way, way worse when you’re telling them that they’re evaluating something that’s already pretty reliable. You need to know it’s not giving false positive, not have a warm fuzzy feeling about it.
Again, I don’t know if someone else has already done that. Maybe they have. I don’t live in the Netherlands. I don’t trust it until I see the numbers that matter though, and the more numbers that don’t matter I see without the ones that do, the less I trust it.
The fine contains a letter, a picture and payment information. If the person really wasn’t using their phone, they can file a complaint and the fine will be dismissed. Seems pretty simple to me.
However, I have not heard any complaints about it in the news and an embarrassing amount of fines has been given for this offense.
For a post on a site like this that kind of anecdote is plenty to add to a conversation, and it does actually make me feel a tiny bit better about the whole thing, but when you lead with statistics you’re implying a level of research and knowledge that goes beyond just anecdotal. It’s not really fair to you or any of us, but using the numbers that sound good to avoid using the ones that reveal flaws is one of the most popular ways for marketing teams and governments to deceive people. You should always be skeptical of that kind of thing.
How do they know that they caught 95% of all offenders if they didn’t catch the remaining 5%? Wouldn’t that be unknowable?
Welcome to the world of training datasets.
There are many ways to go about it, but for a limited number they’d probably use human analysts.
But in general, they’d put a lot more effort into a chunk of data and use that as the truth. It’s not a perfect method but it’s good enough.
The article didn’t really clarify that part, so it’s impossible to tell. My guess is, they tested the system by intentionally driving under it with a phone in your hand a 100 times. If the camera caught 95 of those, that’s how you would get the 95% catch rate. That setup has the a priori information on about the true state of the driver, but testing takes a while.
However, that’s not the only way to test a system like this. They could have tested it with normal drivers instead. To borrow a medical term, you could say that this is an “in vivo” test. If they did that, there was no a priori information about the true state of each driver. They could still report a different 95% value though. What if 95% of the positives were human verified to be true positives and the remaining 5% were false positives. In a setup like that we have no information about true or false negatives, so this kind of test setup has some limitations. I guess you could count the number of cars labeled negative, but we just can’t know how many of them were true negatives unless you get a bunch of humans to review an inordinate amount of footage. Even then you still wouldn’t know for sure, because humans make mistakes too.
In practical terms, it would still be a really good test, because you can easily have thousands of people drive under the camera within a very short period of time. You don’t know anything about the negatives, but do you really need to. This isn’t a diagnostic test where you need to calculate sensitivity, specificity, positive predictive value and negative predictive value. I mean, it would be really nice if you did, but do you really have to?
Just to clarify the result: the article states that AI and human review leads to 95%.
Could also be that the human is flagging actual positives, found by the AI, as false positives.
You wouldn’t need people to actually drive past the camera, you could just do that in testing when the AI was still in development in software, you wouldn’t need the physical hardware.
You could just get CCTV footage from traffic cameras and feeds that into the AI system. Then you could have humans go through independently of the AI and tag any incident they saw in a infraction on. If the AI system gets 95% of the human spotted infractions then the system is 95% accurate. Of course this ignores the possibility that both the human and the AI miss something but that would be impossible to calculate for.
That’s the sensible way to do it in early stages of development. Once you’re reasonably happy with the trained model, you need to test the entire system to see if each part actually works together. At that point, it could be sensible to run the two types of experiments I outlined. Different tests different stages.
I think 95% were correct reports is what they mean. There could be a massive population of other offenders that continue sexting and driving or worse. One monocam won’t ever be enough we need many monocams. Polymonocams.
I suspect they sent through a controlled set of cars where they tested all kinds of scenarios.
Other option would be to do a human review after installing it for a day.
but digging out that info would involve journalism and possibly reporting something the cops wouldn’t like! We all know how that goes.
Yeah that’s a shame, would be really interested to know those numbers.
Expansion of the surveillance state, you go uk.
The DEA in the US has surveillance on every highway exit. If you drive on an interstate, you are logged.
Wait really? Via cameras?
yea, in fact most toll booths are camera operated up north here now, they click a picture of the plate then mail you the toll bill + added expenses for the mail in system. Annoying af but it’s either that or get EZ pass(or not use the turnpike).
it flashes multiple times on the way through so I’m sure it grabs a front picture as well
I love threads like these because it really shows how flexible opinions are, post about ai surveillance state and everyone is against it but post about car drivers getting fined for not wearing a seatbelt and everyone loves it.
This is a weird phenomenon. Feels a bit like how focusing on “welfare queens” / “dole bludgers” can pave the way for similar privacy erosion (and welfare cuts) even though its a tiny percentage of the people. Seems a short hop away from “if you’ve got nothing to hide…”
Except in this case being a poor driver actively puts others at risk rather than just being a drain on tax money.
Seatbelts I don’t really care about, because with that people mostly just affect themselves (or others in the same car), but for other infractions it makes sense.
The real issue is whether you can trust that the data will only be used for its intended purpose, as right now there are basically no good mechanisms to prevent misuse.
If we had cameras where you could somehow guarantee that - no access for reason other than stated, only when flagged or otherwise by court order, all access to footage logged with the audit log being publicly available, independent system flagging suspicious accesses to any footage, etc. - it wouldn’t be too bad.
Compared to all the private cameras that exist in cars these days…
You know the best way to not have absolute power corrupt? Not have absolute power.
If you collect this data there is degree of probability that eventually it will be abused. If you don’t collect this data there is zero chance.
Some > none
Good government is about assuming the worse and decided if you are willing to endure that. If the absolute worse humans you can imagine were put into office how much bad can they do?
Surely the ultimate come away from that is will not ok with people breaking the law and we’re not ok with AI taking people’s jobs. There is no conflict here
So you think most people like the idea of a surveillance state automaticly enforcing it’s every whim with perfect efficiency?
I’m pretty sure that’s something pretty much universally disliked
You don’t know many authoritarian.
Course, they only think this should apply to everyone elae.
I don’t think I said that did I
In it’s current form it’s good technology. It’s all fine as long as you’re chasing after crimes we all agree are bad* It’s the slippery slope I’m worried about. Just a matter of time untill this is going to be used for something malicious we don’t agree with.
*I don’t care if front seat passengers wear a seatbelt or not as long as they’re adults.
The slippery slope is what makes this not okay. It’s a completely unnecessary invasion of privacy in the guise of “safety”.
I’d love to see some statistics showing that these things are anything other than an additional tax on the drivers. This is bad for everyone and it desensitizes you and opens the door to further surveillance I’m the future.
“Slippery slope” is a common argument but usually flawed. In this case, driving is an extraordinarily regulated privilege and despite that, it still results in massive deaths and permanent life changing injury every year. In the US, car crashes are the number one cause of death for children. It’s difficult to draw a line between expanding driving enforcement to gross losses in privacy like many here are envisioning.
It also ignores the benefits to civil rights. Again, I don’t know about the UK but in the US, traffic enforcement by police is very unevenly applied. Minorities routinely get their privacy violated on pretexts while cops don’t even pay lip service to the rules.
Just as an aside, gun violence is now the leading cause of death for children in the US; vehicle collisions are now 2nd, due to gun violence increasing and vehicle collisions decreasing.
It isn’t though.
It isn’t unnecessary invasion of privacy. You have no expectation of privacy when driving around on public streets, and to say you’re allowed to break the law and use personal privacy as an excuse is absurd.
I just wish they would have one where I live to fine all the people using the HOV lane who aren’t supposed to be
Then we watch the numbers plummet and see there’s only actually 5% of people using the lane and finally see how useless the hiv lane is so we can just make it a regular third lane.
The HOV lane is supposed to look empty. If it was packed full of cars, carpooling wouldn’t have any advantage because you wouldn’t go any faster.
It doesn’t work that well around here, cause there’s inevitably that one car that refuses to go faster than the rest of the traffic that it’s separated from. Or slows down to 10mph when the rest of the highway is stop and go, despite there being a barrier. Then someone gets rear ended because no one was expecting the lane to be going 10mph (and were on their phone), and the accident closes down the lane entirely
Basically, by me, the HOV lane is slower than traffic 90% of the time. Even in stop and go, because that lane is actually the one containing the accident causing the traffic.
Well, uhh, sounds like you could use some more traffic enforcement there. Maybe with AI and cameras ;)
I thought the advantage of carpooling was saving money on gas and car maintenance. Also, environment.
I am not pro seat belt laws. It is your life and you should be able to throw it away if you want.
Not wearing a seatbelt reduces the security of others. If you want to throw it away, that’s a different matter and should not be handled through seat belt laws.
Is the freedom to drive without feeling like you’re being watched more important than the prevention of texting while driving?
During my commute, it’s common to see people looking at their phones. I don’t know what the effect is without statistics, but seeing an accident along the way is a usual occurrence.
inattentive driving should be considered gross negligence
I’m more concerned about error rates and false accusations
Doesn’t it say that each image is sent to a human for review before any charges are laid? Might not be the case forever, but at least for now it’s actually a human who ultimately decides whether or not to prosecute a driver.
That’s the important part for me. As long as the whole process isn’t automated I’m fine with it.
This has always been the case for road cameras in the UK from the start from when we first had speed cameras introduced, before they are sent out they are (supposed to be) reviewed by a person first to check for false positives.
No. Your freedom to feel feelings is your problem. If you feel like you’re not being observed right now, your feeling is already wrong.
ITT a bunch of people who have never read an ounce of sci fi (or got entirely the wrong message and think law being enforced by robots is a good thing)
But the law isn’t enforced by robots the law is enforced by humans. All that’s happening here is that the process of capturing transgressions has been automated. I don’t see how that’s a problem.
As long as humans are still part of the sentencing process, and they are, then functionally there’s no difference, if a mistake is being made it will be rectified at that time. From the process point of view there isn’t really any difference between being caught by an automated AI camera and being caught by a traffic cop.
Although completely reasonable, I fear that your conclusion is inaccessible for most folks.
And as a pedestrian, I’m all for a system that’s capable of reducing distracted driving.
How to disincentivize a motorist public is to make driving a stressful affair- currently, it’s other people. Soon, it’ll be catalogs of minor infractions caught, at the millisecond intervals they occur in, forever and the bill to pay it showing up every single week for the rest of your driving lives. Odds are it’s going to be scrapped, made a Boogeyman for a while, and then come back every time people get testy about gas prices
Sounds like youd support precrime too. Stop licking boots
What the hell is precrime?
Calling an image recognition system a robot enforcing the law is such a stretch you’re going to pull a muscle.
It’s going to disproportionately target minorities. ML* isn’t some wonderful impartial observer, it’s subject to all the same biases as the people who made it. Whether the people at the end of the process are impartial or not barely matters either imo, they’re going to get the biased results of the ML looking for criminals so it’s still going to be a flawed system even if the human element is OK. Ffs please don’t support this kind of dystopian shit, Idk how it’s not completely obvious how horrifying this stuff is
*what people call AI is not intelligent at all. It uses machine learning, the same process as chatbots and autocorrect. AI is a buzzword used by tech bros who are desperate to “invest in the future”
deleted by creator
Face recognition data sets and the like tend to be pretty heavily skewed, they usually have a lot more white people than poc. You can see this when ML image filters turn black people into white people or literal gorillas. Unless the data set properly represents a super diverse set of people (and tbh probably even if it does), there’s going to be a lot of race based false positives/negatives
deleted by creator
That might be the case tbh, but either way that would be bad and discriminatory. I might just be overthinking it, it might not actually be that bad, but I know discrimination like that is super common when it comes to how recognition-based ML is trained
But how is that different or worse from a human sitting at the side of the road and writing down number plates for example?
The image recognition system detects a cell phone being used and snaps a photo, records the plate number, etc. How exactly does that lead to racism?
You’re making what amounts to a slippery slope argument, and that’s often a very flawed way of thinking.
According to Sci-fi organ transplants will lead to the creation of monsters who will kill us all for “tampering in God’s domain.”
Maybe fiction isn’t the best way to determine policy…
I for one base ALL my global policy on sci Fi novels 🤦♂️
Since the writers are on strike we can have them just write the entire legal code as the writers of black window are actually taken seriously beyond nerds for once.
I work in an adjacent industry and got a sales pitch from a company offering a similar service. They said that they get the AI to flag the images and then people working from home confirm - and they said it’s a lot of people with disabilities/etc getting extra cash that way.
This was about six months ago and I asked them, “there’s a lot of bias in AI training datasets - was a diverse dataset used or was it trained mostly on people who look like me (note: I’m white)?” and they completely dodged the question…
(this is definitely a different company as I am not in England)
Yes does the AI automatically send every taxi through or is it only when they are on the phone. Has the AI ever seen a taxi driver who’s not on the phone in order to check this?
“Hookay thanks for the presentation fellas, but lemme ask ya: Was your model trained only on iPhones or was a diverse palette of plastic Android phones from the last 15 years also taken into account?”
The UK has never seen a dystopian nightmare they didn’t rush to embrace.
There is a strong authoritarian undercurrent in England.
I’m not against this. I think the fact that car deaths are skyrocketing in the US and the UK is even more absurd since modern cars are supposedly “safer” with all of their safety tech. Plus how are people still doing this fucking shit when death from dangerous driving has been a thing in the news forever now? It’s like people need even more stricter rules to keep them in line instead of thinking like a reasonable adult.
I think deaths jumped a bit post COVID but I don’t think they are skyrocketing. Do you have a source?
I looked it up. They aren’t skyrocketing.
The numbers dropped due to lockdown, then bounced up and are stable.
I hate this cult of negativity - just make up how everything is getting worse in order to hand more power to the government.
The casual and bovine l way it all happen is disgusting.
deleted by creator
I think the fact that car deaths are skyrocketing in the US and the UK is even more absurd since modern cars are supposedly “safer” with all of their safety tech.
SUV vs. Bicycle.
100% agree. It flags infractions, you have people verify what was being flagged, due course follows.
For reference, in Switzerland deaths/major injuries from traffic accidents have steadily dropped since the '70s. Thanks to, as you mention, better car safety tech.
But there has also been a great number of speed cameras and lower alcohol tolerance. Oh and new laws with income-relative fines, temporary to permanent loss of driving license, and even jail for the worst driving offenses probably cooling the jets of even the wealthier road maniacs.
There is a name for that sort, the safer the item is the more reckless the person becomes.
Yup, and some of these are quite serious. But a cop at the side of the road could stop these people instantly. These people won’t find out that they have broke the law for two weeks. Or they could just kill themselves/someone else/both half a mile up the road.
No they wouldn’t. You are telling me you could stop someone on the motorway instantly. You think a stationary cop at ground level would be able to spot a phone held below the window and have the reaction times to intiate a persute?
I never suggested any of this…alls I said was a cop at the side of the road could stop a car. I didnt say we couldnt have a copper parked up on a bridge as lookout or use these cameras.
deleted by creator
Fuck ai and cameras and uk.
FUCK CAMERAS. THAT’S WHY I VOTED BREXIT. SO I CAN TEXT AND DRIVE, NOT LIKE THESE BAGUETTE LOVING SISSIES
Monitor my internet and tell me what I can do harder daddy
What?
I hope it’s sarcasm
It is sarcasm. Britsbongs are too indebted to hang on sync and they’d rather sniff nitrous under a bridge
Meanwhile in New Orleans
https://www.fox8live.com/2023/07/26/nopd-use-facial-recognition-leads-zero-arrests-nine-months/
Really great dialogue and discourse going on in this post. Thank you everyone for your opinions and viewpoints. Definitely have a lot to think over on my current stance. Exactly what I was missing lately from the social media I’ve been consuming (actual discussions with merits both sides hold).
My main problem with this is, that this becomes like the huge online behemoths like youtube etc. I think most people have seen incidents where youtube cancelled a channel or applied copyright incorrectly, and getting a human to review things is next to impossible. The reason is clear, the sheer amount of content breaching the rules is too big to cost efficiently deal with by humans.
One camera catching 300 people in 72 hours. We don’t see how many it triggered, how many were reviewed and found to be false positives.
The problem is going to be if a whole police force takes it up, or it goes national. The amount of hits generated would be far beyond the ability to confirm with humans. I see it going a similar way to youtube. They just let the AI fine people. You report it as wrong, so they send your petition to another AI that pretends to be human and denies you again. The only way to clear things up is to take it to court. But, now the court system is being flooded so they deny people the right to a court case and the fixed penalties will be automatically applied.
This is the dystopia I fear. Actually catching people committing driving crimes? I don’t have a problem with that. Aside from maybe the increasing number of driving crimes coupled with the knowledge these cameras exist could lead to less concentration while people make sure they’re sitting upright, looking attentive, eyes straight ahead hands at 10 o’clock and 2 o’clock. Did I indicate for that lane change back there? I guess that remains to be seen.
Haha, that’s a scary thought. But not unreasonable. Fine first and let the recipient proof they are not at fault,fighting through a series of AI entities.
“You’re through to the AI’s AI Manager how may I reject your complaint?”
Why are people saying this is a hypersurveillance dystopian nightmare? Guys, you are still in public! The only difference between this and having police officers sitting there and looking is this is much cheaper and more efficient. The recordings are still being sent to a human being for review.
The problem is the whole “give an inch, they take a mile.” We don’t know what rights this may take away from us in the future. So in the now, always question
Yeah I understand this argument. In my mind there is no anonymity when driving, (and in my mind there shouldn’t be) and the responsibility you have as a driver have that makes this permissible.
A valid and reasonable point. The problem is that often it spills out of it’s original intent. The “think of children” argument springs to mind
The only difference between this and having police officers sitting there and looking is this is much cheaper and more efficient.
Sure, but that’s a huge problem, because the legal system wasn’t actually designed for perfectly efficient enforcement. It is important that people be able to get away with breaking the law most of the time. If all of the tens of thousands of laws on the books were always enforced we would all be in prison and bankrupt from fines. Some laws are just bad too, and the way they get repealed is when enough people get away with breaking them for long enough to build political momentum for it.
Also, it isn’t like they are going to stop at using scaled-up AI surveillance just to enforce seatbelt use and texting while driving, there is way too much potential for abuse with this sort of tech. For example if there are these sorts of cameras all over, networked together, anyone with access to them can track just about everything you are doing with no way to opt out. Even if you aren’t doing anything wrong the feeling that you are always being watched is oppressive and has chilling effects.
A day in your life must be horrible, friend.
Am I the only one who considers the text on the camera car (“HIGHWAY MAINTENANCE”) a bad joke?
I’m surprised car companies haven’t already partnered with governments to have the vehicles themselves snitch on the occupants. Why install these camera systems all over the place when the vehicles themselves collect ridiculous amounts of data with greater accuracy? I’m sure the car companies would love the additional revenue stream and the governments would love the greater surveillance capabilities.
Cause no one would buy a car from a company that does this
Probably because they wouldn’t see a dime of revenue from this. It would be a new law that just says they have to do it. At best, they would be allowed to pass the costs to customers somehow, likely through our plate registrations at the DMV.
It’s basically a no win for the car companies. Lots of ill will, increased chance of litigation, increased costs for building cars, all for nothing.
In fact, I bet the car companies lobbyists are the reason we don’t have this already.
