Yet...
On the everyday news, in the popular press etc., self-driving cars are here, now.
Disconnect, much?
Brits suffering through the nationwide KFC famine can enjoy with wry amusement the fact that an AI can be fooled into thinking an image of Colonel Sanders and the restaurant's logo are a stop sign. The fast food famine arose after KFC UK last week switched from logistics provider Bidvest to rival DHL. The result was delivery …
One would think that, given the advances in GPS and mapping, that any self-driving cars need to stop 'looking' at traffic signs and have the relevant information already encoded in their maps. There is no complex recognition required, and regular updates can be done (even on-the-fly through mobile ata, same as 'live-ish' traffic updates). Local temporary changes could be signalled through a radio transmitter beaming the local changes to any vehicles passing by.
Of course there still would be security issues, in particular the car would need to be hardened gainst attacks to the update system and to the 'local warning' system. BUT these would work on quite specific and well-known channels - encrypting and authenticating small packets of data is actually a well-advanced technology even if in many cases it's woefully applied. It's certainly more easy to secure than relying on an artificial "Intelligence" that is based on a neeural net where even the designers don't know exactly how recognition/classification decisions are being made
Given that my modern and newly updated gps, unaware of the queensferry crossing. urges me to drive through a concrete divider and down an 45% embanking in a desperate attempt to get to the Forth Road bridge every time I drive to Edinburgh I'd personally pass on relying on gps/mapping solutions, Along the same stretch of road it frequently thinks I'm driving on parallel non-motorway roads and there are so many cases of fleshware drivers blindly following out of date directions to drive the wrong way up one way streets etc that I'd be factoring in a hefty amount of scepticism of GPS in any AI I was building.
"One would think that, given the advances in GPS and mapping, that any self-driving cars need to stop 'looking' at traffic signs and have the relevant information already encoded in their maps. "
Just plough past the guy holding the stop sign at the road works then? After all, it's a 60 mph straight bit of road.
This post has been deleted by its author
Signs that don't move could be made into RFID "signs" which would be more accurate that using machine learning. Temporary signs and traffics lights could also be fitted with RFID. It wouldn't entirely eliminate the problem as cars need to be able to spot things like Police stopping traffic, animals running out etc.
RFID in its typical form would then depend upon every self driving car betting fitted with an RFID Reader. Readers "illuminat" the tag with RF (e.g. 13.56MHz or 125kHz) to power it up. Given the distances to traffic signs, the vehicles might need the 50,000 watt version of the RFID Reader.
Perhaps there are other options.
the vehicles might need the 50,000 watt version of the RFID Reader
I'm sure Photonic Induction could cobble something together that would do the job. Althought 50 kW is probably a low estimate of what he'd come up with.
Yep, I have given this some thought. You would place visual markers on lamp posts or install transmitting ones that map the road layout. These would need to be interconnected and if they weren't working or incorrectly placed the car should then report this back to the company tasked with fixing them. As there are multiple lamp posts on both sides of the road the car would be able to confirm if one was incorrect and not working as they would create redundancy for each other on both sides of the road. It maps the road itself before the car gets there and is constantly mapping for changes, pedestrians and other cars. The perfect back up for the cars own sensors.
Sure this only works where there are lamp posts but it would be a start.
So what happens when a construction worker forgets to turn off the stop sign after packing up. Every autonomous vehicle sees his van as a stop sign while he drives back to the yard.
Or someone could decide that it would be fun to hide one of those stop signs beside a major motorway. Instant long tailback while the plod search for a small device in the undergrowth.
Unfortunately, updating map databases tends to be VERY labour intensive, as any change can require very extensive QA checks to ensure the integrity of the whole database. It took the regular sat-nav databases for the UK about 2 years to include the newly built road on which I live - and Google Maps still hasn't got it after 6 years!
Further, it would be impossible to react in real time to temporary speed limits such as those imposed by road works - which might only be there for a few hours, or even be constantly moving in the case of works such as renewal of white lines etc.
There will always be a need for some sort on local indication of speed limits etc., so as to cope with these situations.
Unfortunately, updating map databases tends to be VERY labour intensive, as any change can require very extensive QA checks to ensure the integrity of the whole database. It took the regular sat-nav databases for the UK about 2 years to include the newly built road on which I live - and Google Maps still hasn't got it after 6 years!
Same with speed cameras - driving on motorways, satnav will warn about 50mph speed cameras. I'm assuming that was down to roadworks and speed restrictions that have long since gone....
One would think that, given the advances in GPS and mapping, that any self-driving cars need to stop 'looking' at traffic signs and have the relevant information already encoded in their maps.
Driving in Spain is fun with GPS. There's the "Where the f**k, did that roundabout come from? It's not on my satnav!" and there was a stretch of road where I would have to do the complete opposite of the satnav.
"Turn left"
"No, you stupid bitch, there is a no-entry sign, I have to turn right"
"Turn around now"
"How???? I'm on a dual carriage way with a big f**k off concrete barrier down the middle - wait till we get to the roundabout signposted ahead!"
Rearranging anything to be misinterpreted as a hexagonal red sign on an intersection (or not as such) takes some doing. What's inside that hexagonal sign isn't as important as the shape and colour.
Ditto an upside down triangle on an intersection. (yield/give way)
The lines on the road give an important secondary clue,
At least some of these attacks are bring over/under thought and OCR is for hoomuns more than machines.
Why would autonomous cars (when they arrive in 2080) need to read road signs? Sure they have to read the road, looking out for wayward pedestrians, footballs, cats and the like but surely all set variables (like speedlimits, stop sign location, etc.) will be coded into the maps they use. I don't get this.
Just as i finished typing the sentence last sentence it occurred to me: roadworks but then again surely any roadworks could have custom (doesn't have to be a sign) transmitter to inform the car of the rules of traversing any ongoing works. This just seems like another "look how we can confuse weak AI" schtick. Be more interesting to read about how they solve these problems.
This is all just a bit "got your nose"
You know how your satnav often gets it wrong due to old data? This is the same * 1000. Autonomous vehicles have to be able to see and understand traffic signs.
Also if roadsigns, even just the temporary ones, had to use custom transmitters they would all need replacing or modifying.
Don't use the satnav mate, but I understand that maps on the device need updating.
I think the whole idea of trying to make a car (AI) understand the world around to perform the simple task of driving is nigh on impossible and thats before you've got to the interesting morality questions that have been posed. e.g. When the car has no option but to crash does it place the lives of the car occupants above those on the street.
If the world does want autonomous cars then I believe there will have to be changes to the infrastructure to make sure they can work reliably and safely. How does a visual AI check that what it thinks it is seeing is correct (as someone pointed kids with spraypaint, we get it all over here). I guess it would check a known database of good data which I would expect the manufacturer to update as required.
Since roadworks are by their very nature temporary (supposedly) then this would not mean replacing All road signs (not my point) just providing a way to interact with the car without having to rely on Visual AI for exceptions such as roadworks.
if they crack a self driving, morally upright car that needs no updates, set road data, can master driving different conditions, no infrastructure changes etc, etc, etc, etc, etc then they might as well build Skynet.
"morality questions that have been posed. e.g. When the car has no option but to crash does it place the lives of the car occupants above those on the street."
This is a straw man. The trolley problem is a thought exercise to which there is no "correct" answer. It's not even known for any given human driver whether they would prioritise themselves / their passengers / anyone outside the car. Quite possibly, any human driver who THINKS they would react in a given way when it's a thought exercise could react in a different way when it's for real. So giving the car AI simple hardcoded instructions for collision avoidance priority will be sufficient.
There are many problems to solve to be able to get self-driving cars. Morality is not one of them.
"Hold the lever to keep the switch gear between the two tracks in an attempt to derail the train & thus save everybody."
Should be "Hold the lever to keep the switch gear between the two tracks in an attempt to derail the train & thus kill some, or all, of the people on the train".
Should be "Hold the lever to keep the switch gear between the two tracks in an attempt to derail the train & thus kill some, or all, of the people on the train".
It's called the Trolley Problem, not the Train Problem. Trains don't often come into conflict with pedestrians at points.
When the car has no option but to crash does it place the lives of the car occupants above those on the street.
Ultimately, yes. Even if that is not the default behaviour, people will produce modifications that will allow such decisions to be predetermined.
That's no different to a human driving though, is it? It requires truly selfless reasoning and balls that'd shame Superman to crash into a tree instead of a person in the road, if doing so is likely to kill you. Maybe I could bring myself to do it if its a child in the road and my child is not in the car, but if my child is in the car, then whoever or whatever is in the road is shit outta luck. Sorry. You'll not find any parent make any other decision either, despite the howls of protest that will doubtless follow.
I don't see why society expects AI to behave better than they would themselves - even if it did, they'd start changing it because they wouldn't like the outcomes - does your smart home dob you into the police because it can detect weed being smoked? Perhaps your smart home could tell your GP the truth about how much you eat, drink (alcohol), smoke, and your fitbit can explain just how little exercise you really do? Any limitations on such behaviours won't be technical.
Note to the utopians - people drive cars, and perhaps more importantly lorries full of stuff, across borders. It's going to be handy for autonomous cars to have some idea of what the human-controlled visiting vehicles are going to do in response to the locally perceived physical environment, including roadsigns, in real time - not just try to predict it based on a theoretical cyberspace model where the no-entry and one-way-street signs *haven't* all just been reversed by jolly rag week students.
not just try to predict it based on a theoretical cyberspace model where the no-entry and one-way-street signs *haven't* all just been reversed by jolly rag week students.
Or the town council want to try out something new like change all the one-way streets around.
"Or the town council want to try out something new like change all the one-way streets around."
Apart from during a transition period, that's probably a very good example of something that could become very easy to do. If the vast majority of cars are AV/AI and permanently connected, it would be obvious have a national mapping infrastructure such all cars have the same map updates, with pretty much real time access.
This would also pretty much eliminate the need for road signs.
Whether we'll ever get to that stage is another matter, but the transition period is going to require far more complex and expensive systems than will be required once (if!!) AV/AI vehicles become the norm.
Taking just the UK, if all road vehicles were AV and connected at all times, even the tiny little rural roads with 10ft hedgerows such as in the south west should be usable since each car will "know" about all the other cars in the area long before they come into view and be able to negotiate who goes where and when, maybe just adjusting speeds slightly so opposing cars arrive at passing places together or re-route, effectively creating dynamic one-way systems.
There was a story (was it here at the Reg?) about Tom Tom ceasing support for various models of SatNav, some of which aren't that old. I can't see support for built in units being any better, so am a bit sceptical about those, and whether they'll be much use once the vehicle is 5+ years old.
The satnav in my wife's 7 year old Landrover cheerfully tells us we're doing 60mph across a ploughed field each morning (they built the bypass 6 years ago). We could of course get the maps upgraded, but it's rather more expensive than the alternatives, so haven't bothered. However, if it was a legal requirement then we'd... probably sell the car. Most likely to someone who simply wouldn't bother, and take their chances. But to write off all that physical hardware, just for a software upgrade? It's a completely different set of equations for natural resource consumption considerations than we've ever used before.
"so am a bit sceptical about those, and whether they'll be much use once the vehicle is 5+ years old."
Hopefully this is something that will be addressed as AI car legislation is introduced. Manufactures will be legislated to provide update for the life of the vehicle. Hopefully the lawmakers will consider what the true "life" of the vehicle is and not assume that everyone gets a new car every year or two like they do. It's already difficult enough for the less well off to keep and maintain a car in the current MoT/anti-pollution regime we are in now. Someone running an old banger isn't really in a position to take advantage of "scrappage" schemes. The cashback/discount is no where near enough to being a new car price down to their level.
"I find it hard to believe that a $30,000 car can't match it."
I find that very easy to believe. I was using a hire car last week, and rapidly resorted to using my phone because of the poor quality of the GPS built into the car. (A Toyota Auris with a bit over 1000km on the clock when I picked it up.)
For the record, I was in Norway and my phone uses "Navman Navigator" and OpenStreetMap data.
Why would autonomous cars (when they arrive in 2080) need to read road signs?
Let me quote you the oldie but goldie from the other side of the wall dating back to the early 1980-es:
Protocol of the weekly meeting of the Police Union.
The Union met today and discussed the following agenda items:
1. Application by Snr Constable Petrov for financial assistance due to temporary hardship.
The Union decided:
1. To approve the application for financial assistance of Snr Constable Petrov
2. To issue Snr Constable Petrov a temporary movable STOP and 30km/h per limit stop signs.
"Just as i finished typing the sentence last sentence it occurred to me: roadworks but then again surely any roadworks could have custom (doesn't have to be a sign) transmitter to inform the car of the rules of traversing any ongoing works."
Again this is not the exact issue, it is the serious disconnect between the claimed ability of AI to deal with the real world that meat-bags do a half-passable job of in order to work for the decades it will take to transition from drivers to robots on the road.
If your AI can't tell a KFC and STOP sign apart, just how good/safe will it be? Even if you think "oh just put up a transmitter for robocars at road works" how will they deal with any other sort of outage/problem that meat bags could deal with by using their (admittedly often limited) intelligence to work around? Finally when there is a a fatal/serious injury crash involving a megacorp's robo car and some lawyer can show such shitty AI discrimination of the obvious, do you think the fines for knowingly unsafe design will be Ford Pinto style or not?
Why would autonomous cars (when they arrive in 2080) need to read road signs?
Well, two obvious reasons.
Firstly because they convey mandatory information to which the vehicle must adhere immediately, not once someone gets around to updating the map. Getting a dozen speeding tickets because what was a 30mph limit on Friday is now a 20mph limit on Monday and your map hasn't been updated yet, is going to be annoying, and expensive.
Secondly because non-automated road users (other cars, trucks, vans, pedestrians, cyclists etc) will be using the signs to inform their decisions - understanding what those road users are likely / supposed to do next will be a key part in not running them over.
The problem seems readily solvable - use a better image set for road signs, and make sure the car understands that not everything at the side of a road is a mandatory road sign - some are just adverts for the village fete, or hangover food.
Some signs cover a wide, changeable, and semi-unpredictable area. "Wild ponies or horses" being a good example, or riding schools. Many horses apparently confuse plastic bags stuck in hedges to be a vortex to another universe, and are likely to misbehave. It'd be best for all if the AI doesn't assume the horses will be in the fields, or moving slowly at the side of the road.
Temporary road closures for say road races (running etc) won't always be known about in advance in CA. or wherever the update is done, and if you're relying on humans to ensure the maps are correct at the time of use, then you're going to get human error. Anyone having lived in Newcastle-under-Lyme will be very familiar with truckers getting lost looking for Newcastle-upon-Tyne, at least in the days before GPS. People make mistakes, so cars relying on maps updated or created by humans are going to need to drive to the conditions in play on the road at the time of arrival, not what they should be according to the map.
"use a better image set for road signs"
The AI also needs the ability to parse and understand the printed instructions on the temporary yellow roadworks signs which, although designed to fit within the regs, are not standardised in the wording.
Also, part of the A1(M) northbound was closed yesterday evening and I had to follow the diversion. Luckily I know the area and followed my nose rather the diversion and re-joined further up away from the congestion on the latter part of the diversion route.
" Many horses apparently confuse plastic bags stuck in hedges to be a vortex to another universe, and are likely to misbehave."
EXACTLY! this fad of riding horses around everywhere will never catch on, we need to keep walking everywhere!
(every old technology was new and scary at some point)
will stop at KFC, that's not a bug it's a feature
Nope, its a bug. I'm quite partial to fast food, and to chicken. But if KFC were the last fast food outlet on earth, I'd still pass up their revolting greasy, horrible offerings. It doesn't help that they also seem to specialise in disinterested staff, and dingy poorly cleaned interiors with all the charm of a municipal toilet. Give me a McD's "chicken legend" anyday.
If my self driving car took me to KFC I'd be demanding my money back.
wouldn't want either a self driving car
What your or I want currently counts for nothing, looking at the obsessive enthusiasm of politicians and government administrators all round the world. They have decided that this new technology (which will obey every single one of their rules) is a marvellous thing, and they can't wait for this new way of controlling the populace. Not just in terms of the obvious like speed limits, but in ways of raising more tax (road pricing), in terms of limiting the use of private vehicles, or even discouraging private vehicle ownership (already an objective for the mayor of London), but also a way of massaging energy use and demand to fit the buggered up energy system the same people have created.
Probably wouldn't want either a self driving car or KFC.
I will take a car which is INTEGRATED with the highway and under highway control any day. Something that has proper algorithms which implement the solution to an optimal control problem. Something which is under the central control of the computer responsible for this particular highway stretch not everyone doing their own thing.
Neural network regurgitated belch? Screw that. I have done numerical methods work in my past lives and I know exactly easy it is to clusterf*** the underlying math. No amount of californicating dope can convince me to the opposite.
It doesn't help that they also seem to specialise in disinterested staff, and dingy poorly cleaned interiors with all the charm of a municipal toilet.
Are you sure those are not just the charm and colour of your locale or hometown?
I know the first KFC to open in my neck of the woods (circa 1989-1990) was closed shortly thereafter due to selling raw chicken - it was 10 years before another opened.
This post has been deleted by its author
"but i turn my nose up at the worlds leading chicken fast food specialist"
Yeah, because everyone knows that "biggest selling" is *always* synonymous with "the best".
Some guy called Colonel Sanders slammed the quality of KFC after it was taken over. What the hell did that ignorant asshole think he knew about cooking chicken that the corporate beancounters didn't?
I last had KFC around fifteen years ago, and it was best summed up as mediocre and overpriced. (Bland water/grease-filled chicken and "large" fries that were medium at best). Maybe they're fantastic these days, but the balance of what I've read online doesn't weigh in their favour.
You're entitled to your own opinion on their merits, but you can take the tedious corporate-fellating inverted-snobbery-invoking line of argument and shove it.
Ironically, I don't think Kravitz actually outsold Hendrix. But following your logic, I bet you consider Nickelback (#) far better than Howlin' Wolf?
(#) Cheap shot, possibly. Not the world's worst, so much as- like KFC- tediously mediocre for something so prominent.
KFC is grease with added chicken bone.
FTFY. The skin is intact, and crispy coated, but they seem to have some special process to remove the actual meat (presumably for use in pies) - so that you get a bite of crispy skin, then there's just bone.
The idea that something needs to be correctly identified down to the nth degree is a little pedantic. The main rules will be "if there is an object in the road, don't hit it".
Whether or not it can see an object in the distance will be down to it's sensor's range and the time it would take the car to stop in those conditions. It needs to work out how far it can see. If it knows that it can only see 30 metres in front (for certain) then it needs to go at a speed which will make it sure that not only can it stop if needed but that it can stop gently, that limits it's speed, down to perhaps 7 mph. It doesn't need to know the top speed limits written on a sign.
On top of what it can see ahead of it, it also needs to take into account the surprise object suddenly appearing from the left from a crowd of moving objects (called people). It doesn't need to identify a child from an adult, male from female, just that it is an object.
On many roads, those moving objects on the path are only 4 ft away from the car. If it takes into account that any one of these objects could suddenly come from the side, and fall in front of it, then just knowing that this is a possibility will keep the car going very slowly just in case.
If the AI car is made safe, it is also made useless, since it moves too slow.
If the car is made useful, it is now going too fast to be safe, risks have been accepted, and so little details like correct identification are now less important.
For me, the question is, do you want to accept a car driven by a human making a mistake or an AI car. The human can be blamed, pointed at and taken to court.
The AI is a faceless corporation backed by government and the legal system that allowed it to be on the road. When it makes a mistake, who are you going to blame, and more importantly how are you going to get "justice" for the dead child? Good luck with prosecuting government / corporation / legal system once they have accepted that the risks are ok.
Leave humans in charge of a car. Just use AI to help avoid human mistake and enforce "a safe speed for the conditions"
This post has been deleted by its author
Why isn't there more progress on automating railways? Perhaps Google and others are more interested in personal information, trip data than actual autonomous cars. Look at the crapware called Android TV, you can't tune the set without accepting Google's T&C. Or the fact that Uber is more interested in user data than making profit from actual trips.
"Why isn't there more progress on automating railways?"
That's already pretty much a done deal. It's the cost and will to upgrade the infrastructure and the inertia of the unions that is stopping it. I mean, FFS, most of the recent rail strikes have been about which human, train driver or conductor gets to open and close the fucking doors at the station!!
"[...] Other reports suggest software issues contributed to the chickens' scratching."
It is said that KFC cut their costs by switching to a general courier rather than staying with Bidvest who specialise in food distribution.
I wonder if the job will now be subcontracted back to Bidvest?
Saw that happen in IT once when a long-standing big POS contract was lost to a cheaper bidder. The new supplier then discovered they hadn't the skills or access to spares for the existing POS system - and had to subcontract it back to the original supplier. The latter then had a profitable situation of basically time and materials with no risk.
Yes, sounds like a stupid bean counter decision. I think people that buy into Franchises are being conned. They supply capital, premises, staff and basically are an employee in all but name. Responsible for everything yet only able to use the Franchise supplies. KFC isn't a fast food chain, it's a marketing company, so inevitable they will end up using a cheaper inappropriate supplier for the outsourced distribution.
Franchises are nasty and exploitive.
How many human drivers are distracted by misleading signs?
It seems to me there's a second research project there - to find that out and to ban signs that could cause accidents.
The other thing is that in this country we don't really take vandalism seriously enough. Defacing road signs can be dangerous. What's South Georgia for if not to establish a colony post-Brexit and fill it with our more sociopathic citizens? At least the Bullingdon Club would feel at home among the penguins.
I get the idea of testing slightly altered signs to see whether they might be misinterpreted by an autonomous vehicle, but calling a lenticular sign an "attack" is not much different from hanging an actual stop sign over another road sign and calling that an attack. That's not an attack, that's just showing a different sign and having it classified correctly.
It is not a problem, it is an opportunity. That's what I was taught by bullshitting marketeers when our systems didn't come up to scratch.
Just imagine. You are sitting at the end of your road trying to get into the continuous flow of traffic on the main road to get to work. No bugger is letting you in because a). you wouldn't let them in and b). you are driving a better car than them. So what do you do?. The answer is flip the new switch on your dashboard that lights up the clever little display fitted to the wing/fender of you car showing a carefully crafted "STOP" sign. AI in other cars screech to halt.
Bingo, you are in. You will also be glad that you fitted bullet proof glass if you are on the left side of the Atlantic.
"The answer is flip the new switch on your dashboard that lights up the clever little display fitted to the wing/fender of you car showing a carefully crafted "STOP" sign. AI in other cars screech to halt."
Funny! But the more mundane answer is that the waiting car will broadcast a request to merge and the approaching cars will all slow slightly to create a gap for your car to quickly merge into with the minimum of delay, probably much less than happens now when a human driver stops to allow a merge and causes a tailback with a standing wave that take many minutes to clear.
DHL probably drove up to the KFC shop, put a card through the door saying, "We tried to deliver your parcel but you were out. Please collect it from your nearest delivery depot" and drove off again.
(unlike other delivery companies, who don't even bother with the card. I'm looking at you, DPD)
As I was driving in and out of Poole yesterday I had a deja vu moment. Bridge support roadworks, just like they had a year ago. But this time with a twist. Alternating between inside line closed signs with use both lane signs. How could anyone think this was sane. How many police have driven by and not questioned the stupidity. As it happens the inside lane closed was correct.
How would a autonomous vehicle cope - switch in and out and in and out and in and out - this is what logically should be done if no vehicles overtaking
"As I was driving in and out of Poole yesterday I had a deja vu moment. Bridge support roadworks, just like they had a year ago. But this time with a twist. Alternating between inside line closed signs with use both lane signs. How could anyone think this was sane."
Depends on the modelled traffic flow predictions. The theory is that if everyone queues up in the one lane the traffic queue will extend back to the previous junction and cause even more issues there. Reminding everyone to use both lanes and merge at the last minute stops that happening. In practice, that last minute merge often slows the traffic even more than people merging ASAP and can end up causing both lanes to be congested back to the previous junction in rush hour-like conditions. I guess it's risk management based on expected levels of traffic.
"Any misclassification of traffic signs can potentially lead to a multitude of disastrous consequences, ranging from a life-threatening accident to even a large-scale interruption of transportation services relying on autonomous cars." (from the original article)
So 'life threatening accident' is the bottom of the scale and traffic jam is top of the scale for 'disasterous consequeces'. Seems a bit odd, espescially when you consider (on our congested little isle) that the former invariably leads to the latter.
Increasing the limit would decrease congestion and thus reduce chaos. Dropping the limit would cause congestion, thus increasing the chaos.
Most people ignore most speed limits most of the time. Seriously, try driving rigidly to the limit and watch everyone else come sailing past, or stack up behind nose to tail. I tried sticking to 75mph on the GPS, so about 80-85 on the speedo on the A1 the other night and I was still overtaken constantly mile after mile.
If AI sticks to the limits rigidly, they'll have to rise, or utter carnage will result. The obvious problem that causes is for the none automated or low skilled road users - cyclists, pedestrians, OAP drivers who aren't really safe at any speed etc. There are solutions, just not popular ones.....
"Increasing the limit would decrease congestion and thus reduce chaos. Dropping the limit would cause congestion, thus increasing the chaos."
There's no one size fits all solution and your generalisation doesn't work in all situations, as the westbound M62 "smart motorway" upgrades in the Huddersfield/Bradford areas shows. It used to be slow/standing traffic for miles because everyone rapidly joined the the back of the queues. Now, with the speeds managed down for miles before the "blackspots", the small amount of congestion caused by large numbers of vehicle joining the motorway, especially on uphill slip-roads, clears fairly quickly because the traffic is approaching at much lower volumes. Ditto for the M1 North at J34 for Sheffield.
I'd much rather be doing 5-10 miles at 50 than 3 miles at 70 and 7 miles at 0-20.
On the road to Beadnell in Northumberland (not a million miles away from the Shilbottle sign mentioned by an earlier commentard) is my favourite defaced sign. At a crossroads there are Stop signs upon which someone has added "Hammertime".
Vandalism...not big and not clever, but does make me giggle.
actually looks quite useful. There are plenty of places signs get confusing because it's not at all clear exactly which part of a junction a given sign applies to, or where different information would be useful to people coming from different directions. While it could often be solved by better placement of standard signs, being able to present a different sign to people approaching from different angles (slip roads onto motorways, for example) has a definite use case.
A car turning left at a mini-roundabout - on exit it switched lanes into the right hand (wrong side) lane of the two lane road. At the next mini-roundabout they were fortunate not to meet another car coming in the other direction.
It took a while to think of a plausible explanation for that potentially dangerous manoeuvre. They had seen the No Entry sign beyond the mini-roundabout - and that would eventually stop them going in that direction. The driver had assumed by having to go left at the mini-roundabout - that they were entering a circular one-way system round the church.
In fact the road is organised as a two-way system that detours round three sides of the church. The more direct road on the fourth side is accessible by two way traffic for a short distance in that direction - but there is a No Entry sign where it rejoins the detoured road. The shape is like a backwards "D".
... our compliment of small, lemon-soaked paper napkins for your comfort, refreshment, and hygiene during the flight, which will be of two hours duration. Meanwhile we thank you for your patience. The cabin crew will shortly be serving coffee and biscuits… again.
There's going to be a ready market for comedy bumper (or rear window) stickers that have been designed, or accidentally found, to cause AI confusion in a particular model of self-driving cars. They'll sell like hotcakes.
Any fragility whatsoever in any of the self-driving AI vision algorithms is going to be the source of much merriment and amusement. Not to mention huge traffic jams. Jobs, empires and billions will be lost.
At precisely 7:15pm local, the TV Local News will mouth, "And now for our daily look at the latest edition of: Self- Driving Cars Being Silly."