This AI is full of holes: Brit council fixes thousands of road cracks spotted by algorithm using sat snaps Blackpool is literally and figuratively using artificial intelligence to pave the way for a better future for Britons, or something like that. It's using machine-learning software to identify potholes on roads in desperate need of repair. Simply "all of it" isn't precise enough, so code is used to pick out asphalt not-spots. …
I got the impression, from the bits reported, that all the miners were going to be re-hired and they'd dig down then across to separate the UK from the bedrock then tender for a tug company to pull us across the Atlantic. Last report I saw was that the tender winner didn't actually have any tugs because Nicola Sturgeon said they'd cost £97m, they'd actually cost £200m and the operator didn't want them anyway. Bloody Scots sabotaging the plans again.
It's so bad near me that i'll have to buy a mountain bike in a few years, because my road bike is getting damaged so regularly from the crap road conditions.
PS: I assume the person who downvoted you either dislikes cyclists, or just plain doesn't like you talking down Surrey & Kent.
Strange how many countries have a policy of when a road wears out to simply strip it back to the foundations and relay it completely. It may be costly but you never see any patching so one wonders if it turns out cheaper in the long run on repairs to road and vehicles.
This post has been deleted by its author
> Blackpool Council did not divulge many technical details describing how the AI technology works exactly.
I ran across a hilarious site today which is the mother and father of all conspiracy sites (I am loving the Reptilian Invasion Timeline (part of the Negative Alien Agenda) )
And it has THIS helpful guide to AI technology. Read and learn!
"I fell down this rabbit hole 5 hours ago and i haven't hit the bottom yet. This stuff is amazing! And deranged. Mostly deranged, but still i can't put it down!"
I've been looking in on these type of sites for some years now. It's interesting how they all seem to feed off each other and self-perpetuate. Even more enlightening is how certain SF memes appear in films of TV shows and then pop up on these sites as new "facts" shortly afterwards. Similar happens with SF books ideas too, but they tend to go with the more generally popular TV and film stuff.
> they all seem to feed off each other ... new "facts" shortly afterwards
Interestingly, and even more worryingly, if you closely read the published research in any virtue-meme-related area, you will see precisely the same pattern. Eg, build a model better fitted to the meme, then declare it as "data", which secondary researchers then use as their own basis. Then off the lakhs of tertiary researchers go, ringing the corollaries on a fiction.
The result is a cloud of "research", all very solid and internally consistent, which can take days plowing your way up the reference tree before you get to the primary work, at which point it all comes apart in your hands. Problem then, is that everyone else sees only this massive cloud of agreement, and trying to communicate the fundamental problems meets massive resistance/disbelief.
And some areas do even cite "inevitable consequences" in their conclusions which are straight out of SF.
> these type of sites
I will say that THIS site stands head and shoulders above the rest, so could be worth you taking another look at. It's somewhere between awesome and hair-raising to see the sheer thoroughness and logicalness and internal consistency, with no ranting, just a simple setting out of "facts". The only problem is the "facts" are decoupled from reality.
It stands as an excellent example for others of how to set out bodies of knowledge. And it stands as an psychiatric example of how intelligence and major conscientious effort can be profoundly discontiguous with reality.
You're not kidding with "head and shoulders above the rest". I clicked to the homepage and, what I'm assuming is a biography of one of the lead contributors or founders, is absolutely phenomenal:
"Lisa Renee experienced a spontaneous Kundalini event several years ago that catalyzed a Starseed Awakening to perceive multidimensional realities and communication with the Evolutionary forces of Light. Lisa Renee has been personally contacted and then prepared, trained and downloaded by Interdimensional beings (Extra-terrestrial and Ultra-terrestrial, not from this Universe) known as the Krystal Star and Aurora Guardians. Guided by Guardians, Lisa was biologically upgraded to be downloaded to comprehend the Science of Ascension through the Law of One and its dynamics upon the layers of energy fields. This understanding of Consciousness Technologies was experienced by her own personal evolution and began her transition into an Multidimensional Guide and God-Sovereign-Free (GSF) Steward during this Planet’s Ascension Cycle. She is an Emissary for the Guardian Groups and a spokesperson for the shift of humanity to Ascension. Along with the Guardian Groups, her mission is to support humanity through its evolution with education, awareness and by discussing the impacts of the energy shifts upon the planet, human beings and human consciousness. She is an Ascension Guide, Planet Gridworker, Starseed Advocate, Humanitarian, Writer and Educator for Disclosure and World Humanism."
Just...wow.
"After all, there are 10 kinds of people..."
Binary is deprecated. Now it is the infinite number of points obtained by recursively interpolating between two end points. Even end points are now deprecated - in favour of opposing extremes eventually joining in a circle. Then again - tangential points can only be be expressed on the surface of a sphere.
I suspect even the spheres may need at least a "time" dimension.
Blimey, that's even better than the orgone blasters site (although much easier to read, lacking the multicoloured text and thousands of exclamation marks).
Makes David Icke look quite sane. I particularly liked the bit where they go on about remaining still like a hunted animal....while travelling to Egypt. Teleportation, clearly...
Pick a road, any road, send a squad out and fill in all the potholes you see with your v1 eyeballs
Then drive to another road and do the same again.
if you find a bigger hole that needs more works or bigger equipment / more planning (ie send the bigger team tomorrow) then with a variety of Satan’s tools available that allow you to report roadworks, accidents and roadkill, then it will not cost much to add ‘pothole’ and big pothole. to the list and simply work from that . Both highways people and the public could use the same system.
After all, you would assume that highways people spend their day on the various highways, so this shouldn’t. Need tech, it just needs people who go out and do their day job, so that in turn we can go out and do ours
On a plus side though, hopefully it will lead to the end of the badlyvdesigned web apps that take ages to register a single pothole and then just mark it as closed complete the following day when the pothole still remains where it was.
</rant>
Pick a road, any road, send a squad out and fill in all the potholes you see with your v1 eyeballs
That's fine unless they do what happened in my street. They parked their wagon over the biggest blooming pot hole of the lot, filled the others in and drove off.
I fail to see how. You can't remotely fill in a hole, so they have to be "on the road" to do their job.
I'm fairly confident that something like Google maps, Waze and a bunch of other satnav tools that are already in place are cheaper than using satellites and some specialist 3rd party to run that service.
There would presumably be problems looking through trees and other such things that block visibility of the road from space and I guess it doesn't work in the rain either, since the road will look smooth as the rain fills in the holes with water. The same goes for traffic jams, they can't see the road through the cars sitting on top of the holes either.
What happens if:
1. you have an apathetic public who just expect others to report pot holes and/or hyperactive idiots who insist on reporting every tiny imperfection in 'their road'.
You end up with your teams fixing low priority small holes to make nuisances happy, while bigger holes go unfilled because nobody reports it, but everyone complains about it until eventually you get a big legal claim against you.
2. Your teams wander the roads, but consistently miss one or two roads through biases or incompetence. Again stuff might be missed.
This technology solves an efficiency problem: by identifying and flagging the highest priority holes to fix as soon as they appear, rather than waiting for someone to get unhappy about it, or one of your teams to stumble across it.
The bin lorries drive down our road once a week.
- The driver stays in the cab. Maybe they could make a quick note of knackered roads? Assuming they're not required to be constantly monitoring the other operators.
- An operator might turn an ankle in a pothole. So they're incentivised to make their working environment safer ;-)
"The bin lorries drive down our road once a week."
I would guess that most councils will have outsourced refuse collection to avoid issues with staff costs. Because its outsourced, any change to the contracts will take 2-3 years to renegotiate and result in a significant cost increase for refuse collection.
To pay for the increases, no potholes will be filled for the 2-3 years.
At which point, it would have been cheaper to just reseal all the roads over X many years old...
How about installing a suspension activated switch in the wheel arch of these vehicles? When they drive over a pot hole of a certain depth the switch is activated, it logs the GPS co-ordinates and sends an automated report to base? It could even take a photo as the vehicle drives over the hole.
Of course, for this to work, the drivers will have to intentionally NOT avoid the pot holes, which may prove costlier for them and unpopular with any passengers.
"When they drive over a pot hole of a certain depth the switch is activated, it logs the GPS co-ordinates and sends an automated report to base?"
The system would keep flagging the pavement outside my house. The bin lorry often cuts the corner by riding up over the kerb. Mind you - eventually they do have to repair the damage caused to the pavement.
They had a trial a few years ago of putting cameras on a bin lorries.
Live in rural Hants and all the local roads are effectively what I would call "cart track". Going over 30mph is a risk to the vehicle.
I find it hard to believe that sat photos are a better solution than one of the below
- cameras on bin lorries with "AI" to spot holes
- Deal with Google, Waze or other to correlate user pot hole experience (is driving, phone shock detected, with position etc)
- making it easier for users to report issues
I'd think having existing road users as automated detectors would be best.
And what exactly are the traditional methods, waiting for somebody to phone about it ?
Honestly, I'm not sure AI is required in this scenario. Get satellite image, compare pixel colors, write the algorithm that decides which pixels are on a road and not the right color, flag it.
But they got it done with AI statistical analysis machine, good on them. I hope that that means a better driving surface for more people this year.
It's all about the marketing, and AI is the current buzzword that gets them kudos as a council (from other councils).
This isn't that hard, get an AA road atlas and a highlighter pen. Drive down each road filling in potholes and marking the road on the map as you go. When you colour in all of the roads, start again.
The savings are from the van that drives around and paints a square around potholes weeks or months before the actual fixing crew go outside. That's never been a necessary part, they only do it because councils grade which holes they will fill. Councils have never heard the phrase "a stitch, in time, saves 9" so they wait until things are really bad before addressing them.
Pay one man in a van to drive all the streets once a week with a sat nav or a mobile phone to assess and record all the hole positions (assessment is the manual un-costed part of the AI process). I guess with on-costs that's £30000. Perhaps they use five of them, or one senior executive to do this instead, say £250000 ... I'm still £750000 short of savings. I really can't see how they can save £1m pa in *finding* pot holes ...
Assuming it is done using similar methods across the UK, the current method uses students or other minimum wage contractors walking the streets and marking a paper reference and the street with spray paint.
Once the streets are "walked", the results are entered into a system (the one I'm familiar with was simple Excel...) that calculates the number of problems per street and then ranks streets based on those.
Costs are estimated per street and then, based on budgets you get X many of those streets done.
To ensure transparency, any work that is carried out is usually done via a tender process introducing delays and additional administration and checking once the work is done.
My guess is that paying a man in a van to drive around will be contracted out and then require additional supervision to check and manage. And no, I'm not joking unfortunately.
"require additional supervision to check and manage. And no, I'm not joking unfortunately."
Your mistake there is the word "require". It's not required, it's just what councils do. Accepting that what is currently done is what is necessary is the reason councils are as bad as they are. We don't need to separate the finding and filling either. Send two workers out in a truck with everything they need to fix potholes. Have them document what they fixed and which roads are done. That's all you need. No work list, no database, no walkers, no additional layers of managers.
AI is fixing the wrong part of the process.
Not tried it personally, but this app seems more appropriate to the task (assuming it works, which the comments seem to imply is a bit hit and miss) - crowdsource the data. The app monitors GPS position and mobile accelerometer readings, sends a report with any spikes. Council (assuming they receive the data) then aggregates and sends a team out when number of reports on a road within a certain time window exceeds a threshold
Some are hard to avoid, but yes, I would try and avoid gaping holes that would clearly wreck a wheel if driven over at any speed. Perhaps the use case on the app is to get the ones that haven't yet made it to be this bad, however, and are perhaps a bit cheaper to repair
So since the AI is 'just' doing the spotting, was it costing the council over £1 million to simply flag the potholes needing repair?
I'd be happy to spend some time driving around and noting holes for them. I'll even throw in a 10% discount.
"And what exactly are the traditional methods, waiting for somebody to phone about it ?"
My local council has a page on their website for reporting potholes. They have notes on how to report and you can even upload a photo.
It works: On a road I regularly use, a utility company had dug a hole, not made a good job of filling it and a pothole had developed as lorries drive over it. The council had repaired it within a few weeks of my report - and they addressed some other issues on the same road.
We should have a change in the law that requires utilities and others who dig up the roads to be made to fix their own errors or at least, pay for their errors to be fixed, instead of leaving it to local councils.
" Get satellite image, compare pixel colors, write the algorithm that decides which pixels are on a road and not the right color, flag it."
Many things have permanent different pixel colours: road kill; previous patches; oil spills.
Possibly they should include a specific coloured layer below the surface. When it appears - then it is a pothole developing.
It means that they are only fixing damage so large that it can be seen from space!
The "up to 30cm across and up to 10cm deep before any action is taken" metric used by my County Council would mean that few car wreckers would have any chance of (commercially) being seen from space.
Exactly. The article says that "all of it" isn't precise enough, but that's precisely what is needed. When a road starts coming apart, half-arsed patches that disintegrate within a few weeks or days aren't just annoying for the people who want to use the road, they're more expensive in the long term than resurfacing the whole thing properly. The problem is that the cost comes all at once instead of little bits at a time, so it doesn't fit in nicely with elections and yearly budgets.
Seriously, how much does it cost to employ a guy 40 hours a week to drive round in a van, taking in EVERY road over the course of... hell... a month, let alone a year... and press a GPS tracker button whenever he finds one.
If this is "money-saving", then it's still a ridiculous waste of money.
I'll do it for £50k + vehicle expenses + £30 for the tracker + £20 a month for always-on data connection.
I make that about a £350,000 saving... AT LEAST over the AI, let alone whatever bureaucracy ritual they were doing before.
There are at least 455.8km to maintain.
Generally you have to walk the roads to check them rather than getting to sit in a vehicle. If you notice large potholes it's generally because insufficient funding was available when the pothole was small unless it was caused by another issue (water leaks).
"There are at least 455.8km to maintain. Generally you have to walk the roads to check them"
Say to walk all of Blackpool's 455.8 km of roads you need to actually walk 500km (if you don't want to spend the £1 million on an AI solution to optimise the walking graph required to cover all the roads)
Human walking speed is about 5km/h, so the whole walking bit can be done in 100 hours. Double that to allow for noting / marking, the whole lot can be done in 25 working days. Even allowing for all sorts of administrative work, vacation, sick leave etc, the council could pay 1 person something like £25k a year to cover the whole road network under their responsibility every 2 months or so.
How on earth was it costing the council £1 million (a year?) to just IDENTIFY the potholes?
So are you suggesting that satellite-images (of active roads with vehicles) pumped into AI is somehow better than a guy poodling down every road at 10-20mph, actively looking, or even walking the streets, segwaying, skateboarding, whatever?
How the hell are you going to see any more from a satellite than a guy actually driving down the street?
And guess what - if a driver looking for them going down a road looking for potholes, and able to stop and check any time he likes (hi vis optional), doesn't see them... either he's not doing his job, or they don't affect the roadway, or the satellite ain't gonna see them either (e.g. under a parking space).
Your method involves finding the potholes that should have been repaired before they became issues and focuses on major issues that can only be patched.
The Councils method assesses the health of the road and where possible identifies major issues (cracking or subsidence) when they haven't affected the road surface.
I suspect the Councils approach to this is that by avoiding labour intensive checks, they can carry out more surveys on a regular basis to identify deterioration while it can be repaired and the life of the surface can be extended versus requiring a new surface - hence the quoted £1m saving over "more traditional methods"
Note: this is based on London council techniques from 15-20 years ago. Methods may have changed...
They're not employing any more or any less men on this as before, I fail to see how this method can be cheaper than the old, of having their employees keeping both eyes open. And that is not taking into account of the cost of buying, adapting and running the software and the infrastructure. It looks to me as if the costs are merely being moved from one cash bucket to another. The classic government system of "saving" money.
Anyone who has tried to get compensation for damage caused by potholes will be familiar with the excuse that "We are not liable because we inspect the road for damage on a regular basis". Although this sounds feeble, I understand that it's a legal escape route.
Sending people out to inspect roads is expensive, and it may be difficult to cover enough roads to avoid liability. But if you can say "our AI inspects every road in the area continuously", then you don't have to pay anybody.
The fact that so-called AI systems can never explain their conclusions may work against this, but most people don't know about that.
Around my way the regular inspection is every 6 months for each section of road. At which point a bloke drives around in a little Council van and diligently measures every 'large' pothole to see if on that day it meets their exacting standards for repair. If so, it will be added to the repair list. After a (long) while another man will come out and try to identify the hole and paint a yellow square around it.The next day it will be badly patched.
If you can find the claim form on the County Council website (it is very well hidden) you have to supply a photograph of the damaged vehicle tyre/wheel in the same frame as the hole with the yellow square clearly visible. An FOI request by the local paper showed that they rejected 90% of claims in 2018. Clearly their lawyers are more competent than their repair teams!
While the politicians have formed an orderly queue to get their names associated with this project, its interesting to note an absence of commercial interest here - if Google, IBM or Microsoft were involved, we'd never hear the end of it. That makes me think that some exceptionally talented person employed by the council likely working on their initiative found out that this might be possible and managed to get the council to roll out a full exercise for testing and productionizing.
Unsung hero, I salute you and feel your pain.
"“We hope that Project Amber will act as a showcase of what can be achieved by local authorities working in partnership with the supply chain to achieve lower whole life costs in highway maintenance through the use of surface treatments,”
WTF is wrong with "We hope Project Amber will demonstrate how repairs can be carried out more cheaply"? Anything has to be better than that rambling orgasm of manglement speak.
Even "Tech's cheaper, innit" leaves a less putrid taste in the mouth than that pretentious bollox.
Of course doing things the old fashioned way goes like this:
"Thank you for calling the Blackpool City Council help line. Press 1 to get lost in the menu maze. Press 2 to hear this in Welsh."
All those poor minions who were employed ignoring reports from the public are now at risk of having their jobs outsourced to Finland. But soon the union will require that all AI output by channelled through them, and things will return to normal.
Remember: each new good system gets co-opted next to the bad systems and becomes useless.
The manual method seems to involve looking at the roads far too infrequently, spray painting around the hole... then waiting for the spray paint to be washed/worn away. Rinse and repeat.
If they did more preventative maintenance, it'd save money in the long term.
It's ridiculous that where there's two holes side by side, one three months more advanced than the other, they only patch the first one. Then have to come back another time to do the second. I'm sure that comes from some overly bureaucratic tendering/subcontracting process... :-(
....from what I've been reading that my local council must paragons of virtue. We really don't have a pothole problem that I'm aware of. Certainly not inside the Borough boundaries and not on the more rural roads that I use. I do travel a lot, mainly on motorways and A roads, but I can't say I've ever come across a "wheel breaker" in 30 years of traveling the country. I must be incredibly lucky!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
