Cyclists are a lot smaller and a lot less predictable than cars - so obviously harder for the algorithms to hunt down and hit - but they are improving.
Autonomous cars may be further away than believed. Testing of three leading systems found they hit a third of cyclists, and failed to avoid any oncoming cars. The tests [PDF] performed by the American Automobile Association (AAA) looked at three vehicles: a 2021 Hyundai Santa Fe with Highway Driving Assist; a 2021 Subaru …
This post has been deleted by a moderator
Lord Sourpuss, may you grow up...to the full size of an adult.
Those crammed-interior, clown-car toys are a nuisance on the road.
While you saved on your Tata, the rest of us had to accelerate around your lawn mower and pay jacked insurance rates.
Just wait until the cheap deathtraps from China are imported. The cheap and child-sized will love them, inflicting their poor choices on the rest of us.
Randy Newman was right.
Yes… I can with great certainty predict they will cut the Roundabout against the flow of traffic and into my vehicles path outside the Hermes Parcel Depot in my town.
I can also predict about around 50% of them have no cycle lights at night in general too.
Your Vehicle Assist needs about another 25 years in development to bake this nonsense in.
"Yes… I can with great certainty predict they will cut the Roundabout against the flow of traffic and into my vehicles path outside the Hermes Parcel Depot in my town."
Jump the curb at traffic lights. Fly through red lights. Blast through the speed limit and rider ability to brake. Go the wrong way down the one way. Basically a law unto themselves.
>Basically a law unto themselves.
Just as, I expect, a similar proportion of car/van/truck drivers are. I have certainly encountered motor vehicles doing all of the examples you give for cyclists.
Every morning as I walk to work, I pass a local primary school on the opposite side of the road ... having to dodge the cars driving along the pavement (thankfully, towards me) so that they can park that little bit closer to the pedestrian crossing, presumably so little Timmy/Flossie doesn't have to walk more than 50 yards.
"Just as, I expect, a similar proportion of car/van/truck drivers are. I have certainly encountered motor vehicles doing all of the examples you give for cyclists."
I certainly hope you havnt seen cars/vans/trucks jumping the curb at traffic lights (Often to avoid waiting and join the road they were turning onto) but yes some motor vehicles may do some of the bad behaviours. The difference is huge however as the motor vehicle gets reported by its reg plate and someone gets fined or worse. A cyclist does it and nothing happens. Police will even sit and watch it happen, and if they are in a car what are they to do?
I dont agree with the behaviour from either but when the police occasionally put together an 'initiative' to actually do something about law breaking cyclists I wonder why its only the occasional show instead of a persistent effort.
> I certainly hope you havnt seen cars/vans/trucks jumping the curb at traffic lights
Was the 'at traffic lights' especially important? The big metal pole in the pavement make it less likely for 4+ wheels there, but as my walk to work anecdote shows, car drivers are more than happy to use the pavement as an extension of the roadway when it suits their convenience.
> The difference is huge however as the motor vehicle gets reported by its reg plate and someone gets fined or worse.
Evidence for this assertion? Sure, a reg plate makes it easier to trace an offender, but prosecution more than likely ain't gonna happen unless the police are there (or have a camera) to see it happen. I could film my walk to work (if I wasn't busy dodging the cars) and send it to the police, but would anything happen as a result?
"Was the 'at traffic lights' especially important?"
Sorry it was a specific manoeuvre regularly pulled but I didnt explain it. When coming to traffic lights at a turn and the lights are red the cycle jumps onto the curb. Assuming they want to go left they ride along and join the moving traffic, assuming they want to go right they use the red light to cross to the other side and then cross to go right. My bad for not explaining it.
"Evidence for this assertion? Sure, a reg plate makes it easier to trace an offender, but prosecution more than likely ain't gonna happen unless the police are there (or have a camera) to see it happen."
Already thats drivers but not cyclists. If you report it and others do too you will probably find cops camped out with a mobile camera to bring in the money. If you have dangerous driving on camera then yes there is a chance the police will investigate. And thats not to defend bad driving, the question is what happens if you change that to cyclist.
The walking stick is a great implement for dealing with cyclist who go through red lights...
I saw the archetypical "little old lady" on a signalled pedestrian crossing dealing with a lycra-clad psycholist by introducing her cane to the spokes of his front wheel as he sped through the pedestrian crossing against a red traffic light (apparently they only apply to powered vehicles or something). The bicycle stopped abruptly, but the cyclist continued his forward motion until he met the ground with a satisfying thud. The little old lady then retrieved her can from the remains of the (obviously expensive) bicycle, and continued on her way.
All the other pedestrians nearby gave her a suitably rousing round of applause!
"...the motor vehicle gets reported ..."
That doesn't happen where I live. The local cops' bar on "things we care about" is set pretty high. I watched a car drive the wrong way on a major, one-way street. The cops parked at the curb watched it, too, and all they did was blip their siren at them.
Depends what you read and believe from this mythical hype machine.
In certain situations (and I'm thinking motorways predominantly here) I am fairly confident that we can have higher level autonomy in a reasonable timescale, one of the biggest challenges will be working out how to safely return control to a disinterested meatsack (either in the event that that's needed en route, or more typically at slip roads). In neither case will "drop control at the human" be sane.
In certain other situations (emergency braking) they are already a great help (though much of that is, as is typical with safety improvements) being absorbed by drivers relying on it - risk homeostasis.
The "hype machine" isn't mythical.
The question is what do /real/ people (not tech experts) expect out of cars and car makers, and how they react.
I know a person who as a teenager believed the news reports that "self driving cars are just around the corner!". She poo-pooed her driver's ed, and is now a very inept 23 year old driver, because those self-driving cars didn't show up on time.
We have people believing their damn Teslas can drive themselves WITHOUT supervision -- take a look at your average Tesla driver on the highway -- assuming they haven't tinted the heck out of their windows, they are often seen reading stuff on their phone or tablet, not even pretending to look out the window or being ready to take over.
"The "hype machine" isn't mythical."
What you see as a hype machine clearly isn't what I see.
But then I am looking forward to assistive technologies giving independence to people who are currently denied it.
I am looking forward to vehicles on the roads being reasonably predictable - and safer than the current average motorist (not nearly as a high a bar as you make out).
This test is like criticising a GSCE student for not being able to do degree level maths.
I'm thinking motorways
Of course, if you're only thinking motorways, you would put the smarts into the road as well as the car - it would be doable for such a limited network and it would solve all sorts of otherwise difficult problems. You could even couple your vehicles up into virtual trains.
And for the remaining roads, a simple speed limiter is probably all you need to significantly improve safety.
I think the "autonomous driving" crowd need to decide whether they're attempting to eliminate the cost of taxi drivers or improve the efficiency and safety of road transport. I suspect there's far too much emphasis on the former when the latter is much more tractable.
Of course, if you're only thinking motorways, you would put the smarts into the road
I don't know whereabouts in the world you are but here in the UK they've been turning motorways into "Smart Motorways" for the last decade or more and have generally caused complete chaos in the process. If you over look the fact the whole smart motorway dream has been a fiasco the sheer disruption and cost of the work has been excessive.
How much more disruption and cost do you think re-engineering the motorway system to help autonomous vehicles would cause? 10 times as bad? 100 times? Besides what ever tech we embed in the road system will be out of date before they've even fitted a 25 mile section of the M1.
To work well in practice you'd need an international agreement on how the tech would work, most countries would need to sign off on the same spec so good luck to you there.
"I don't know whereabouts in the world you are but here in the UK they've been turning motorways into "Smart Motorways" for the last decade or more and have generally caused complete chaos in the process."
The UK's definition of 'Smart' in this context beggars belief. It's designed to increase capacity at the expense of safety; nothing more; and it's an outright insult to anything carrying the label 'Smart'.
" UK "smart motorway" is?"
A motorway where the hard shoulder is turned into a permanent running lane.
There are emergency refuges to pull over assuming that you can get to one if in difficulty
Otherwise you end up stopped in a live running lane hoping that someone on the camera's spots you, puts on the matrix signs to slow the traffic (backed by speed camera's) and close your lane - and then pray that the motorists spot them and/or you.
God help you if you break down in a spot where there is nowhere to get out of the way of your car - say a bridge or spots where there are walls right up to the edge of the motorway.
They also have variable speed limits to control traffic flow when things are really busy or the people in the control room get bored.
Also on any moderately busy* stretch (Human nature being what it is), all lanes leading up to an accident site will become impassible to the emergency vehicles which previously would have used the hard shoulder aka the emergency lane.
* the very reason for allowing the hard shoulder lane to be used for normal traffic. (being far far cheaper than building another lane is just a coincidence)
> A motorway where the hard shoulder is turned into a permanent running lane.
So, what's "smart" about that? *scratches head* Oh, I see, marketing rebranding of "cheap widening of the roadway". I'm sure the Titanic had a "smart lifeboats" system.
quote "And for the remaining roads, a simple speed limiter is probably all you need to significantly improve safety."
A man with a red flag walking in front?
Tried it. Wasn't acceptable.
And yet the official figures from the US show that for mile driven Teslas has less accidents than other Human controlled cars.....
An autonomous vehicle does not have to perfect it just has to be better than the average human driver.
"And yet the official figures from the US show that for mile driven Teslas has less accidents than other Human controlled cars....."
In the US, there are vast amounts of roads where pedestrians, cyclists and cars don't mix. I'm not aware of a specific comparison of the accident rate per mile driven for accidents solely involving pedestrians and cyclists and no other vehicles, which is what this study is simulating. (i.e. we're comparing apples and some form of root vegetable here. Oranges, being a fruit are too similar for this comparison...)
Wow - so you can drive a car all the way without stopping - that probably makes you the slightly tired, not quite concentrating motorist we're all concerned about.
I can drive all the way to cornwall and back from the midlands, and I do it in an EV that isn't a Tesla. It takes me fractionally longer than it used to in an ICE vehicle, but I use substantially less energy and arrive much more alert...
With the current "best" in charging (i.e. 800V systems, which CCS chargers everywhere deal with happily) the breaks are far more limited by people's own timing (planned breaks, loo stops etc)
To do your hypothetical journey in a current fast charging EV would (ABRP planned, Ioniq5 selected) add a 25 minute break at Lancaster (half way) in a 4 hour 20 minute journey. That's exactly where I would want to break for a few minutes anyway, stretch legs, visit the facilities, relax the eyes and brain.
You'd then park in a car park with a charger, so would set off later in the day with a full charge and do the same on the way back.
Of course in the mean time you have waited for precisely zero time refuelling for the rest of the year.
>>>I use substantially less energy<<< - Apart from not needing to stir a manual gearbox I can't see any way you'd use less energy while driving :)
>>>current best in charging<<<* I think your hypothetical journey will be a slam dunk overnight when there'll be no waiting for chargers, sunny weekend in mid July? not so sure.
*ISWYDT - very good
Well since you picked an ICE with a pretty decent range, and since seriously rapid charging is now being seen as the way to go... I thought I'd pick the current cream of the crop (and it isn't Tesla, not by a long way).
Oh, and the infrastructure is all being rated to support this, that why the CCS standard is so nice. The car tells the charger what voltage and current to deliver, up to a thousand volts and three hundred and fifty amps...
When you only need twenty minutes.. you'll find the chargers are not all that busy, but even on a nice weekend... what's the issue? You plan for one stop, but you pick where you stop so that you can move on if there is a wait, exactly the same way I used to when I drove an ICE vehicle.
I can't remember the last time I had to wait at a charger.
I use less energy because an EV is inherently significantly more efficient than an ICE.
So much more efficient that you could use the dino juice input material in a power station and I'd get more miles from the energy contained therein than you would in an ICE using the same source material. Of course not all the power we use - currently less than half of our power (and we're exporting 10% of our demand) - is dino juice, so that's an effective doubling down again (at least in consumption terms).
"With the current "best" in charging (i.e. 800V systems, which CCS chargers everywhere deal with happily) the breaks are far more limited by people's own timing (planned breaks, loo stops etc)"
I just watched a video on YouTube from somebody that bought an Ionic 5 and took it on their first long road. Charging maxed out at 228kW and there were plenty of electrons replaced before they returned from using the restroom, getting food, having a break. Since they would have been stopping a couple of times anyway, the effective added time required for charging was two hours on something around a 2,400 mile trip. They were also lucky to have booked hotels with charging available so for a couple of nights they weren't losing travel time.
"Technology Connections (can't remember the other channel he took along) - and he wasn't lucky, he specifically chose hotels which provided that service."
Yep, you are correct. I think it was luck that one had a DCFC and not just L2....... or was I not paying enough attention.
It's been a couple of years since I've done a road trip for pleasure, but I do keep track of time when I do. A minimum stop for me in an ICE is 20 minutes for petrol, restroom and misc. A stop with a meal is at least 45 minutes and a big reason I like to drive is so I can stop along the way. Any roadside attraction that wants to attract more passersby would do well to have EV charging. With fast enough charging, a long trip isn't necessarily going to take longer. I don't need to be able to drive the 500 miles my car will go on full tank since my bladder will give out long before that. When I get closer to getting an EV, I'll gauge what my endurance is between comfort breaks at that point and not value cars with more range than that when I make my choice. Sucks to get old.
Yes - the DC chargers were unexpected.
When looking at range and comparing it with comfort breaks, I'd take ~80% of the real world (as opposed to WLTP) range, since you'll want to rapid charge from 10-90 (ish) rather than 0 (nervewracking) to 100 (slow at the end).
"I can drive my ICE car from Birmingham to Glasgow and back on 1 tank of diesel. No recharge necessary."
"It's people like you what cause unrest" John Cleese
Wait until you are old enough when stopping to use the loo happens long before the fuel tank runs dry.
Official figures - in the US - so are they reducing shunts on freeways - lots of cars are trundling in the same direction at similar speeds - so clearly a good use case.
Driving around UK B roads, sharing the roads with cyclists, horse riders, tractors, sheep, cows, wild ponys, pheasants (even the odd peacock here abouts) - or around the back streets in our towns - a very long way to go.
Air France 447 was not a prime example. The (idiot) pilots put the plane outside the operating envelope and into a deep stall all by themselves. They were not 'relentlessly trained' either; Air France afterwards re-instated basic-flying and gliding lessons to their students after this crash.
"In certain other situations (emergency braking) they are already a great help"
There are plenty of videos showing Tesla vehicles not doing an braking as they pile into lorries, other crashed cars and emergency vehicles with their flashers on. Besides that, slamming on the brakes isn't always the best thing to do.
You know it's bleedin' obvious, I know it's bleedin' obvious and the Yank lurker knows it's bleedin' obvious. Now how do we persuade all those pouring money into it and governments offering their citizens as crash test dummies that it's bleedin' obvious. Somebody needs to say it.
Please go and say that on a Tesla forum. They will not like it one little bit.
The Tesla cult/disciples are all in on Full Self Driving and RoboTaxis.
I think that you are 100% correct. The legal things alone will take years to sort out. Tesla will not want to have the liability for each and every crash that one of their cars is involved in unless Elon buys major insurance companies in every country where this is allowed.
Exactly my thoughts.
I cannot open PDFs so I don't know if the AAA tests were rubbish or this article is, but most definitely these are not, as the article misreports, "autonomous vehicles".
As you state, they're level 2 systems that seem to perform as described in the manual (I have an L2 myself).
Fully autonomous driving *is* hard, which is why we don't have L4+ cars on the road yet, but this test / article says nothing about the capabilities of L4-L5 platforms.
If it were even possible to engage these L2 systems in a non L2 environment then surely the systems should be outlawed. If the system is only capable of operating with any level of safety within tightly controlled usage cases it needs to be possible to only engage them in those limited situation.
To your point on the test parameters, at least the head on test is suspicious, as it took place entirely withing the speed envelope that most of these systems EXPLICITLY WARN the driver about in the owners manual. Low speed obstacle detection has been an industry wide problem, and my car will happy slow down from 65 to a nearly complete stop, then happily go back to accelerating and creep right into the back of the car ahead of me like it suddenly threw on an invisibility cloak.
Part of this is actually a tricky sensor fusion issue, where the radars become dodgy at low speeds and then the car has to decide weather to trust the cameras or the radar inputs. So it actually comes as no surprise that the Tesla at least tried to brake(as the newer ones lean more on their cameras). Though to be fair, the F35 still has similar issues. The article also fails to mention if the driver assist was trying to kick out at the time(alerting the driver it is time for them to get their head outta their backside).
That said, all of the vendors need to raise their game on tests like this, and Tesla should probably install a separate computer to track emergency vehicles, which they seem drawn to like moths...
It's rather like the French approach to intersections.
It's clear that intersections are dangerous places, as there are a lot of accidents at them.
Therefore, one should approach any intersection at speed in order to minimise one's time in such a dangerous environment.
and Tesla should probably install a separate computer to track emergency vehicles, which they seem drawn to like moths...
I thought they had fitted a separate tracking system for them, I thought that explained there success rate in hitting the buggers.
"Part of this is actually a tricky sensor fusion issue, where the radars become dodgy at low speeds and then the car has to decide weather to trust the cameras or the radar inputs. "
Not many engineers are good at using and programming Kalman filters.
Kalman filters made a neat read. Thanks and praise to the Reg readers who aren't afraid to wade in with the devilish details.
I feel like I am now an inch closer to understanding some of the artifacts that got spit out in some of the higher level systems I worked with in days gone by.
So to make that work for sensor data that is not generated with the same fundamental correlations(like radio/radar data vs camera data) you'd what, need to interpret both into a common object model then apply the filter to both models outputs to resolve dithering between the two sets? That might explain why the cheap car "radar" is a problem as the sensors stop returning valid data below certain range/speeds so you are dithering against a null set effectively. Seems like that would force a hard bound on the results of the equations where the results rapidly diverge from reality.
Interesting, but if my boss comes to be and asks what I am looking at I will only say: "Bove my paygrade sir, am nought but a humble backup tape jockey. No concern of mine." lest they try to promote me to R&D.
"As you state, they're level 2 systems that seem to perform as described in the manual (I have an L2 myself). Fully autonomous driving *is* hard, which is why we don't have L4+ cars on the road yet, but this test / article says nothing about the capabilities of L4-L5 platforms."
*You* know this.
*I* know this.
Car manufacturers' marketing departments however, do not seem to know this.
> Car manufacturers' marketing departments however, do not seem to know this.
Indeed, and that's the issue here, because what people know about a given car is what marketing tells them.
Not everybody out there even knows there are "autonomy levels". You tell them "autonomous", and with some movies helping, they imagine their car has an invisible chauffeur (the dummy which inflates behind the wheel is probably an option).
"Car manufacturers' marketing departments however, do not seem to know this."
The really clueless are the punters buying these gadgets without having the slightest what the limitations are.
I'm good with basic cruise control. It keeps my leg from cramping up on a long journey and from the slow creep into felony speeding when there isn't traffic around to keep pace. Anything that applies the brakes or fiddles with the steering would freak me out.
If the technology worked then the companies creating the driver-assist mode would be able to create an operating system and web browser that we could all use without any risk at all of getting infections ... oh wait, are they writing the driver-assist code in Java?
"These are L2 vehicles at best, but appear to be L3/4/5 tests."
They're L2 vehicles, but the marketing blurb (AutoPilot, CoPilot, EyeSight and so on) most definitely implies they are L3 at least; and Tesla's Full Self Drive doesn't just imply it, it outright states it. Therefore it's fairly reasonable to put these claims to the test, even if we know the likely outcome.
Tesla's "full self drive" or whatever they call it these days has not been deployed in production yet, but that one *is* meant to be L3 or higher.
Which, in Europe, means you should be able to use it legally in a couple hundred km of motorway on northern Germany from later this year. Elsewhere you will presumably still be able to engage it but with you remaining in control.
Which suits me fine, really. Driver assistance is a wonderful thing from a safety standpoint, provided that you know what you're doing.
As someone else mentioned, marketing claims are a problem, as is lacking or inappropriate regulation. But that's a different side of the issue that should be addressed properly.
My background is both as an advanced driver (ex emergency services, and more) and as a commercial pilot, and I feel that driving licences should have different categories for different levels of automation (not just the current and becoming obsolete auto / manual transmission), as well as recurrent training and retesting.
Road accidents are a serious risk that is unfortunately way too normalised and without training, the more safety you add the more people compensate by taking bigger risks.
> Driver assistance is a wonderful thing from a safety standpoint
> the more safety you add the more people compensate by taking bigger risks
Fortunately you contradicted yourself so I don't have to do it myself... Assistance systems are a double-edged tool: While they do indeed save lives, they also encourage idiots to take more risks, because "the car will save them anyway".
Now I have among my extended family some (let's say) rather poor drivers (little to no situational awareness, stubbornness, tendency to follow whims). I can't really say driving assistance would help them much, if anything I know it would confuse them to no end when the car takes stupid decisions there is no obvious reason for (misinterpreting lanes, obstacles and speeds), and once they have decided their car is an adversary that needs to be taught who's boss, run for the shelters!
> That's why I included the third argument: better and recurrent training
Yes but considering previous mentioned real-world examples, I know it wouldn't help. It's really more like a condition, like being tone deaf: No matter how many music lessons, tone deaf is tone deaf. So, no amount of training will turn them into good drivers (after all they already have daily training driving to work/shopping/leisure). With age they eventually come to realize and admit their handicap, and actively try not to put themselves into situations too close to their limits, but that's about as far as it can go.
(Didn't downvote you BTW.)
There are a lot of them, hundreds or thousands mixed in with cycles and even more motorcycle scooters and motorcycles. It seems works OK with scooters in the middle of very busy roads.
Having a lack of angry UK car drivers screaming "You don't pay road tax" seems to help.
The other day I was in a convoy following a bicycle doing 15 mph in a 50 mph zone. After several minutes we all (20 or so cars) managed to overtake it safely according to the new rules, then we got to traffic lights on red and the cycle went past all the cars on the inside to the front. We all then had to follow the cycle yet again for another mile or so after the lights changed with no chance to overtake again safely.
Did it not occur to you that this simply demonstrates the futility of the overtake at any cost mindset? Or that maybe you, and all the other drivers annoyed but not in the slightest inconvenienced by the non-delay might have been better off riding bicycles themselves?
You seem to have completely reversed my rather obvious meaning. It's not the cyclist who feels strangely compelled to overtake at any cost. It's always a car driver.
I'm a cyclist. I am commonly courteous to other road users. Where road design makes overtaking difficult I pull over when it's safe and convenient to let faster traffic past.
I'm also a car driver. I understand that my powered vehicle will easily pass the bicycle as soon as conditions allow. Any time lost waiting for that moment will be easily made up afterwards.
I ride motorcycles too. I've been known to rollerblade. There are twats in all parts of society.
Perhaps simply because they are more numerous, car drivers seem to feature heavily.
"A small amount of common courtesy by cyclists to avoid tail backs is all that's needed."
Good luck with that. I'm not down on cyclists in particular. Those same sort of people can be complete bastards without ever being on a bicycle. The same thing applies to people in caravans that amass a long tail on winding roads and never pull over to clear the queue when they drive slow.
"Did it not occur to you that this simply demonstrates the futility of the overtake at any cost mindset? Or that maybe you, and all the other drivers annoyed but not in the slightest inconvenienced by the non-delay might have been better off riding bicycles themselves?"
If the cars had been allowed/able to safely overtake earlier, they would have got through the lights at least one cycle ahead of the cyclist, and would have been way ahead. Being stuck behind idiotic, selfish and inconsiderate cyclists for miles on end will result in non-insignificant delays; as well as increasing the risk of a serious accident (and quite probably shortening the lives of everybody involved as a result of the increased blood pressure).
And no, the answer is not 'relax'; the answer is 'be considerate to your fellow road users'. Something many cyclists know VERY little about.
Believe it or not, I'm writing this as an avid cyclist who is frequently disgusted by the shockingly antisocial behaviour of my fellow cyclists on city roads.
> the answer is 'be considerate to your fellow road users'
Come on, you know the right answer is "I'm the center of the world, always right, and I have to punish those who aren't like me". (Not speaking specifically about cyclists here...)
Trying to understand and respect other people has never been trendy, even back in the "Love & Peace" period. It has always been "Us vs.Them", and for those bicycle riders "Them" is everything having an engine (including motorbikes in my experience).
So they simply can't resist the heady feeling of elation and fulfillment blocking a dozen of those hated car drivers gives them. It literally makes their day; They will be lions in bed that evening...
Getting through traffic lights one cycle ahead will never put you "way ahead". It will, in my experience, simply get you slightly quicker to the next red light where i will re-pass. Perhaps you resent this. Nevertheless, it's still open to you to do likewise.
So your average speed, including stopping for traffic lights, was the same as the cyclist. No hold up at all then for, as sure as death and taxes, there will be another set of lights appearing shortly! This used to amaze me most mornings on my commute, drivers hurtling past just to brake sharply at the next set of traffic lights.
I worked a racking and stacking gig (de-comm / shift / re-comm) for the better part of a year back in the noughties. A large part of the job was driving a van out to the new data centre in Camberwell and it was exactly the same story but with higher speeds.
"So your average speed, including stopping for traffic lights, was the same as the cyclist. No hold up at all then..."
No; it was only the same because they'd been stuck behind the cyclist for miles. If they'd been able to overtake earlier, they'd have been through the lights way ahead and there would have been no issue.
This is the mess we’re in, we have compressed our lives’ timelines so much that we require cars to function properly, and cars that must get there as quickly as possible. So much talk about getting ahead and beating red lights. EV’s are not going to address this level of angst on the road, reducing car use is, moving more towards local active travel and services is going to make more peoples’ lives a lot less stressful! Anyone not seeing this is not looking outside the (tin) box.
> This is the mess we’re in, we have compressed our lives’ timelines so much that we require cars to function properly
Yes, but it has been like this for over a hundred years now considering that cars have been about for ages and you have the trains too that have been about way longer.
If everyone switched to bikes, well there wouldn't be enough room anyway, but lets say they all did. Well that would be like having to "go back to the good ol' days" a hundred years or so ago when everyone used horse and cart or simply walked.
I agree with your point but I think it's damn near impossible to roll back the clock so far and silly to think this is anything new.
Perhaps it is more useful to think about reducing the distance needed to travel to work more than extending the time needed to travel there.
Literally not true, though only a frequent occurrence on certain roads, and less often in city driving, as most people will take an alternate street when they are being trolled. And when the driver is on the clock, that time adds up. I let a box back delivery truck pass me on a narrow two lane road that follows an 11 mile winding grade. For long sections of it there is no safe place to overtake without risking a headon in the oncoming lane.
The spandex clad roadie in front of me decided his daily workout was more important to sharing the road, and eventually had six cars backup up behind him before an ambulance showed up. It was so far back that captain entitlement still didn't notice for another couple of turns. Apparently the guy does this every day, refuses to use a turnout either.
Different cultures than the mountain bikers I ride with, though they have their own antisocial tendencies I disagree with.
The roads are enough of a S@ show enough without the ego warfare, and it's a problem that needs to be addressed regardless of the rider/driver/pedestrians chosen form-of-transit based clade/gang/tribe.
Since shouting at the people on the the other side of the line isn't working, we need to address the behavior of those that see us as part of the same side. So I call out other cyclists when I need to, as it really is unjustifiable to lane hog when it is safe to tuck in, and a dick move to pile up cars just because you like standing in the clips. I also called out my friends for bombing single tracks and intentionally spooking horses back before they started closing most of the trails to bike traffic.
"Nobody has ever been stuck behind a cyclist for miles, unless they were driving along a cycle path. Total bollocks."
I live in a rural area, lots of twisty roads with short visibility zone before next bend (and all the fun of rural traffic - pedestrians, no pavements, horses, wildlife, farm animals crossing, tractors etc)
I drive an approx 1l engine car, so poor 0-60 acceleration, thus safe (for me & cyclist - I'm not risking an overtake when around next bend may be a tractor with a nice and spike laden attachment on the front!) overtaking opportunities on winding roads are very rare, so yes I can be stuck behind a cycle for miles
> So your average speed, including stopping for traffic lights, was the same as the cyclist. No hold up at all then
Given that logic you can replace "cyclist" with "pedestrian". Surprisingly, a car driving behind a pedestrian is just as fast (slow) as a pedestrian. Works for slugs too I guess.
Surprisingly, a car driving behind a pedestrian is just as fast (slow) as a pedestrian.
The difference is that pedestrians tend to keep their slow pace to the spaces created for them (footpaths) and leave the roads for the faster cars. Cyclists, on the other hand, frequently ignore the cycle tracks created for them on the basis that "I pay my taxes too, mate", and toddle along at 15MPH in 30/40/50 MPH zones. A car driver doing that would be booked by the boys in blue for obstruction.
> "I pay my taxes too, mate", and toddle along at 15MPH
I agree but, they dont pay road tax.
Also "toddle" along is subjective, depending on your point of reference. IN a car you are faced with a toddling speed, but yet if you are another cyclist or a pedestrian then the many cyclists doing 15MPH are a menace for the opposite reason.
Right, that would be the part where we point out that that's per cyclist, and the total over a ride can ad up.
Don't look at this from the POV of someone riding a kick scooter 5 blocks to hit starbucks. What's a few mins either way right.
Think of this as a handicapped person trying to get home from work in a wheel chair accessible van so they can use their own (accessible) bathroom without playing does-this-place-have-a-hadicapped-bathroom roulette. Now suddenly that time adds up fast, and the unnecessarily delays do more than sting.
Don't assume that taking unqualified potshots at car drivers isn't also punching down hill. You don't get a free pass on disrespecting other peoples time when you don't know their actual circumstances. Sure there is going to be the 1% of living streotypes like M3 bmw guy and Chevy Suburban Soccer mom. But unless you are in front of a BMW dealship or a school pickup line, most of the people you are talking about didn't earn the condescension.
If we are ever going to get a working transit system on either side of the pond, we will need the votes and voices of those who don't have the luxury of biking everywhere. We will have to build a balanced system that is optimized differently but still basically works for everyone. Wheaton's law applies. The trash talking is making that harder, not easier, and the loudest voices in the bike world are pushing for the least workable solutions.
I used a major A road where the traffic was travelling at 50 mph until the cycle held everyone up. Neither my wife or me would be willing or able to cycle the forty miles of the journey anyway. My wife needs a walking frame to move! The cyclist could have been more considerate of other road users by not undertaking the line of traffic at the lights and then holding everyone up a second time. Not wishing ill on the cyclist but I can guess what will happen to him if he does this on a daily basis.
I'll just add that I noticed one of the drivers ahead who overtook the cyclist for the second time wasn't very courteous and was considerably closer to him than one metre, taking a risk getting past against oncoming 50 mph traffic. Two wrongs don't make a right. There are assholes in cars and on bikes, when the two "meet" there are no winners only losers.
As many as one cyclist in 20 observing Highway Code H1 would be a miracle. “be considerate to other road users “
And I’m a cyclist. I try not to hold up the traffic, it’s bad for CO2 and it’s a really uncomfortable feeling. Find a safe place and stop for a few seconds. It makes little difference to my journey but it improves dozens of others.
Agreed, except for this bit:
> already sells a service with nobody behind the wheel.
Nobody *immediately* behind the wheel. But their cars are basically shadowed by a van with a field support guy and a spare driver, making them much more expensive to operate and not viable for mass deployment just yet (it is of course still amazing that they can do what they already do at all and it's a necessary step in the evolution of this field of technology).
Technically it makes them a partly autonomous vehicle with additional remote controls, and probably manual controls nobody is using at the time.
Wierdly, quite alot of modern cars are, though the latency of their celluar radios make this impractical outside of Defcon demonstrations. If it has electric power steering and wireless data for the entertainment system/onstar/vehicle locator at least.
Thankfully this almost never happens.
But their cars are basically shadowed by a van with a field support guy and a spare driver...
I'm not sure if that's actually true anymore. I've watched YouTube videos (at least last year, maybe in 2020) of people in Waymo cars who had to wait for ages for someone to come and help when the car got stuck. They have support drivers available, sure, but I'm not sure they're following all their cars around these days. First line of support seems to be to a call centre these days.
>>Imagine how well a left-driving remote gnome is going to do in right-driving locations.
Assuming you are referring to remote gnomes located in India (a sensible left driving location; only sensible countries drive on the left), I submit that the 'side of the road' issue would be the least of your worries, given the popular image of the driving environment on the sub-continent.
No hard data to confirm this, but I suspect that lane handedness has less of an impact on drivers from areas with free-for-all driving cultures. Learned behavior that focuses on hyperawareness of others and collision avoidance may offset some of the familiar habituation linked "sidedness".
I suspect environment dictates. Injecting that kind of chaos into the "alles ordentlich" culture of the German road system won't have the same impact as in France and I'm not sure the Italians would even notice. Germans in particular seem to drive very well, and expect others to as well, becoming quite flummoxed when confronted with 3rd world driving styles.
I'm not willing to kill people to get that data though, and don't envy the chap filling the human-subject research paperwork.
"The leading self-driving car system is Waymo"
They pull it off by having done very fine mapping of their operating locations and having that data onboard. The cars aren't independently autonomous and won't operate very reliably if they leave their home areas. What they really show is that driverless vehicles might be viable in well mapped city centers. I think that PRT, trams and other public transportation might be more efficient. I'd love to see PRT being used as first/last mile transportation in dense areas. It would be perfect at the Las Vegas convention center instead of a fleet of Tesla vehicles with drivers. For large companies such as Apple, Google, Microsoft, PRT cars could pickup and drop off directly inside buildings and ferry employees and visitors from remote parking and public transportation terminals.
Are there any tests of 'self driving' technology having to avoid pedestrians? (I have to declare an interest here, as I sold my old car for scrap last year.) If I recall correctly, one of the first fatalities was of a lady wheeling a bicycle across a road not recognised as such.
Additional source for the UK: The highway code.
If there is no pavement, keep to the right-hand side of the road so that you can see oncoming traffic.
Norwegian traffic laws also have a similar paragraph stating which side of the road pedestrians should walk when there is no footpath, as do, I presume, most other countries.
'See' is the pertinent part here. on the right hand side oncoming traffic will be the nearest when passing - you will see any without having to turn and look.
On approaching blind bends, it's far safer crossing to a position where 'See' gives both parties the earliest warning of the imminent meeting regardless of which direction the car is approaching.
There is also the principle of being able to stop your vehicle within the distance you can actually see, together these two points can keep everyone safe.
Yeah, over here it's explicit when their isn't a dedicated sidewalk/foot path, and on unpaved roads. We even learn there is a right "way" on the side walks even though you are allowed to walk either way.
Where it gets messed up is cyclists. Picture a gravel road descending a hill, with up hill bike and pedestrian traffic. Who goes where? On paved roads the bikes act like cars, stay off the sidewalks, and the pedestrians walk against the flow of the nearest traffic so everyone sees them.
Never seen a line in the DMV handbook covering it outright, and the "rules of road" in the real world seem to run the opposite, with bikes following hikers who dread the too late call of "on your left" before being crack attacked by a mountain bike tire from behind for the third time that day. Both seem to switch to walking like they were driving.
Hmm ... that wouldn't be as effective as you hope:
(1) it takes the pedestrian too long -- relative to the speed of the vehicle -- to process the info, conclude, "yes, I should shoot", draw, aim, and fire, and,
(2) inertia -- the driver may be disabled or dead, but that vehicle, like a Cape Buffalo, just keeps going, and hits the pedestrian anyway.
There was an article (here?) not long ago. Sort of a promo from one of the companies that make software. The code in Tesla and Toyota was not up to snuff. In all cases, the cars plowed through pedestrians in a crosswalk. The use case was the car had to avoid a stopped vehicle in front of them, swerving around. Not a good testament.
The decision training matrix breaks down when you start stacking hazards like that. If it made it around the car in the first place the driver should be taking over with the existing systems though so it's out of scope for the systems they were testing.
Also, to be fair human drivers also have a huge problem with stacked hazards and are much more likely to plow over pedestrians in a crosswalk when they just swerved to avoid hitting the car in front of them.
It also means that they (car bot or human meatsicle driving) were traveling to fast for conditions and didn't leave enough following/breaking distance.
And yes that means that the unreliable ML driven system is backed up by an unreliable human one. It's not a perfect world, that's why I advocate for the safety systems not the convenience systems, and as soon as they are killing less people than the people are killing people. Others seem to want to keep moving the goal posts till the car is expected to be a perfect driver, omniscient in all ways, and instantly solving intractable philosophic problems on the fly.
May they be haunted by the ghosts of those whose lives could have been saved, says I. Based on the numbers that may include mine.
We will have autonomous vehicles approved on expressways only for many years before they are allowed on city streets - i.e. we'll see autonomous long haul trucking years before autonomous taxis become common. No bicycles or pedestrians to hit, and you sure hope you never encounter an oncoming car!
As far as I'm concerned that's what I'd really want out of an autonomous car. I wouldn't mind at all having to drive it to my local interstate highway entrance then have it take over driving until I reach the exit nearest my destination. Even if I have to take over when switching from one interstate to another (it can be kind of complex as a human to interpret the signage and get into the appropriate lane) or in a construction zone having it take the wheel for 90% of the trip would be a big win. If I had the car drive itself I'd probably choose to take my car just about anywhere in the US (I'm fairly centrally located) instead of dealing with how terrible the flying experience is these days.
In France, which I freely admit might not be fully representative of the world at large, I once encountered a vehicle driving the wrong way down the hard shoulder. It was at night, so I first thought that the oncoming headlights were an optical illusion, but no, just a twat.
I've subsequently encountered people reversing around the Paris ring road and, memorably, a cyclist panting along a Spanish autopista.
Expressways are not risk free.
> I once encountered a vehicle driving the wrong way down the hard shoulder.
Back in my day that's how you were supposed to drive if you were an emergency vehicle (+ possibly highway maintenance?) attending a call. Though it was blues and blues on, which you would have probably noticed.
If they haven't moved to the Rettungsgasse system yet they probably will soon, though driver discipline leaves a bit to be desired compared to Germanic countries.
> a cyclist panting along a Spanish autopista.
On summer weekends that's the quickest way to make progress on them autopistas
"I've subsequently encountered people reversing around the Paris ring road and, memorably, a cyclist panting along a Spanish autopista."
Hats off to the reversing Paris ring road dude to just, you know, stay alive.
I've lived there for 7 years and no amount of money would ever make me do something that crazy !
Shouldn't it be the robot monitoring the human? DON'T rules are a lot easier to specify than fancy situation recognition. "If you speed once more Dave then you'll be reported to the police." "Why is your foot still n the throttle when there's red lights ahead?" "Get off your sodding phone." "You have noticed that Fire Engine with flashing blue lights haven't you?" But nag-bots aren't sexy. The UK government is in thrall to the hype.
My car automatically slows down when a slower vehicle is in front (and matches), or if it knows a junction/roundabout is approaching, or if it detects the speed limit has reduced. But I believe this is more to do with making the most of regenerative breaking and convenience rather than a safety/assist feature, and hence it never brakes enough to stop the car entirely (like the Subura/Hyundai)
Looking at your caption photo you've got to ask yourself how the cyclist got into this situation (OK, its obviously posed, but....). The only possible explanation is that the cyclist ran a stop sign and went into an intersection without looking. The cycle isn't really a cycle, either -- its obviously an electric bike or rather an electrically powered moped so you'd expect the rider to have at least undergone some training and be wearing minimal protective gear.
I drive a car but I also ride motorcycles and have a pedal cycle. All two wheeled vehicles deserve respect from their operator -- as soon as you go over about 12mph you've got enough kinetic energy to yourself or some other serious injury if you crash. So you have to learn how to ride responsibly. A lot of cyclists in our part of the world don't and given that their machines have relatively poor handling and flimsy brakes compared to a motorcycle and they carry little protective gear its no wonder that literally all of the serious accidents that have happened to people I know on two wheels have been on pedal cycles (mountain bikes are the worst -- even more things to hit when you come off). Pedal cyclists can help everyone, and themselves, by remembering that they're road users, too, and by at least carrying a mirror -- it might be OK for a slow moving vehicle to hold up traffic in hte UK but here if you've got more than five vehicles behind you as a slow mover you're obliged to pull over and let them pass.
...and expecting technology to save your butt if you don't ride defensively is wishful thinking.
> you've got to ask yourself how the cyclist got into this situation
That's a picture of one of the test runs. They're testing the car's response to intersection, not who's in the right, which direction has priority, which road user has what level of training, or how heavily armoured anyone is.
Of the three vehicles tested, one never noticed the dummy cyclist on an intersecting trajectory. It's not that it braked too late, or didn't swerve enough, or anything like that. It repeatedly failed to detect the cyclist, to throw up any detection alert, or brake. It struck the cyclist every time.
Conditions, speeds and distances were sufficient for other vehicles to respond. One was very consistent, one wasn't, but the Subaru Forester failed every time. You can't excuse that by asking questions about whether the test dummy was careful enough or having a go at cyclists generally, or declaring that you "have a pedal cycle".
I'm with you in terms of cyclists needing to take responsibility but in terms of testing it's not just what was called the golden path in another commentary that has to be tested, it's the ability to cope when things aren't has they should be. That should apply to all testing but especially where safety is concerned.
ex motor-cyclist here. I got into that situation because the motorist ran through a stop sign. Then gave a false address and drove off.
I understand that this particular posed picture looks like the car did not face a posed stop sign, but I was in a large intersection at a rail crossing, with the edge of the intersection just out of shot.
Despite being someone who works in marketing, marketing is a huge problem in this. Making people aware of your product and company, and persuading them to buy your product instead of the competitors is fine; but deliberately misleading people, and encouraging misunderstandings that benefit your company such as autopilot is autonomous, is really bad.
I'm also disappointed by encouraging fashions and fads to motivate people to buy more things or buy things they can't afford.
"The cyclist and vehicle targets were both lightweight and designed to be harmless to the test vehicle and driver."
Was the radar able to spot them? Were the cars guilty of crashing into cardboard boxes?
Having said that, we already have TMS - Too Much Software, and AI should only take over when proven to be safer than the average human.
Immediate solution - rename these shoddy AI systems. Why are Tesla allowed to call it AutoPilot? No wonder their owners think it will do everything. Trade descriptions.
Bicycles nowadays are mostly plastic, carbon fiber, or even wood/bamboo. You can't rely on them having any big amount of metal on them, not to mention pedestrians are usually non-metallic too, so your car should definitely not rely on metal detection to avoid diving into people or animals (either on foot or on bikes/scooters/skateboards/whatever).
Jasper Carrott as part of his stand-up comedy in the 80's used to read out what he claimed to be actual excuses given by motorists for accidents, recorded on insurance claim forms:
"The guy was all over the road; I had to swerve a number of times before I hit him!"
"The pedestrian had no idea which direction to run, so I ran over him!"
"I backed out of my driveway straight into a bus, it was five minutes early!"
"Backed by Mercedes insurance not your own" is an interesting one.
It means they either have a great deal of confidence in the current technology that they calculate it won't lead to significant claim-related losses, or they have so much confidence in the future of the technology that they're willing, short-term, to underwrite any losses.
With a simple lane assist with my Toyota Corolla, come the snow comes the warning that the camera is not active. It doesn't have to be heavy snow either. Also, with construction (essentially on every trip, local and otherwise, lanes are not often identified.
In fact as an added technology that I thought was promising, it has remained virtually unused. (Two or three times in five years)
My dad used to say that added electrics (by which he meant windows at that point) meant more to go wrong and expensive to repair.
The only really useful "tech" foe me has been seat warmers.
Lets face it there is no way that we can have autonomous vehicles sharing roads. A handful of kids with mobile phones could grind a whole city to a halt by strategically jumping in front of them unless steps are taken to prevent the sort of access to roads that cyclists and pedestrians currently enjoy.
The authorities are currently using slight of hand to do this under the radar by discouraging car use in our old town and city centres they are moving our new towns and cities to the sides of motorways (where pedestrians and cyclists are already banned). Once that is done the old towns and cities will be redeveloped and it is problem solved.
Microsoft has pledged to clamp down on access to AI tools designed to predict emotions, gender, and age from images, and will restrict the usage of its facial recognition and generative audio models in Azure.
The Windows giant made the promise on Tuesday while also sharing its so-called Responsible AI Standard, a document [PDF] in which the US corporation vowed to minimize any harm inflicted by its machine-learning software. This pledge included assurances that the biz will assess the impact of its technologies, document models' data and capabilities, and enforce stricter use guidelines.
This is needed because – and let's just check the notes here – there are apparently not enough laws yet regulating machine-learning technology use. Thus, in the absence of this legislation, Microsoft will just have to force itself to do the right thing.
Comment More than 250 mass shootings have occurred in the US so far this year, and AI advocates think they have the solution. Not gun control, but better tech, unsurprisingly.
Machine-learning biz Kogniz announced on Tuesday it was adding a ready-to-deploy gun detection model to its computer-vision platform. The system, we're told, can detect guns seen by security cameras and send notifications to those at risk, notifying police, locking down buildings, and performing other security tasks.
In addition to spotting firearms, Kogniz uses its other computer-vision modules to notice unusual behavior, such as children sprinting down hallways or someone climbing in through a window, which could indicate an active shooter.
In Brief No, AI chatbots are not sentient.
Just as soon as the story on a Google engineer, who blew the whistle on what he claimed was a sentient language model, went viral, multiple publications stepped in to say he's wrong.
The debate on whether the company's LaMDA chatbot is conscious or has a soul or not isn't a very good one, just because it's too easy to shut down the side that believes it does. Like most large language models, LaMDA has billions of parameters and was trained on text scraped from the internet. The model learns the relationships between words, and which ones are more likely to appear next to each other.
In brief US hardware startup Cerebras claims to have trained the largest AI model on a single device powered by the world's largest Wafer Scale Engine 2 chip the size of a plate.
"Using the Cerebras Software Platform (CSoft), our customers can easily train state-of-the-art GPT language models (such as GPT-3 and GPT-J) with up to 20 billion parameters on a single CS-2 system," the company claimed this week. "Running on a single CS-2, these models take minutes to set up and users can quickly move between models with just a few keystrokes."
The CS-2 packs a whopping 850,000 cores, and has 40GB of on-chip memory capable of reaching 20 PB/sec memory bandwidth. The specs on other types of AI accelerators and GPUs pale in comparison, meaning machine learning engineers have to train huge AI models with billions of parameters across more servers.
Opinion The Turing test is about us, not the bots, and it has failed.
Fans of the slow burn mainstream media U-turn had a treat last week.
On Saturday, the news broke that Blake Lemoine, a Google engineer charged with monitoring a chatbot called LaMDA for nastiness, had been put on paid leave for revealing confidential information.
In the latest episode of Black Mirror, a vast megacorp sells AI software that learns to mimic the voice of a deceased woman whose husband sits weeping over a smart speaker, listening to her dulcet tones.
Only joking – it's Amazon, and this is real life. The experimental feature of the company's virtual assistant, Alexa, was announced at an Amazon conference in Las Vegas on Wednesday.
Rohit Prasad, head scientist for Alexa AI, described the tech as a means to build trust between human and machine, enabling Alexa to "make the memories last" when "so many of us have lost someone we love" during the pandemic.
Google has placed one of its software engineers on paid administrative leave for violating the company's confidentiality policies.
Since 2021, Blake Lemoine, 41, had been tasked with talking to LaMDA, or Language Model for Dialogue Applications, as part of his job on Google's Responsible AI team, looking for whether the bot used discriminatory or hate speech.
LaMDA is "built by fine-tuning a family of Transformer-based neural language models specialized for dialog, with up to 137 billion model parameters, and teaching the models to leverage external knowledge sources," according to Google.
Analysis After re-establishing itself in the datacenter over the past few years, AMD is now hoping to become a big player in the AI compute space with an expanded portfolio of chips that cover everything from the edge to the cloud.
But as executives laid out during AMD's Financial Analyst Day 2022 event last week, the resurgent chip designer believes it has the right silicon and software coming into place to pursue the wider AI space.
As compelling as the leading large-scale language models may be, the fact remains that only the largest companies have the resources to actually deploy and train them at meaningful scale.
For enterprises eager to leverage AI to a competitive advantage, a cheaper, pared-down alternative may be a better fit, especially if it can be tuned to particular industries or domains.
That’s where an emerging set of AI startups hoping to carve out a niche: by building sparse, tailored models that, maybe not as powerful as GPT-3, are good enough for enterprise use cases and run on hardware that ditches expensive high-bandwidth memory (HBM) for commodity DDR.
GPUs are a powerful tool for machine-learning workloads, though they’re not necessarily the right tool for every AI job, according to Michael Bronstein, Twitter’s head of graph learning research.
His team recently showed Graphcore’s AI hardware offered an “order of magnitude speedup when comparing a single IPU processor to an Nvidia A100 GPU,” in temporal graph network (TGN) models.
“The choice of hardware for implementing Graph ML models is a crucial, yet often overlooked problem,” reads a joint article penned by Bronstein with Emanuele Rossi, an ML researcher at Twitter, and Daniel Justus, a researcher at Graphcore.
Biting the hand that feeds IT © 1998–2022