Tesla autopilot saves driver after he fell asleep at wheel on the freeway In an exciting first, the autopilot feature in a Tesla car managed to save rather than kill its occupant. At 0300 PT on Friday, the highway patrol pulled alongside at grey Tesla Model S travelling at 70 miles an hour on a freeway down toward Silicon Valley and noticed that driver – 45-year-old Alexander Samek – appeared to be …
This post has been deleted by its author
Tesla's autopilot can't even cope with stop signals, or junctions. Turning it on in any kind of urban environment is an invitation for something horrible to happen.
And this isn't just isolated to Tesla. Even the best autonomous vehicles in the world suck. There is too much hype and too little consideration of how they'd handle situations that a human could solve easily but are essentially intractible to a machine.
Wow - that's some hateful downvoting. Was referring to this story:
https://www.independent.co.uk/life-style/gadgets-and-tech/tesla-autopilot-drives-hospital-life-threatening-medical-emergency-a7179886.html
Didn't realise it was made up. it was being reported pretty widely.
"Pity they can't actually do that, nor ever will."
Of course they'll be able to drive the blind or blind drunk home. WayMo's test vehicles can probably do that today. The issue is whether they can do so sufficiently reliably in all conceivable driving conditions to avoid collateral damage. Or at least get the level of injury to innocent bystanders down to socially acceptable levels.
And there is the problem - the acceptable levels to the public will be zero casualties, even though that's impossible outside of a controlled environment. Every single incident (regardless of fault) will be interpreted as a failure of self driving vehicles, where what we should be setting the bar at is as good as a human driver.
The makers of autonomous vehicles have only themselves to blame for this - they have sold the idea to the public on the basis that they will eliminate traffic accidents. If they turn out to be no better than human drivers, then people will quite rightly ask what point there is in having them.
And there is the problem - the acceptable levels to the public will be zero casualties..... where what we should be setting the bar at is as good as a human driver.
I'm not entirely convinced that that is where the bar is set. I think it's more the expectation that self driving vehicles should be as good (or better) than human drivers perceive themselves to be.
If you've been driving (say) 20 years and never had an accident, that's as much luck as anything. But most humans will assess that and see it as evidence of their own driving acumen (ignoring the fact that even the best driver can still be rear-ended - or worse - by a total pillock). Or, of course, they have been in accidents but it was totally the other guys fault (even if it wasn't).
So it's a high bar to meet.
I think the best you'll get in terms of acceptance for now, is there'll be 3 classes of driver
- Those who don't want self-driving cars full stop
- Those who want other drivers to be in self-driving cars (because other drivers are idiots)
- Those who want self-driving cars now (because they hate driving, or want safety improvements etc)
I'd hazard a guess that the majority of drivers probably fall into that middle group. They view their own driving as better than average, so you're going to have a really hard time selling to them on safety improvements because they expect the car to be better than their image of themselves - which means near perfect.
I actually have similar concerns to that but it's not so much an assessment of my own driving abilities that drives it. It's partly more based on observations of the quality of the code we see released in other areas by some of the companies involved (e.g. Google and Android) and much more driven by concerns about skanky companies cutting corners to make money (see Uber).
It doesn't mean we won't get there in the end, but it's a rocky path, and I think it's a lot further off than most of the players/advocates would like you to believe.
"what we should be setting the bar at is as good as a human driver."
Dunno about that. I know a guy who has been classed as 'at fault' in a couple of crashes (one involved someone swerving into his lane whom he hit less than a second later, the other involved a moment of bad timing to check his mirror - had it been a moment earlier or later, but he picked the one wrong second to check the mirrors). Normally he's a great driver and every one I know feels safe with him.
I know another driver who has never had or caused an accident, and he is scary terrible. People won't ride with him or offer/ask to drive if they do. He tail gates. He refuses to indicate. If he's going slow and someone tries to pass he'll speed up to block them in. If someone tries to merge he'll speed up. He spends a lot of his time looking out the window. Just this last week he came within inches of hitting a parked car at speed but his passengers screamed and he looked back at the road only to find he wasn't on the road any more. Yet he's never once caused an accident (at least none within visual range at the time, who knows what has happened with frustrated drivers further down the road).
I'd settle for no 'distracted' type accidents and other accidents at least on a par with, if not better than an average driver. Mis-reading the road conditions is not a bad thing. Not reading them at all is quite bad. Not even lifting your head from your txt is terrible. If we can at least take that much out of the equation.. (BTW, manufacturers - make sure the entertainment systems and communication systems are completely divorced from the driving systems - you don't want a hack or a fault on the first two chewing up resources that causes the 3rd to fail!)
You can still get charged for that:
Every person . . . who is drunk while in charge on any highway or other public place of any carriage, horse, cattle, or steam engine, or who is drunk
so you can lose your license when drunk whilst being taken home by your cow
http://www.legislation.gov.uk/ukpga/Vict/35-36/94/section/12
My Grandad once lost his sergeant's stripes for being "drunk in charge of a bicycle".
One of my mates got fined for speeding on his bicycle. The cops were so impressed that he'd managed 36mph on a winding country lane that they didn't notice he'd inadvertently admitted he was cycling home from the pub.
One of my mates got fined for speeding on his bicycle. The cops were so impressed that he'd managed 36mph on a winding country lane that they didn't notice he'd inadvertently admitted he was cycling home from the pub.
One for the Widows and Orphans fund, or a shaggy dog story. Almost all UK speed restrictions apply to mechanically propelled vehicles only
Where I live if you are drunk in a vehicle - even in the back seat - you can be arrested if you have the keys on you. I was told by a friend who is a lawyer that if you want to sleep it off in your car, you need to get inside, lock the car, and toss the keys out of your reach before you go to sleep. And even that's not foolproof as if a cop sees you entering the vehicle drunk you can be cited for drunk driving since you can't prove that you were not intending to drive or would be leaving the keys out of your reach.
Obviously you could still be arrested for public intox if your car is not on private property, but that's a minor inconvenience compared to what happens to you for drunk driving.
As good as who? I don't know about you, but I'm a much better driver when I'm fully rested, paying 100% attention to the road, etc. than I am if I'm tired, distracted, have several people in my car yelling about something etc.
I've always said the bar needs to be set at a 90% reduction in crashes and fatalities per mile - measured across ALL conditions (i.e. not like Tesla's bullshit numbers where they claim Autopilot has a lower than human accident rate but compare their accident rate in conditions Autopilot can work in versus ALL the conditions humans drive in)
You might think that if humans have X fatalities for a certain amount of driving, but autonomous cars would had X-1 fatalities doing all the driving in those same conditions that we should want autonomous cars. WE WILL NOT! People need to see MAJOR benefits to give up control, so it needs to be better than the BEST human drivers (because let's face it, 95% of people think they are in the top 5% of drivers) paying full attention. A solution that is better when I'm tired and distracted but WORSE when I'm at full attention is not something I will accept!
One for the Widows and Orphans fund, or a shaggy dog story. Almost all UK speed restrictions apply to mechanically propelled vehicles only
Some issues there.
1) Cops seldom know the law, aside from "is it faster than the posted limit".
2) The law may not be as you think it is.
3) A bicycle is mechanically propelled. The pedals, crank, sproket, chain and whatever the rear hub uses is mechanical.
3a) I could make a rocket-powered vehicle that has no mechanical parts to it. I could theoretically even have no valves to control fuel input, and just burn fuel at a constant rate. I'd be able to break speed limits while not being able to be ticketed?
3b) Using such a law as an excuse would probably have been used by someone to claim their vehicle was 'coasting' at the time, thus not mechanically powered, thus not subject to such laws. Any such past law is probably since repealed.
4) Just because it's what the law says doesn't mean it's what the cops act on. And most people are not so well versed in the law that they'd have the knowledge to challenge the cops. Most people consider speeding as a strict liability offence; their vehicle was travelling faster than the limit therefore they're guilty.
One of my mates got fined for speeding on his bicycle. The cops were so impressed that he'd managed 36mph on a winding country lane that they didn't notice he'd inadvertently admitted he was cycling home from the pub
No he didn't. He might have been fined for cycling carelessly or without consideration for others or somesuch but he was not done for speeding. More likely he was simply stopped and given a talking to and inappropriate speed may well have been a factor in that.
https://www.slatergordon.co.uk/media-centre/blog/2015/06/can-cyclists-break-the-speed-limit-or-does-the-law-only-apply-to-motorists/
https://www.slatergordon.co.uk/media-centre/blog/2015/06/can-cyclists-break-the-speed-limit-or-does-the-law-only-apply-to-motorists/
From the same article :
"It is, however, possible for local bylaws to impose speed limits on cyclists. For example, on Hampstead Heath in London there is an 8mph speed limit for cyclists, and in Richmond Park the speed limit of 20 miles per hour for vehicles also applies to cyclists."
As I mentioned above, cops will give you a ticket even if the law says otherwise. Few people know the law well enough to challenge the ticket (in many countries speed camera vehicles must be legally parked or the ticket is invalid - how many people know this? The tickets stand unless challenged and either revoked by the appropriate officer or challenged and revoked in court).
Few countries have so few laws that any one person can know all of them, even in a field they specialise in :(
ISTR (from more years back than I care to remember) a newspaper report about a bloke who tried to get around the law by using a hovercraft to navigate along the river that was common to his house and the local pub... Dunno if he got away with it, but it makes for a good tale.
"how can anyone be certain that the drunkard in the car won't try to take control?"
That actually happened to a friend of mine - she was driving round a corner when the drunk passenger panicked and grabbed the wheel, resulting in the car rolling. Luckily nobody was seriously hurt other than the car.
The primary goal of autonomous vehicles is to increase productivity by reducing employment in the transportation industry. There is no money to be made in making fewer cars, the sharing concept was naive (to be kind). Which innocent to kill is similar nonsense; they will kill indiscriminately and at their discretion ( although I liked Alastair Dabb's kill the country musicians first). If driverless cars flow down to the individuals, it is more likely to be driven by commercial interests forcing non autonomous vehicles to be illegal; so that a container full of steaks isn't wasted when one of Musks monsters flattens a family.
I disagree. The reason I want one is for efficiency, yes, but my efficiency while riding in it. I can use the time spent in transport working on things, or reading, or something useful. Obviously, that's not safe to do yet, but there are real benefits to the users of the personal cars in addition to companies doing transport. Nobody thinks it is acceptable if a car will just crash into things, which is why I'm not expecting to get one of these for at least a decade, but there are many reasons to want one, even if you personally don't.
Create new company using robot trucks from the start, undercut prices, companies with drivers forced to downsize or close - nothing a union can really do about that, except try and complain to the government to get the automated vehicles banned.
For the foreseeable future self driving trucks will still have 'drivers', if only to prevent robbery, but the big point for transport companies is those trucks will be able to keep moving, at least, 22 hours a day
Good luck trying to leave the motorway when the slow lane has a twenty long convoy of trucks in close formation and you're in the middle lane.
In the autonomous world the trucks will be self driving and cooperatively communicating with surrounding traffic. they can then predict the arrival of the car at the exit and make space.
I remember this news about successful delivery of 50,000 beers by an autonomous truck
https://www.wired.com/2016/10/ubers-self-driving-truck-makes-first-delivery-50000-beers/
Been hot topic at the shop since then and occasionally. One old guy who brought his truck for headache rack and steel force tonneau installation on his F-150 was a delivery truck driver for 40years. He wish there was a self-driving truck back then. Security of the truck is one of their main concern when moving and stopping for rest.
Good luck trying to leave the motorway when the slow lane has a twenty long convoy of trucks in close formation and you're in the middle lane.
Similar ting happened on the German autobahn when they had a no-overtaking rule for trucks, the trucks all ended up in a big line behind a slow one and people died when they tried to cut in to exit and got it wrong.
So, if there is no driver giving signs of life, how the feck does the car manage to carry right on as if there is?
@Elonmusk: The default behaviour is not happening. That is a bug the size of an elephant. Look into it, or get cuffed when the court case that notices your failure to do so comes up.
"So, if there is no driver giving signs of life, how the feck does the car manage to carry right on as if there is?"
I was wondering that too. Is it possible to fall asleep while holding the steering wheel and not not let go? If this is one of those edge cases Tesla didn't consider, then they need to add some extra sensors to check. The "black box" should indicate if the steering wheel sensors where properly activated by the drivers hands.
Remember that movie where the kid "learns" to drive, "borrows" his grandad's car, a drunk man steals it... falls asleep with both hands inside the wheel... (so there is body heat and a heartbeat at or near the wheel), but the guy is drunk out cold...
Straight from 1987... (quite obviously named "License to Drive", because he never actually got the license.)
https://en.wikipedia.org/wiki/License_to_Drive
"if there is no driver giving signs of life, how the feck does the car manage to carry right on as if there is?"
Spot on. This is not the correct behaviour, the car should *ALREADY* have slowed down, pulled over and flashed hazard lights.
However with regard to the article highlighting the times that autopilot has crashed - it's 5 times with 2 fatalities. You can be 100% sure that if there were any others, they would be well-known and very public, because, Tesla and because, Autopilot. So how does 5 crashes and 2 fatalities in approx 3 years and god-knows-many-miles since deployment compare with human-driven cars.
Without that comparison the criticism is meaningless. Self-driving cars don't need to be perfect to be deployed, they just need to be better than the average human (which is actually quite a low bar), and will improve from there.
"Without that comparison the criticism is meaningless. "
Autopilot has 4x the accident rate of non-tesla drivers, and tesla drivers have 3x the accident rate of non-tesla drivers.
https://www.greencarreports.com/news/1107109_teslas-own-numbers-show-autopilot-has-higher-crash-rate-than-human-drivers
https://medium.com/@MidwesternHedgi/teslas-driver-fatality-rate-is-more-than-triple-that-of-luxury-cars-and-likely-even-higher-433670ddde17
Yes, they do.
I accept that a human driver is imperfect because I don't have the choice.
I do not accept that a software-driven car can make any mistakes. Every instance of doubt should immediately result in either a slowing down or an emergency stop of the automated vehicle to ensure life preservation. Resuming trajectory to happen only when all sensors give the all-clear. If that is impossible, request of a traffic drone to check and authorize continued movement, or call in a support team - or the police if necessary - to resolve the situation.
Being in a car is not a God-given right. It is a privilege that stops when human life is endangered. A human can disregard this, and end up in jail with one or more deaths on his conscience for the rest of his life.
There is exactly zero reason to give that privilege to a computer.
An immediate slowdown or emergency stop at any uncertainty is extremely reckless and will kill people. Not could, will. It sounds like a great idea for a car on an empty road or test track, but think about the consequences if you're being tailgated, or are in dense 70 mph freeway traffic, or have a passenger who isn't wearing a seatbelt.
Remember that the majority of collisions with Google's self-driving cars occurred when they followed the rules of the road as written, but the person behind them wanted to run a yellow/red light.
So how does 5 crashes and 2 fatalities in approx 3 years and god-knows-many-miles since deployment compare with human-driven cars.
Depending on which report you read the exact figure varies, but they all seem to be around 1 billion miles of cumulative Autopilot, and around 10 billion Tesla fleet miles in all modes.
https://electrek.co/2018/07/17/tesla-autopilot-miles-shadow-mode-report/
And according to OECD IRTAD data for 2017, road casualties in the US are 7.0 deaths per billion vehicle km.
Tesla Autopilot: 2 deaths per billion miles
US average: 11.2 deaths per billion miles
So at crude face value it looks like Tesla's doing well, except that's not adjusted for vehicle type and driver population, nor for the fact that Autopilot is something you'd engage on the open road. All of which I'd expect to bring the US average figure down dramatically, perhaps by a factor or 3-4, based largely on UK experience that shows that major non-urban roads are significantly safer per vehicle mile than urban and minor roads.
Overall, I don't think the evidence strongly points one way or the other when measuring fatalities per vehicle mile.
A sensible 70 MPH in a 65 MPH zone is a dead-giveaway too. CA driving culture is to travel 45 MPH in the fast lane until cars start honking, get angry, speed up to 90 MPH, resume using your cell phone, swerve around, realize that it's dangerous to drive 90 MPH while using a cellphone, slow down, then resume traveling 45 MPH while using a cell phone in the fast lane.
"Hmm, if Musk has them set to not speed, then it will actually create a hazard, as any Cali driver or cop will attest."
From a cops perspective, that sounds like a perfect catch-22. Not speeding? Get ticketed for causing and obstruction. Speeding? Get a speeding ticket. Either way, they get to pull pretty much anyone over at any time and maybe find "cause" to search the car etc too.
Probable cause..
Not being white, with intent
Yup, that there should cover it.
In a somewhat related story - did you hear about the nutter who took a Model T Ford across the US after 15 minutes driving one*. Spoiler alert he's still alive and he wrote a book.
*Model Ts have a 4ish speed kind-of-manual gearbox and three pedals. The slight problem being that the gearbox has two ratios the transaxle the same - and the middle pedal is actually reverse gear. The gearbox is controlled by the left side pedal (L-N-H) and you do have brakes but Microsoft might call them press & pray. Did I mention it will try its best to kill you if you try to use the starting handle?
> "From a cops perspective, that sounds like a perfect catch-22."
In theory, yes, but any ticket less than ten MPH over will generally be dropped by judges, and the cops know this. The de facto "speed limit" is now ambiguous and open to subjective interpretation, but the law does allow cops to write you up anyway if in their opinion you were driving at an "unsafe speed," regardless of your actual speed. Thus the stated limit is more of a guideline anyway.
The trick is to stay in the flow, which generally is well over the limit much of the time, except when traffic is heavy, which is most of the time.
At one point, cops wouldn't ticket for driving over the speed limit, but would definitely ticket for driving slower than everyone else. If the speed limit was 60 and everyone was going 80, you'd be ticketed for driving the speed limit. I don't know if the policy has changed as it's been decades since I set foot (or tire) into CA.
If you've ever had the pleasure of being pulled over by the CHP then you'll know that they make it pretty obvious that's what they want you to do -- the bright red light, the siren and so on. I'd guess that sometime soon there will be an update to Telsa software that identifies a cop car on your tail that wants you to stop and if you don't react within a short window it will bring the car to a halt.
As for it only being a few miles from where a Tesla came to grief on a gore point (that's what they call 'em around here) I'd guess that the firmware has been tweaked by now to try to unambiguously identify this situation. For those of us who are used to California freeways, especially in urban areas, its easy enough to get confused by the layout of junctions -- older ones can be incredibly confusing if you don't already know the road layout -- so I'd guess that Tesla has taken this problem in hand. (For readers who aren't used to freeways, they just drive on the M6 or something like that, then there's nothing close to the level of confusion and general intellectual mayhem that passes for US road signage in the UK -- motorway signs are clear and easy to read, the roads are comparatively well lit and people don't space junctions so you have to swerve over umpteen lanes of traffic for a ramp that's a mere half mile or less from the previous one.)
"there's nothing close to the level of confusion and general intellectual mayhem that passes for US road signage in the UK"
I've heard similar from friends and family who have driven in the US. A sat-nav with lane guidance is essential if you don't know the roads. They tell me that it's not uncommon for the *only* signs relating to a junction to be actually at the junction itself, no advance warning signs at all.
You haven't driven in Birmingham (UK) then. Sometimes continuing on your route will mean going straight ahead. Other times you'll need to be in the right lane as your route goes into a sudden right turn/slip road but most of the road you're on carries on forward ( to somewhere). Other times it's to the left. And then there's the roundabouts with 5 or more exits, several of which actually go in a different direction to how they first appear, but all of which require several lane changes.
And the A38 "Aston Distressway" where the same two lanes come on and go off again, with a gap in the barrier for both incoming and exiting traffic to share....as well as no central barrier and a centre lane of red tarmac which is into town in the morning, out of town in the evening and shut at other times, controlled by overhead gantry signs. Sort that lot out Tesla......
> (For readers who aren't used to freeways, they just drive on the M6 or something like that, then there's nothing close to the level of confusion and general intellectual mayhem that passes for US road signage in the UK -- motorway signs are clear and easy to read, the roads are comparatively well lit and people don't space junctions so you have to swerve over umpteen lanes of traffic for a ramp that's a mere half mile or less from the previous one.)
<
Eh, no. Its just different. Different history. Its a bit like saying that streets in London are a brain dead stupid bloody mess because they are are not built on a gridiron pattern like in most of LA. Its not like the first streets in London were not on a gridiron. Which the locals then promptly utterly screwed up over the next 1800 years.
The highway system in California has a very well worked out logic of its own, given the constraints the various parts were built under. At various times. The only problems I've had in the past were non matching on ramps / off ramps. And thats about it. The fact that the transition from the Hollywood Freeway to the Ventura Freeway North involves crossing almost a dozen lanes in just over a mile is just one of those physical facts of life of road engineering. All well signposted. If you know were and when to look. But there again I learned my big city driving skills in LA 30 plus years ago in a car that made a Lada look like a hotrod. A good way to learn to drive like a local.
The UK motorways are OK signage wise but the French do it better. Best traffic engineering I've seen anywhere.
How can it ‘normally’ save lives when this is the first publicized case where they actually did, in contrast to the well-known cases where they got drivers killed instead? For that matter, if the CHP hadn't noticed the guy and managed to force the car to stop, there's a good chance that it would've run into the deadly barrier a few minutes later, so we really can't say that the car saved him.
Is the constant snark helpful in these kinds of article? Seems to me the autopilot is quite impressive in a lot of situations. Have a look a some of the dashcam YouTube footage of it avoiding various collisions. Not perfect, but nothing is. What is the accident rate per mile driven compared to other cars?
I guess sarcasm is easier than finding stuff out.
Seems to me the autopilot is quite impressive in a lot of situations.
It gets a lot of snark because it's an automated lane control supercruise that's calling itself an autopilot, despite the demonstrated incapability of detect a stationary object in the road, like a large red vehicle with blue flashing lights and a siren.
In other words, it's a bit of software that is truthfully is not quite ready for an alpha test that is widely deployed to people who don't understand the shortcomings and endanger the lives of other people. If it only killed it's own drivers then we probably wouldn't mind so much since the choice would be on the person driving it, however many of us object to being killed by somebody elses bad software decisions, with a liability on the software company of saying "sorry" in the most insincere way possible.
What is the accident rate per mile driven compared to other cars?
Quite poor apparently.
https://www.greencarreports.com/news/1107109_teslas-own-numbers-show-autopilot-has-higher-crash-rate-than-human-drivers/page-2
It appears to be be possible (and has for some time) to circumvent the driver awareness system by wedging a water bottle or something of similar weight into the steering wheel, which fools it into thinking there's a hand on it. I should think that's what happened here.
The correct approach seems to be that taken by GM, which uses a camera system to track the driver's gaze and figure out if they're paying attention to the road or not. I'm also pleased to see that Volvos are immune to this little trick and use a different technique to detect a hand on the wheel.
Hyundai also have a system that requires driver input when it's Lane Assist is in Active mode. In my experience simply holding the wheel isn't enough, it very quickly spots if you are trying to fool it.
I'm surprised Tesla didn't include some way to detect a cop car, especially as this:
https://en.wikipedia.org/wiki/Traffic_signal_preemption
has been available in parts of the US for some time.
News about Tesla is a Rorschach test for the media.
If you read all the different media reports about this incident, its very easy to see that each reporter has seen what they wanted to see in the story.
Some are praising Tesla's Autopilot for saving lives. Others are critising it for allowing the situation to occur.
The reality is that it's not much of a story really, and it isn't really about Tesla. It's a story about a stupid guy who thought he could get away with drink driving, and about some cops who saved lives with smart thinking. An interesting story, to be sure, but not worth the vast number of column inches it's earned because journalists want to push their Tesla agenda.
Tesla's autopilot is supposed to disengage and slow the vehicle if the driver is not demonstrating attentiveness. So no, it didn't save this guy's life. It endangered lives by continuing to function even when it shouldn't have.
Tesla's system is simply broken. Drivers can fool it by placing weight on the wheel and can take their hands off for up to 30 seconds which is WAY too long, encouraging all kinds of unsafe activity and inattentiveness. If it were fit for purpose the car would require drivers wiggle the wheel, or perform some mandated task every so often that can't be fooled and they would bleep if hands were off the wheel for even a few seconds.
They're still treating driver engagement and attentiveness as an afterthought. This story merely demonstrates that. The next time it happens, the outcome might be vastly different.
People are ripping Musky to bits but the reports quotes the Muskid:
"Exactly. Default Autopilot behavior, if there’s no driver input, is to slow gradually to a stop & turn on hazard lights. Tesla service then contacts the owner. Looking into what happened here."
Surely the Musky-one's first two sentences explain what should have happened followed by a statement that they're looking into whether it acted correctly or not...
Luckily the reporter's sentence then completely confuses the quote by adding the non-statement "Which is, of course, not what happened.' Did it not take the correct action? Is Musky not looking it to it? Did it take the correct action but wrongly performed it?
Clarification would be good ...
I don't have a Tesla, and probably wouldn't use Autopilot if I did, but if I'm driving on the 2nd lane of a two-lane Autobahn, even if I pass out I would not want it to slow to a controlled halt and put the hazards on, because pretty soon I'd be rear-ended by a large saloon travelling at near warp speed, and that would not be good for anyone's health.
Make the police sirens emit a fax tone, and put microphones in the car with a modem plugged to it.
A one-way digital signal is embedded in the 9600-baud carrier tone, ordering the autonomous car to slow down and pull over, like a kill switch. Every autonomous car within the siren's range will hear the signal, and offer the driver a chance to IGNORE the command in 60 seconds. If he DOESN'T ignore, the car autonomously pulls over, or just stops.
Or make it a wifi or 3G signal that only autonomous vehicles can pick up (all Teslas have a 3G chip, don't they?), and ONLY the police has access and authorization to use it, preferably adding the license plate or VIN to make it an ADDRESSABLE signal. (Tampering with those would be illegal as consequence...)
Still, you can keep the 60 seconds command to IGNORE it, and answer to the POLICE why you would refuse to pull over.
Sorted.
HERE'S YOUR IT ANGLE.
Next!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
