“The Silicon Valley race to build a self-driving car may revolve around one simple question: to map or not to map,” Alexei Oreskovic reports for Reuters.
“Google Inc, is creating intricate maps that detail every tree and curb along the road – an expensive endeavor that other companies could find difficult to match,” Oreskovic reports. “Newer entrants such as ride-share service Uber and Apple Inc could take a shortcut and develop a car capable of piloting itself without such elaborate and expensive blueprints, industry experts say.”
“Raj Rajkumar, one of the leading experts on self-driving cars at Carnegie Mellon University, said the map-based approach makes sense for a company with Google’s resources but is not required,” Oreskovic reports. “All autonomous cars rely on basic electronic maps for navigation and lane centering. But Google’s cars use far more detailed 3D maps, which the company creates by using laser scanners. Google analyzes the data, determining where traffic lights and stop signs are, for instance, so that the vehicle ‘knows’ exactly where it is. But the maps can quickly become stale, he said. Fresh snow could change the landscape.”
Read more in the full article here.
MacDailyNews Take: Google’s method would be fine in a warehouse for forklifts or something where changes to the landscape can be anticipated. The real world, not so much. Google’s mapping efforts seem like overkill. The focus, for real world autonomous driving, should be vehicles that can pilot themselves without painstakingly-made maps that can and will change often in myriad ways.
What’s wrong with driving yourself? Why the rush to autonomous vehicles? I can’t imagine a autonomous roadster on a curving mountain road, driving used to be a thrill.
Hi silverhawk1, there are three main benefits of autonomous driving that I can see here goes:
1) Taxis are suddenly so much cheaper – the main cost for taxis compared to driving your own car is paying for the people inside them to wait around and then drive you. Suddenly everyone can get a taxi everywhere for almost the same price as driving themselves. If taxis then start picking up multiple people who are going in the same direction it actually becomes CHEAPER than driving yourself (presuming you’re not going to know where to stop and pick up strangers) and so we move towards a much more efficient way for everyone to get around.
2) Better driving – the moment that a driverless car can pass a driving test and is allowed on the road is also the moment that the roads either become a lot faster or a lot safer. This is logical when you think about stopping distances. Most people take 1.5 seconds to hit a brake after being presented with information that should cause them to brake. Driverless cars may well take this amount of time as well to recognise some ambiguous danger, but you could imagine many scenarios when they could recognise a danger faster than this. And if they were presented with a car in front’s brake lights they would be able to brake almost immediately because a car braking in front of you necessarily means that you should also brake. After a couple of frames (a tiny fraction of a second) they would also be able to estimate the amount the car in front was braking and replicate it. This can allow cars to drive a lot faster safely, or drive at the same speeds and allowing the same stopping distances that human drivers are currently allowed by law with much much less chance of any form of accident.
3) Cars can be like limousines – currently cars are designed with a huge limitation. The driver must be able to see where they are going at all times – that means they need a front windscreen, they need a chair facing forward on a specific side of the car with pedals and a steering wheel, and they need windows that they can see out of to see the mirrors. The only counter example to this design is the limousine, where the section with a driver is separated from the section with the passengers. This gives us some idea of how the inside of a car could be improved if we didn’t need this seat that necessarily faces forward and gets in the way of more sociable seating and you can then think of what else you might like to view in terms of TV, the internet, video conferencing etc. instead of needing to watch what’s going on outside of your car.
Some very good points, Joseph. It would make sense that fewer people would even own cars, as one would always be available and affordable.
In addition to reacting more quickly, the cars would be talking to each other, so your car would know the car in front is braking before it even happened. It would also know if the other car is accelerating, or dodging a pothole, avoiding a hazard, etc.
Cars headed to the same area would form convoys.
Intersections would be handled smartly. That’s the most dangerous area of city driving now, but not with cars that are in communication.
Congestion would be handled so much better. One of the biggest problems with heavy traffic is people trying to get an edge by changing lanes. That would go away.
Merging when a lane is closed would be like a zipper. One car from each lane, alternating.
Etc. etc. etc.
I enjoy driving, but I can’t wait.
It will be so much safer that some point, having an accident while driving your own car will be considered to be negligence.
Imagine trucks driving themselves safely cross-country 24/7 at optimal fuel-efficient speeds, without taking any breaks (other than for fuel), while minimizing traffic congestion for commuters. The costs of trucking would decline substantially. And the per-mile cost of trucking would plummet, since the machine would not need to take coffee, biological, or sleep breaks and the truck would travel essentially non-stop, so would cover many more miles per day. With optimization programming, truck fuel efficiency could increase, with associated large reductions in CO2 and other by-products of burning fossil fuels. (Tractor-trailers now get only about 5 MPG.) Accidents — most of which are caused by driver error or fatigue, or failure to adapt to changing road or weather conditions) — would decline.
Imagine the same for driverless cars: occupant specifies destination and sits back while the car gets them there safely and while optimizing fuel efficiency. Picture sending your kid to school (or wherever), or picking them up, in their own driverless car. (Since car occupants won’t be driving, they won’t need a license.) Or ordering an on-line driverless car rental to pick up your mother-in-law in the city for a weekend visit with the family in the suburbs.
Driverless cars could have a huge favorable impact on us all.
Trucks driving themselves = expensive disaster. We have near-autonomous freight vehicles today. They are called trains. These trains are more efficient than any truck for moving freight long distances.
But thanks to the oil and automotive lobbies, we have allowed interstates to be clogged with large tractor trailers going cross-country at a much higher freight cost per mile than trains.
An upgrade in train infrastructure with overpasses instead of train crossings would allow freight to move at ~150 mph instead of 70 mph, and then smaller much more efficient trucks could deliver goods the last few miles. But that goes directly against the oil lobby’s goal of maximizing profit and one by one converting existing railroad right-of-ways into pedestrian trails.
Autonymous vehicles are an EXPENSIVE bag of hurt. You will probably see pilotless commercial airplanes before you see land-based autonymous vehicles/missiles. There are just too many obstacles to hit on the ground, the maintenance required for the arrays of sensors would be maddening. What do you do when your truck goes beserk in a lightning storm or starts having intermittent sensor failure due to corrosion or some latent undetectable failure?
Perfect in a laboratory, not so much in the real world. The human brain is wonderful at analysing danger before it actually happens and thinking ahead with anticipation at the forefront, computers far less so (and thats assuming all its sensors are working properly and its programming is near perfect).
When the London underground started it didn’t have windows, as you say why would you need them. However passengers found it extremely unpleasant in such an environment and though we are more sophisticated today a majority would still find it so, as they would travelling backwards, as many a bus and train survey has shown. Yes we may be able to adapt but that would be a generational thing and thus a very long drawn out a process causing yet more problems.
The unexpected is regularly catching out computers even when at the forefront of technological use. A Spanish HST drives around a corner at twice the speed it should with no computer intervention, Planes land short of the runway (or crash into the sea) because the computers mis understand what the sensors are telling them or the sensors are faulty or effected by conditions and when a warning is sounded even the pilots have no understanding of what is happening having been unprepared for it. These are simple scenarios as compared to cars on a road, so I think we are a very long time from autonomous cars common use, other than on specialist roads or tracks in controlled environments.
And then there is the argument of cost. The cost of updating maps, servicing and updating all the technical aspects of the vehicle and insurance premiums for when things go spectacularly wrong, would not instantly make them cheaper than a car and driver. I remember being told how robots would replace most industrial and indeed other jobs, but as we have found out after years of their use, outside of certain functions (especially where quality and consistency is demanded) they are still more expensive than ‘cheap’ human labour, otherwise the West wouldn’t be industrially a wasteland outside of the high end or specialist tasks or where image is everything. No I can’t see this ‘future’ happening any time soon.
As a small aside think of the problem of an automated taxi picking up others on the way and how would it ascertain that those new passengers aren’t exploiting the very private windowless environment to be a perfect place for crime? Yes yet more intricate technology required which no doubt Google will be happy to provide. Overkill… and literally if we aren’t careful.
Absolutely, some fantastic points. I’m not in the least saying that we’re ready yet (or indeed nearly there yet…) just trying to explain the sort of benefits that we’re aiming for. Apple will know when we’re ready, the rest of us less well researched people can only look at what’s already out there and hope for the next big thing.
I live in a suburban setting with a 2.4 million metro population. What is the break down in a USA population between urban, suburban and rural?
Sometimes the driving is the goal. Sometimes you just need to get to some other place and could just as well be reading or eating lunch during your transit time, sitting in a comfy chair, not having to manage the details of controlling a car.
Some GREAT replies to this post but the other benefit of driverless cars are for the poor and our seniors. Many seniors are trapped in their home. Out here in the center of the country we have far less access to public trans. Many older folks have impaired vision and need would have a much higher quality of life if they could get around more freely without risk to themselves and to others.
Many smaller communities would LOVE to offer “public transportation” to poorer residents but can’t really afford expensive bus systems. You can imagine a fleet of 5 to 10 self driving cards available for the disabled and poorer residents.
Access to Self Driving cars will radically improve the quality of life for many diverse groups.
People could buy things online or by phone and have them delivered via driverless vehicle. Think about all the things we now use cars to run errands for: food shopping, food take-out, drugstore pickups, whatever. Imagine making a reservation at your favorite restaurant, or buying tickets for a show, and automatically arranging for a driverless car pickup and return home. Could be revolutionary.
Driverless car fleets could also be a boon to all-electric vehicles, as fleet operators could easily overcome the (current) driving range limitations of all-electric vehicles since the “driving computer” could easily optimize trips, monitor driving range, and recharge (or swap out) batteries behind the scenes with no loss of service continuity to any passenger or beneficiary.
Quarter mile race would be awesome. Bet your favorite to win!
I’m not so sure about the MDN take on this one (unusually :-P)…
If you think about the way that humans drive we’re a lot more tentative in areas that we don’t recognise, and very confident driving in areas that we’re used to. I think of course that cars should BE ABLE to drive in areas that they’ve never seen before, but dismissing the benefit of being prepared for what’s usually next is throwing away a lot of incredibly useful information.
Add to that the benefits of crowd sourcing this data – so that your car will instantly know that the road ahead has a traffic problem or a change in the landscape because another car just posted an update about it… I mean Siri works because it has a connection to an awesome backend that can do the processing for it. I don’t see why a similar backend can’t keep the ‘expected road map’ up to date for any car and thus give the car the easier job of driving on a road it’s in some way ‘familiar’ with compared to one it knows nothing about…
Too many crazies in the world to trust “crowd sourced” information.
“Trust, but verify”
Crowd sourcing doesn’t mean everyone has an equal voice. Contribution amounts and trust gained by accurate reporting (verified by officials and other users), someone’s later contributions can be weighted greater than either a newbie trying to prank the system.
Relying on 100% official info is also crazy… crazy inefficient, that is. Thanks to TomTom’s incompetence, Apple’s Maps info near me are up to two years out of date and missing critical permanent road closures (following its routing there will literally put you through someone’s house now), and will also detour you 15 minutes by not using a “new” bridge. Google Maps were updated within days if not hours, thanks to user contributions through their mapping tools and Waze.
Crowd sourcing can also mean feedback without user intervention. Joseph Roffrey noted realtime traffic slowdowns, this is already gathered just by running the app, without user input at all, though with Waze you can report specifics about the situation.
Hey silverhawk1, I’m sorry I know ‘crowd sourced’ is a bit of a buzz word, but I don’t think I explained it in detail.
This sort of thing couldn’t be faked. It’s just a case of – your car drives down a road and it sees two things
1) the ‘mapping effort’ done by Google or Apple or whoever that is available online that shows what the world is expected to look like
2) the video from it’s onboard cameras showing what the world actually looks like.
The driving is easier because it can then plot it’s course somewhat in advance and only need to update it based on differences seen in the world as the camera feed comes in.
The ‘crowd sourced’ data is what then comes next – whenever it sees something it didn’t expect it uploads this information automatically and securely and it is added to the “mapping effort” ‘s online map – and because this map is hosted by Google or Apple it can put different weight on info submitted by different cars if some cars are worse at reporting dangers than others for example.
There are a number of advantages to this approach – consider an unusual case – a permanent danger is on a road in a hard to see location, that is almost always invisible but sometimes can be seen (e.g. behind a tree that keeps blowing in the way of the sign).
A live camera car would almost always never see it.
A ‘mapping effort’ might only notice this danger on the 3rd or 4th update.
A crowd sourcing map might be wrong to start with, but after a single car’s camera manages to see the danger it will then be updated into the map and not removed until someone sees that location (and not the tree in front of it) with said danger having been removed.
Another situation to be considered are changes that are only there at certain times of the day.
A live camera will have to be just as careful all of the time – imagine you had to presume that children were just as likely to be jumping out into the road at any moment and in any location, it would make you drive a lot slower right?
A static ‘mapping effort’ would help the car drive through the area but would have to have areas officially labelled as ‘be careful at these times’ and it would be very difficult to be more specific than that.
A dynamic ‘crowd sourced’ mapping effort would have records of when dangers were last seen and exactly where and how frequently such dangers are seen such that it wouldn’t be at all impossible to imagine a good algorithm being able to tell how likely it is that there are children about to jump out on each and every corner on each and every road, and so decide on how extra cautious it should be (e.g. how much slower than the speed limit it should drive and roughly how many obstacles it will expect to have to process in the next area)
My perfect solution would be a car that could be autonomous when driving on major routes between cities ( freeways, motorways and autobahns ), but then required the driver to take over once you leave those roads.
The task of keeping a car in the correct lane when all the traffic is going in the same direction and without pedestrians likely to be crossing the road is much more manageable. It would allow me to relax and pay less attention for the major part of my journey. When it comes to driving within urban areas, a human driver is likely to be the better option at the moment, although I’m sure that technology will be able to handle that in a few years.
Google’s approach of a hyper-detailed map is clearly never going to be a global solution, it’s just a fudge to make it work in a very limited area. A practical solution would require an autonomous car to be properly aware of it’s environment – even when it has changed from what it was expecting ( snow, accidents, new roadworks etc ).
A rock rolls off a cliff in front of you. If you swerve right to avoid it, you take out the cyclist beside you. Swerve left and you collide head-on with oncoming traffic. Brake hard, and the gravel truck following you too closely crushes you.
What to do? Driving means making decisions like this immediately and in the moment. I can’t see an algorithm handling this kind of complexity or any one of a million situations like this, because programmers simply can’t anticipate the randomness of reality. I don’t like how automation is degrading the skills of airline pilots (Air France, Airasia as recent examples) and I don’t like what it will do to driving.
Leave it for warehouses, not for open roads.
Hi Bubbles,
I think it’s fantastic that so many people are thinking like this, it means that software engineers are going to be properly held accountable for their work and that everyone is being properly tentative about allowing vehicles to make moral decisions.
If I might be allowed to defend the world of automated driving on this one just a bit though I’d really appreciate it!
In the example you’ve given, first of all the car would ignore the gravel truck, because currently I’m pretty sure that cars aren’t taking responsibility for what goes on behind them (and I think that legally they’re not required to either at least in the UK as long as they signal their intent correctly), but let’s take a look at a less ambiguous situation.
A cycle swerves in front of your car with an old man on it, and there’s incoming traffic to the left with hundreds of people and a woman with a baby to the right and no time to brake – what should you do?
Well a lot of philosophical talk is going into this at the moment, but the truth of the matter is, with current technology, what the car would do roughly every 60th of a second is as follows.
1) It would look at which route it can change the steering wheel to travel along and probably the first thing it would make that decision based on is the direction that would lead to the least likely to be dangerous situation (i.e. the one least likely to cause a single death in the near future) if all of the directions are definitely going to cause a death based on current data and all of those deaths would occur as quickly as possible, then it would probably carry on straight, but the complexity of the numbers is such that that’s very very unlikely and so it would select an optimal route.
2) It would look at whether on that route there is likely to be a collision or not, and if there is, how hard it needs to break to minimise the severity of that collision. If it’s definitely going to kill someone then it will do so at the lowest speed it can.
And yes it would do this every 60th of a second.
Now you’re right, it can’t do this completely accurately, because visual processing is not at a position to perfectly recognise objects on a road as ‘dangerous’ or ‘not dangerous’ yet. But this year cameras on cars will go live that are now able to accurately recognise any of – people, large animals (this one’s new), sides of the road containing oncoming traffic even without lane markings (quite new), the oncoming traffic itself, parked cars, cars in front of you, road signs, traffic lights, and other things like that, in a busy street, this can mean a LOT of objects being processed simultaneously, but they’re pretty good at doing this and updating each object accurately in 3D (if the car is fitted with pairs of cameras which some are otherwise in estimated-3D). No they’re not perfect yet, but situations like the one you’ve given are already being processed by cars on the road as we speak, and you can be sure that that’s not the most difficult driving situation. In fact, difficult driving is involved in things like overtaking, and understanding complex lane changes at difficult junctions, in the situation given, to be honest, even 5 year old systems would know that they just have to brake and hope that the gravel truck saw the rock before you did.
Google and Apple are completely different business models. Google gives away its free spyware in order to collect and sell data about you and the people you interact with. The self-driving car is just another one of their sideshows designed to distract you from their business model. Of course Google continues to drive past your house scanning every detail, but that’s for advertising disguised as being for cars. Name one piece of hardware that has made Google a significant profit.
before everyone goes overboard with “self-driving cars” an important legal question needs to be answered: who is legally responsible when that car runs over someone and kills them? The human driver? The car manufacturer? The software programmer? The map data people? Just who?
In our current legal system, I can’t imagine that question being settled anytime soon.
“Lane assist” systems can be helpful but the responsibility in firmly still in the driver’s hands. I suspect someone will win a landmark case when a “Lane Assist” equipped car kills someone and the driver blames the manufacturer. One thing that is being painfully learned is fully automated “autopilot” systems can be counter productive in large commercial airplanes. Pilots can be “lulled to sleep” by autopilots and can have trouble in an emergency. And Pilots are among the most highly trained. Quite different then your average driver.
Maybe in the next decade we might have “follow me on the expressways” systems but we are a LONG way from a self driving car. I won’t be in one under any circumstances before 2030 on an open road.
Not just trouble in an emergency. Asiana 214 was caused by a reliance on automated landing systems that were unavailable in San Francisco that month as they were upgraded.
The pilots *caused* the emergency by not properly aviating (at least one pilot should’ve been looking out the window during the approach instead of the instruments), which ironically puts them back at the same level in your analogy as a driver who causes an accident because they’re not paying attention to the right things.
The sad thing is, even if automated cars are statistically 1000x better than average drivers, after the very first no-win scenario accident causing deaths lawyers will be eating each other alive to sue the companies with deep pockets, and set the industry back 10 years. Never mind that a human driver likely would’ve made a worse decision and killed more people in the same scenario.
Apple vs Google? More like Apple and Google vs Mercedes? Mercedes (who invented the Gas auto) already have this working? And all without having to map the curbs and trees. Can even maneuver roundabouts and country roads.
Look Ma! No giant rotating camera stuck on top of a BUG car!!!
Mercedes has made the Google Car look like a high school science project. 😆
Oh and I want that voice for Siri! 😉
Fuck you! Don’t upload self-starting files. Have some respect for your fellows.
1) clean up your language — that would demonstrate that you have respect for us too.
2) fix your plugin player settings so that audio and video don’t start automatically. it’s a user setting that YOU control.
3) have a nice day.
As I’ve pointed out to this punter elsewhere:
Self-starting YouTube videos is ENTIRELY thanks to:
1) Google making this the default in everyone’s Google accounts.
2) Not bothering to TURN IT OFF inside of one’s Google account while on ANY YouTube video. There’s a slider right there on screen. The setting is maintained in your Google cookie.
IOW: grhaki is entirely capable of stopping the self-starting all by himself.
Or IOW: Continue posting videos to MDN. It’s not a poster’s fault if they self-start.
We already have a pretty good guess that extremely detailed map imaging is EXACTLY what Apple is doing with ‘The Mystery Van’.
As the article infers, covering the intimate details of a driving course is something of a waste of time because of the often constant changes. Parked cars change hourly? Are they all going to be stripped out of the images.
Then of course there’s the cost. We know damned well such a project would be performed ONCE then left to rot over time. What’s the good of that? Driving along a details route imaged 10 years ago? That’s no help.
…The Google Maps image of where I used to live is 8 years old! My former, lazy, cheapass landlord had a blue tarp over the roof to stop the leaks! Still there for posterity. How lovely. (o_O)
Can’t afford your own home Derek? Spending all your hard earned cash on shit gadgets and rent. Lol
If only trolling was meaningful beyond showing off one’s self-loathing. Keep that in mind, anonymous coward ‘Cripz’. That’s the only reason I get sad about hater drool. 🙄
I am really enjoying the fine contributions here about autonomous driving cars and the technical challenges required to reach that.
I worry though about what might happen when some nefarious organization manages to sleaze their way into this. Say some war mongering, torturing nation that stoops to cyber espionage and cyber sabotage needs to have their addictive war fix. They could hack into select mobile devices (in this case certain self-driving car) that would result in speeding, stalling, sudden turns of the vehicles of their perceived enemy. Taking advantage of the subsequent accidents, deaths and chaos they could then do what they perceive as their patriotic duty and proceed to invade the country. Obviously this would be considered a benefit for such morally bereft entities, but for citizens of the free and civilized world, well not so much so.
Hey Road Warrior,
Great question if a bit morbid 😉
Most of the ‘driverless’ models at the moment are unhackable – that is the car runs it’s own app that is completely standalone and not even connected to anything (they can’t even be upgraded unless you put a physical new computer in!). Nobody is really talking about remote controllable automated cars (I would’ve thought that this would be much more useful a feature for military than for lay-people) but even if someone were to do this, I very much doubt that they would build a driverless car that COULD be remote control driven into a wall or someone else. One of the great things about these things is that they’re built from the ground up to avoid obstacles. Even if someone was to add a ‘remote control’ feature, I’m pretty sure there’d be a failsafe in the car that set off a brake if it was going to hit a wall or do anything dangerous… That sort of thing would only need to be ‘turn off-able’ for ‘stunt cars’ or military vehicles or similar, and otherwise would just be built in as standard.
The only way to really achieve what you’re talking about would be to upload a virus or similar in an update to the car’s operating system. Which means basically that you’d need to target everyone or nobody and it would have to be done by the company that supplied the cars themselves. But I mean, there’s nothing to stop Mercedes or Audi or whoever building a load of cars even now that can’t be programmed to suddenly lock the doors and put the foot down until they hit something at the push of a button. So we’d only be a tiny bit more in danger, because of course Apple is from America and they fit your description for war mongering, torturing, cyber espionage and the like… 😉
Hey Joseph Roffey thanks for a very nice post.
Making the car standalone is definitely one approach, but even with this, radar, video cameras and laser scanner technology all have limitations and will have to be tested under many conditions for safety reasons but I have confidence that those who wish to be nefarious about hijacking a standalone car or using it as a weapon will have a new field upon which to inflict their insecurities upon others using their wannabe terroristic signature that the free and civilized world is now becoming acutely aware of. To wit, a remote self driving car that appears to be standalone would be a lovely tool that could be used to cause a dreadful accident or a whole series, you like they with sequels. This may stifle the entire industry why they are sent on a wild goose chase to find out what is wrong. I don’t underestimate the evil that resides in those that conduct themselves in such a manner.
Yes there may appear to be less likely a chance for hijacking should the vehicle rely on determining where it is, but as soon as there is GPS access or road traffic condition access there is that possibility. Not to mention a lock out key. Sudden engine shut off and car lock at 100 km/hr? Great idea for death. There is nothing from preventing unscrupulous organizations from infiltrating the chips of the vehicles sensor system, writing a virus for it, and having the virus engage when certain readings are scanned by the radar sensor. Tough to track down what exactly would be happening and of course those organizations that like to go in and do that would have no qualms about doing that. Removal of fail safes or circumventing is always a possibly, which can be circumvented by the way the signals were interpreted. In terms of your point of the logistics of targeting, a good point but again, having a virus located globally (all cars) and activating it locally (say exclusive to a country) is not impossible to do, actually that’s a bit of an modus operandi for those organizations I am referring to.
I am not referring to Apple in this scenario by the way. Insofar as I know they are not an organization interested in war mongering, torturing, cyber espionage, cyber sabotage and the like. I’d trust a self driving car from Apple to take me to where I’m going. I’d also trust a Google self driving car to take me to where I’m going with several detours to places it thinks I might buy stuff. I’d also trust an organization to cause a major panic by sabotaging car company X so that they could provide cars from company Y. I’d also trust them to have a leader to perfectly rationalize it with something like “Yeah so we cause the crash of a few volks, but we must look fordward.”
Great post thanks for providing such good and pertinent information.
This past week, Steve Gibson of the ‘Security Now!’ podcast had with him Lee Pike and Pat Hickey from Galois.com. They’re the two guys who hacked a car on the CBS program ’60 Minutes’ a few weeks back. They forced a car’s brakes to fail, making poor Leslie Stall bash through a cone barricade. They did it remotely.
https://www.grc.com/sn/sn-497.txt
An excerpt of the conversation:
STEVE [Gibson]: …all cars have this problem.
LEO [Laport]: Yeah.
STEVE: And they increasingly have this problem. There are somewhere – as many as 50 so-called ECUs, sort of autonomous subassembly computer units, in a high-end car. And even economy cars will have maybe half that, in the low 20s. And so the reason the make and model wasn’t shown was, first of all, they don’t want to really upset anyone. And it’s really not fair to discriminate because all cars today are like this.
LEO: Well, wait a minute. All cars are like this?
STEVE: Take it away, guys.
LEO: I mean, you can hack my car?
PAT [Hickey]: Probably not in specific. What we’ve been part of is security researchers have found vulnerabilities. We reproduced some of those to help the “60 Minutes” folks demonstrate that.
LEO: So each model would have a different exploit.
LEE [Pike]: That’s right.
PAT: Yeah, or makes might have some components that are common across several model years.
LEE: And many are going to be shared because it’s all coming from the same suppliers, the subcomponents.
IOW: We’re already fscked. Car manufacturing at the moment, just like ‘The Internet of Things’, pays almost NO attention to security beyond what’s legally enforced. I have no doubt there are exceptions. Certain car companies are going to jump on car hacking security problems and solve them. But it doesn’t take any new autonomous driving car to bring murder-by-car-hacking into our modern world. It’s here right now.
Here’s an article from June of 2013 questioning whether one journalist’s death was due to car hacking:
http://rt.com/usa/michael-hastings-cyber-car-218/
Error repair: “Car manufacturing at THIS moment…’