Should We Be That Worried About The Potential Hacking Of Self-Driving Cars?

Holly Whitman

Holly Whitman is a writer and journalist based in Washington DC. She loves to share her thoughts on the intersection of politics and culture, and writes on everything from feminism and human rights to climate change and technology.

Related Post Roulette

134 Responses

  1. Kim says:

    Oh, noez! They’re going to need to be 100% accurate!

    no, they aren’t. Just better than humans. (The ability to detect spoofing is probably going to be required, sadly).

    Oh, noez! Hacking might KILL someone!!!! That’s a new issue!

    Excuse me? You’re aware we have medical devices that are computerized, right? We keep electronic medical records (some of which, like allergies, could kill a patient if erased).

    Hackers being able to kill someone is nothing new. Hackers being able to kill someone with transportation is nothing new (try bollixing up air traffic control).

    … and then, at the end of the article, “everything will be fine by the time this hits market.”Report

    • dragonfrog in reply to Kim says:

      Hacking with the ability to physically, directly, kill someone is a new-ish issue.

      Yes, to your point, I have my limited-edition Barnaby Jack tinfoil hat around here somewhere. The Therac-25 deaths were accidental, but a hacker might have been able to achieve a similar end. Factory robots have probably had the ability to assassinate assembly line workers for a while now. etc.Report

      • Kim in reply to dragonfrog says:

        How long have we had autopilots on planes?
        (granted, there are relatively fewer obstacles that you can hit in .2 seconds if you’re over a thousand feet up).Report

        • dragonfrog in reply to Kim says:

          That’s a pretty good example. Relatedly – how long have we had autopilots with connections to a network accessible either from the ground or the passenger cabin?

          Not that hacking involving burglary to gain access isn’t real, but it raises the bar something significant.Report

  2. Richard Hershberger says:

    Realistically, by the time the legal hurdles are worked out, the security on these cars will be impeccable.

    This seems pretty non sequitur-ish, if only because the legal hurdles aren’t all that high, as evidenced by the absence of insurmountable hurdles for Google to get cars on the road. Yes, we will need to work out where liability lies. This is a routine question for any new technology.

    As for the trolley questions, my eyes generally glaze over when the discussion turns to this. These are questions for undergrad bull sessions, with little to no relevance to the real world. That being said, talk about the human non-driver taking over to decide who gets run over is obvious nonsense. The human non-driver is playing Candy Crush and hasn’t a clue about the situation developing. That is the whole point of the car being autonomous. I am profoundly unimpressed by any scheme for a driverless car that involves a human devoting full attention to the driving environment, hands poised to take over control in a split second. This is utopianism, and like all utopias it falls apart as soon as actual humans enter the picture.Report

    • I am profoundly unimpressed by any scheme for a driverless car that involves a human devoting full attention to the driving environment, hands poised to take over control in a split second

      Yeah, the WORST time to have a human take over is when they are just now being startled to action from a sedate state. That’s a recipe for slow reflexes and panicky overcompensation.Report

    • Brandon Berg in reply to Richard Hershberger says:

      This seems pretty non sequitur-ish, if only because the legal hurdles aren’t all that high, as evidenced by the absence of insurmountable hurdles for Google to get cars on the road.

      For experimental purposes. It’ll be a different story when they want to commercialize it. Regulators tend to be more permissive as long as you don’t do anything really déclassé, like try to make money.Report

    • dragonfrog in reply to Richard Hershberger says:

      I agree that Realistically, by the time the legal hurdles are worked out, the security on these cars will be impeccable is unsupported by all available evidence.

      We’ve had computers for a while now. Microsoft, Apple, Sun/Oracle, IBM, HP, the foundations around various open source products – groups that do nothing but computers – they’ve had decades to get security “impeccable” and it hasn’t happened yet. And now the car companies are going to show up on the scene, with computers as an afterthought to their main line in engines and brakes and suspension, and show the IT giants how it’s done?Report

      • Richard Hershberger in reply to dragonfrog says:

        Yup: that bit has rather an air of “trust us: we’re experts” to it. But frankly, so does a lot of the autonomous car discussion. There are some impressive results, but my spidey sense says that they are busily solving the easy 95% of the problems, while assuming that the remaining 5% will be a piece of cake.Report

      • Kim in reply to dragonfrog says:

        How long does it take to crack a decent wireless password? A few months of data, and a week or two on a supercomputer.

        And that’s considered “good” security.(The old security was something along the lines of 30 minutes of data…often getting it in 5 minutes).Report

      • Morat20 in reply to dragonfrog says:

        You can limit the network connectivity of cars. And also, not to put too fine a point on it, self-driving cars won’t have the largest window for hacking — users logged in and doing stupid stuff like setting their password to “admin” and telling people the answers to their secret questions and downloading attachments from dodgy emails.

        You notice one of the big hacks was attacking a car’s sensors — equivalent to shining a blinding light into the eyes of a human driver. (The computer will not, at least, panic when that happens).

        Now a self-driving car’s computers won’t be air-gapped (there’s a whole lot to be gained by allowing self-driving cars to communicate amongst themselves), and as long as there’s a communication channel there’s a method in from the outside (in addition to spoofing sensors).

        But unlike your home PC, that is a very limited channel operating with a limited format. You can firewall the heck out of that, without worry that the owner will do something stupid (like explicitly allow a virus to open giant holes in it and call all it’s friends over for some fun time).Report

        • dragonfrog in reply to Morat20 says:

          Yes, and… I think that’s about all the car systems have going for them – an over-reliance on users not doing stupid stuff.

          But it’s not like cars aren’t linked to devices on which users do stupid stuff. Bluetooth to their phone unmanaged pocket computer, audio CDs they burned on the home computer (yes, <a href="; even that was used to hack cars). Also directly network connected features of the car – OnStar, keyless entry and ignition, alarm system, wireless tire pressure sensor, in-car WiFi, GPS that downloads traffic updates, remote disablers installed by the dealership, etc. etc. etc.

          Embedded devices have consistently had dreadful, dreadful security design. SCADA devices in plants, Smart utility meters, networked security cameras, networked parking meters, networked printers, electronically controlled safe locks, airplane-seat entertainment systems, shiny IoT fridges and lightbulbs and catfood dispensers and sex toys and thermostats and alarm clocks and quadcopters and exercise sensing wristwatches – every one of them appears on the market and it’s the 80’s all over again. Not one company seems able to learn the security lessons of the past four decades except by repeating forty years worth of mistakes in fast-forward.

          You can firewall the heck out of that, without worry that the owner will do something stupid

          You could – but by and large automakers don’t. From “From our perspective, we have rarely seen segmentation used for security boundaries, instead network segmentation is used for non-communicable network buses.”Report

          • dragonfrog in reply to dragonfrog says:

            Which, to the last point, is maybe not surprising when you think about the feature list of an even moderately fancy new car:

            The dashboard displays speed, fuel, and engine information, so there has to be communication between dashboard and engine. It also displays the GPS map, so there has to be communication between dashboard and GPS. It also displays track and album information, so it has to communicate with the stereo. The music also pauses and lets the car stereo act as the phone speakers when a call comes in, so there has to be communication between stereo and phone. And the list goes on.Report

            • Oscar Gordon in reply to dragonfrog says:

              But the communication from system to display needs to only be one-way. The computer pipes speed, oil pressure, battery state, etc. down an output only connection, no input allowed.

              The rest of that back & forth is understandable, but should not affect the safe operation of the vehicle, short of hacking the NAV system, which again, should have a very limited communication pipe.Report

          • Oscar Gordon in reply to dragonfrog says:

            I can understand why some kinds of embedded devices have a connectivity, but others baffle me. Why does the Bluetooth or USB connection of my media player/phone have ANY path to the vehicle telemetry or command system?

            I wonder how much is people just not really thinking about security because it’s all so new, kinda like how the early internet had not thought to security.Report

            • Let me dust off my old “technology forecaster” hat for a moment and make broad predictions about automotive embedded processing. Note that these are predictions of what will happen, not necessarily what should happen…

              There will be more processors, with much more code, communicating over fewer buses. Maintenance alone will require that the processing elements support a complex query-response model. Consider how many past security holes have been software that does something stupid with some kinds of malformed queries — that trend will continue. Firmware updates over the bus will become commonplace. The car makers will wait ten years too long before they realize, like Apple has, that their software signature keys are amongst their crown jewels.Report

              • Oscar Gordon in reply to Michael Cain says:

                Well aren’t you a pessimistic prognosticator…Report

              • OTOH, I’d predict that the kind of incident someone describes below — software in the stereo causes one of the brakes to do the wrong thing while the vehicle is in motion — will be more often caused by buggy software than by hacker code. Hacker software in the stereo will be much more likely to lock all the wheels while the car sits in a parking lot at 5:15 pm, and unlock only after $25 is paid to some untraceable account.Report

              • Kim in reply to Michael Cain says:

                Hacker software is much more likely to just use the spare processing time on all the processors. and that’s nearly undetectable.Report

            • dragonfrog in reply to Oscar Gordon says:

              Check out the papers at (this is a collaboration between vehicle security researchers at UC San Diego and University of Washington).

              If I understand Comprehensive Experimental Analyses of Automotive Attack Surfaces right, for example, the reason your stereo’s Bluetooth connectivity has a path to control the brakes, is because there are only a few communication buses in the car, and several devices that must bridge buses.

              So you get

              Consequently, by modifying the “bridge” ECUs (either via a vulnerability or simply by reflashing them over the CAN bus as they are designed to be) an attacker can amplify an attack on one bus to gain access to components on another. Consequently, the result is that compromising any ECU with access to some CAN bus on our vehicle (e.g., the media player) is sufficient to compromise the entire vehicle

              This was from 2011, so self driving cars weren’t really in the picture yet – but that meant surprisinly little. Pretty much the only thing not computer controlled in the non-“self-driving” cars examined, were steering and the parking brake. So, they could control the fuel injection, transmission, and each wheel’s brake individually, overriding the driver’s actions.Report

          • Morat20 in reply to dragonfrog says:

            You make the mistake of thinking “communication” necessarily is “two way” communication.

            Telling my NAV system how fast I’m going does not require a response.

            Outside stuff (your phone querying your car on location) does, for instance, but the response can be highly limited.

            Designing with security in mind DOES require designing with security in mind, but even if you’re networking the car’s controller to every other bit of the car (even over Bluetooth, although why would you do that? The critical stuff will all have to be hardwired anyways) — security is easy because giant amounts of it is one way (master computer reporting to displays) and the two-way stuff should be entirely hardwired (master computer to sensors, wheel controllers, etc).

            Which is different than adding wi-fi to a toaster, because there’s not REASON to add wi-fi to a toaster, so they leave it pretty unsecured because they have no actual goal for that. “Oh maybe we’ll…tell it to do stuff later” so they leave open APIs and access and stuff.

            Whereas with a car, it’s an integral unit talking to itself — it’d be like trying to hack your hard drive by adjusting the monitor settings.

            It’s not difficult to ensure that access to the critical bits (steering,brakes,sensors) can only be done by physically accessing their controllers or master controller. Because there’s literally no reason to design it any other way.

            All the bells and whistles should be, for lack of a better term, ‘read only’. The sensors will tell your display everything they’re doing, but the display won’t — can’t — talk back. (And the sensors won’t be listening even if you tried).

            And this isn’t complicated design, really. And it doesn’t take security experts — and you can bet mass-produced self-driving cars will be built from the inside out like that, because the risks of some idiot sticking in a CD filled with a hack that has to be designed for his model car are pretty small when it’s “won’t start” but a lot bigger when it’s a car that drives itself.Report

            • Oscar Gordon in reply to Morat20 says:

              In an autonomous car, the NAV system will have to issue commands to the car, telling where to go, how fast, etc., so a hacker could hack the NAV system, update the maps, and tell the car to take the left turn off of Albuquerque Gorge. Of course, this assumes that the car is only listening to the NAV system, and not paying any attention to its sensor suite, which would probably note that the car has left the paved road, or that the passenger is completely oblvious and has no kill switch at his/her disposal.

              In short, a hack of a system not designed by a 5 year old would need to involve compromising multiple systems in order for it to be anything more than an annoying prank. Not impossible, but hard enough that the hacker would have to be pretty motivated.Report

              • Morat20 in reply to Oscar Gordon says:

                NAV system does require updates, and they probably have to be continuous updates (not the “pay 200 bucks and we’ll use a CD to update it” crap either — like regular updates, daily, automatically) so that’s a vector.

                You’d not be hacking the car, but the maps sent to it. Which requires either credential forging (“Trust me, I’m really the map update software. Here’s my totally legit certificates”) which isn’t easy or accessing it at the source.

                Map update process would be bank-level security — the usual certificate public key dance to exchange a private key, which is…difficult to break into. When banks get hacked, it’s because either people got hacked or certificate authorities got hacked. Same with a self-driving car.

                Honestly, if you could hack the update process you’d not bother with crashing cars and instead steal a whole bunch of money or otherwise commit far more impressive crimes. (Assuming car designers wrapper their map updates using the very common security solutions for financial transactions).

                Getting to the maps themselves would require hacking people, which is doable of course. I’d worry more about security on the end that generates the maps than anything else.Report

              • dragonfrog in reply to Oscar Gordon says:

                You’re not thinking enough like a hacker.

                The hacker can provide the GPS system with a map file that uses a 3000 byte NOP sled leading into some shell code at a spot where the map parsing code copies without bounds checking into a 2048 byte buffer.

                Then it really doesn’t matter than you’re on the GPS and not some other device – you have code execution, in the car, with a connection to the communication buses you need to control everything. There has been, so far, no sign that onboard systems exercise any scepticism at all with respect to signals they receive over onboard buses. They do whatever you tell them.

                The front left break is not saying “Oh, a message for me! I accept commands from only the ABS controller. Let me check the digital signature of this command to ensure I am only taking instruction from a GM-signed ABS controller firmware before I enact it.”

                It is saying “recipient bytes 07 A3 – that’s me! Command byte D2 – that’s ‘set brake pressure’, next byte is the amount. Amount byte 00 – no brake, done!”

                Did that 07 A3 D2 00 come from manufacturer-issued code running on the ABS controller, or did it come from a hacker’s shell code running on the stereo? The brake controller has no way of knowing. Every legitimate piece of control software in the car could be shouting “brake, dammit, brake hard!” but if the stereo keeps saying “stop braking” it will.Report

              • Kim in reply to dragonfrog says:

                This rather assumes that we don’t have hackers on the autonomous car building circuit. I can assure you that they’re actively involved in the high level design.

                (Having hackers involved doesn’t guarantee the absence of backdoors, but it does make it a lot less likely for “the stupid” to win).Report

            • dragonfrog in reply to Morat20 says:

              I fully agree that it should be possible to get the security right. I just think that pretty much the entire history of computing should by now have taught us that
              (1) it will take a lot longer to get there than it should, and
              (2) new features will be constantly added and released faster than they are securedReport

            • dragonfrog in reply to Morat20 says:

              You make the mistake of thinking that communication on a car’s CAN bus that isn’t necessarily two-way, has any enforcement of one-way-ness. See the paper Iinked above for example.Report

              • Morat20 in reply to dragonfrog says:

                I think self-driving cars represent a big enough break from current car designs to force a total redesign of the controller setup.

                The lack of communication channels will have to be remedied as a complete redesign, rather than evolution, co-opting, and patch designs.

                In short — for production models, they’re gonna have to start from scratch and design the controller, sensor, and communication setups from the ground up anyways. (Test models are by nature ad hoc and basically patched, hacked, or cobbled together).

                Which means they’re unlikely to be dependent on cobbled together communication paths.Report

    • Adrian in reply to Richard Hershberger says:

      As for the trolley questions, my eyes generally glaze over when the discussion turns to this.

      Yeah, I’m with you on this one. The Trolley Problem is just another abstracted concept that doesn’t make our real issues, decisions, and dilemmas any easier whatsoever.

      That said, I tend not to look at innovations like these as the starry-eyed scientist might or the average person who wants to, as the OP mentioned, save time “by enjoy[ing] mind-expanding activities like reading a book or catching up on work tasks while on their commute.” Are people really this optimistic about other people? I don’t have any statistics or facts to back up by cynicism, but I can only imagine about one out of thirty people taking advantage of said situation by ‘bettering themselves.’ Which leads me to ask the person who is salivating at the opportunity to save more and more time, ‘for what exactly are you saving all this time?’ Which is why your “candy crush” point served as more than just a side point – it seems to me to be closer to the truth than some reading mind-expanding literature – whatever that means.Report

      • Glyph in reply to Adrian says:

        Which leads me to ask the person who is salivating at the opportunity to save more and more time, ‘for what exactly are you saving all this time?’

        …why do we care? If they want to play Candy Crush or sleep, so what? Isn’t that still a net benefit to me and society, if I arrive at my destination relaxed and happy instead of tense and tired?

        And anyway, if one person out of thirty spends his commute time reading Moby Dick, that’s one person more than did so before autonomous cars.Report

        • Mike Schilling in reply to Glyph says:

          I actually read that on the boat. Fortunately, no whales tried to turn us into Fishmeal.Report

        • Adrian in reply to Glyph says:

          I guess my point is I don’t care what they do, but let’s be careful about using the argument ‘well people will have more time to better themselves and be self-enacting creative human beings.’ It’s far too optimistic. Even, as you point out, unnecessary.Report

      • Mike Schilling in reply to Adrian says:

        There’s already data on this: what are people who commute on pubic transit doing? I commute via ferry boat, and I don’t see a lot of books.Report

        • Kim in reply to Mike Schilling says:

          I see tons, but then again, I’m on a bus, and I do occasionally look over people’s shoulders. Lotta kindles, some hardbacks, lotta Game of Thrones.Report

          • Mike Schilling in reply to Kim says:

            I see almost entirely phones and tablets, with key-pressing patterns that don’t look like occasional page turning. A friend of mine who’s been riding it for decades says it used to be much more social, but mobile devices have ruined that.

            Personally, my time is split maybe 40% reading, 25% napping, 15% work, 20% playing or surfing.Report

            • Kim in reply to Mike Schilling says:

              I go from a relatively wealthy neighborhood, straight by a college, and into downtown (with tons of finance/math people on the bus).

              Some people are “checking news” (which I take to be all the “reading small chunks of info” key patterns). Occasional person playing Minecraft (on a computer) or playing “single finger games” on their telephones.

              There are a lot of people who talk, too — less than there used to be, but if you get a conversation started, people will chime in.Report

            • Chris in reply to Mike Schilling says:

              Once upon a time, I was usually the only one on a bus reading a book, but now I see tons of people doing so. I think that’s partly generational (a lot of the people I see reading are probably in their 20s or early 30s), and part of it might have to do with the fact that I ride through a major university campus (though I always have, so even that can’t explain the change).

              Buses did tend to be more social here. Until cell phones with mp3s or streaming music became ubiquitous, I almost always knew all of the regulars on my buses by name, and we were always talking. Now I only know one regular rider, and her only because she and I wait at the same stop, so we talk there. We almost never talk on the bus, because it’s so quiet that it feels awkward.Report

            • Oscar Gordon in reply to Mike Schilling says:

              People on the WA State Ferries have been known to stash paperbacks under their seats on ferries for reading during the crossing.

              The ferries have been known to, upon occasion, flip the ferry as a prank so people find new books.Report

              • Kim in reply to Oscar Gordon says:

                god, that’s adorable. you see dolphins often out there?Report

              • Oscar Gordon in reply to Kim says:

                Not often, more Orcas, Grey whales, and seals.

                PS The link is kinda funny, in that a juvenile orca decided to mix it up with some grey whales, and not only got smacked down by the grey whales, but the orcas mom arrived and scolded him for being un-neighborly.Report

              • Kim in reply to Oscar Gordon says:

                hmm… yeah, that might have been seals I saw last time I was up there (wound up taking tons of WA public transportation while on my honeymoon.), not dolphins. (when you see something smallish (not orca sized) jumping in the wake…)Report

            • Kazzy in reply to Mike Schilling says:


              I read books and one I figure out my Nook, I’ll read on that. I also read websites (like OT) on my phone. So you’ll often see long periods of reading followed by prolonged periods of typing as I write a comment.

              I don’t have games on my phone. Like, any. If I’m not reading, I’m usually listening to a podcast of some sort.

              But this is NY. No one talks unless they have to.Report

        • Glyph in reply to Mike Schilling says:

          I drive, and I NEVER see books. So what’s the point here?Report

      • Chris in reply to Adrian says:

        The Trolly Problem, as a Gedankenexperiment, is of course meant as an intuition pump of sorts, not as a guide to action, but I think it’s clear what people mean when they raise it in such contexts: how will an automated driver make decisions when no good options are available, only varyingly bad ones?

        Forget the trolley context, and imagine a scenario in which an autonomous vehicle is faced with a large obstacle at speed, a situation certain to cause damage and likely injury to any occupants of the automated car. What does it do: slam on the brakes and strike the obstacle at the lowest speed possible? Swerve to the left into oncoming traffic? Swerve to the right off of a steep embankment? How does it make this decision? And if it makes the “wrong” one (whatever that turns out to be), who’s responsible?

        I imagine there are plenty of folks and Google and other places working on this right now, of course. But it’s not a dismisssible problem, given how often the world throws Scyllas and Charybdises at us.Report

        • Mike Schilling in reply to Chris says:

          It’s a case where

          1. Logically, self-driving cars don’t have to be perfect, just better than people.
          2. There will be much hand-wringing (and lawsuits) when self-driving cars aren’t perfect.
          3. There will be much hand-wringing (and lawsuits) when there was no perfect answer.Report

          • Chris in reply to Mike Schilling says:

            Yes. I’m thinking particularly of 3.Report

          • Iron Tum in reply to Mike Schilling says:

            4. Any situation which would make operating a motor vehicle (drunkenness, distraction, whatever) illegal will also make it illegal to be in the “driver’s” seat of a driverless vehicle. It will also be illegal to be in a driverless vehicle without someone in the “driver’s” seat. Sobriety checkpoints will continue unabated and the rate of removing people from vehicles will increase to compensate for the lack of probable cause.

            4a. Because of the reduction in DUI arrests from non-checkpoint stops, the fines for a DUI infraction will be increased by 5000%. Ditto for speeding violations.Report

        • Troublesome Frog in reply to Chris says:

          I’d find those questions a lot more interesting in the context of autonomous cars if I thought there was some reason to believe that human drivers were already doing a bang-up job of handling those no-win situations. I strongly suspect that they are not. The most likely outcome of those rare situations seems like it’s probably the human being flailing out a bit and then saving his or her own skin. This is something an autonomous car should be able to do pretty well, and it should reduce the overall number of no-win situations while it’s at it.

          In fact, let’s assume that the car does the “morally wrong” thing (whatever that is) 100% of the time and humans get it right 50% of the time. If the car also cuts the number of times it gets into those situations by 80%, doesn’t that make the question basically moot?Report

        • Adrian in reply to Chris says:

          how will an automated driver make decisions when no good options are available, only varyingly bad ones?

          I guess my point here was that this is the wrong question. It assumes that that what we create – driverless cars driven by ‘robots’ – will be free of human error. And it will inevitably make a decision that, like all human decisions, we have to live with. Which is why the ‘who takes responsibility’ question is important – it must necessarily be the passenger who owns said car, no? But the reverse question might be insightful as well, I think: why would anybody hold a driverless car at fault if it is supposedly better than human drivers? There’s never going to be an agreement on the proposition that ‘driverless cars are better than human drivers’ and if there is, the problem of responsibility will be moot because, like the benevolent king, every decision will flow from some ‘thing’ better than us in every sense of the word. And we will be able to all nod in agreement that ‘well at least this is preferable to humans driving.’ I can imagine court cases using expert witnesses as attempting to determine what a human would do in said questionable circumstances to determine fault of the company or, as it would be, fault of no good possible outcome whatsoever. I am reminded of the three laws here… it will always be a matter of programming (i.e. protect the driver first, protect pedestrians not in vehicles second, etc.)Report

        • Michael Drew in reply to Chris says:


          I think one interesting issue this discussion raises is that SDT is going to create a distinction in the case of many car trips (including all single-occupant trips) that has never existed before between the occupant and primary controller of the AV technology, and the entity that is in direct control of its maneuvers.

          So, presumably under normal circumstances there are situations in which car occupants sue drivers of vehicles in which they rode over injuries caused by negligent driving. But presumably the bar is fairly high there, because there is an element of consent in choosing to ride with someone. And of course, drivers don’t get sued after accidents by people in their vehicle unless they have passengers with them at the time of the accident.

          The question will be, when you get into an AV alone and instruct it to go, will you be assuming the legal position of someone who is riding in a normal vehicle with a driver of X degree of skill & responsibility – or perhaps accepting just a bit more responsibility because the consent you have given will be spelled out somewhere and be even more thorough than that implied by getting in a car with another driver? Or by being the human who is directing the AV to operate, are you not just consenting to assume the risk of being a passenger, but instead retaining some (perhaps large) portion of the responsibility for the operation of the vehicle, as if you were still the driver to some degree?

          As you say, these questions will have to be worked out by lawyers and firms developing the technology, and very likely is being worked out (or at least worked on) right now.

          Shorter: it’s never been the case before that you’ve had solo occupants of vehicles suing the driver of a vehicle they were riding in (and driving) when they were injured due to accident. Because, same person. Nor have you often had, for example, kids or wives suing the operators of vehicles in which they were injured when the operators were their dads or husbands. Because, Dad. But now there could be people to sue in those circumstances. So that could cause some weirdness for a while, unless it’s all nipped in the bud by explicit liability limitation. (But even then, lawyers make their $ by getting around those kinds of provisions.)Report

          • Chris in reply to Michael Drew says:

            Should be an interesting time. I have real hope that driverless cars will significantly reduce the number of deaths and serious injuries from motor vehicle travel, not to mention pollution from emissions. I even think it will push us towards more public transportation as the cult of driving diminishes.Report

    • LeeEsq in reply to Richard Hershberger says:

      Good luck with that because a big appeal of driverless cars to many people is that they aren’t going to have to pay attention on the road and will be able to read, work, or play Candy Crush. I imagine that others are going to want to do more interesting activities while being driver around. We have many people who do not pay full attention with current driver cars. Making sure a computer does most of the work will only increase the number of people not paying attention.Report

      • Kim in reply to LeeEsq says:

        99% of people driving are not paying full attention at any point in time.

        Hand off wheel, listening to music, talking

        foot not in optimized “best reaction distance” posture.Report

        • Richard Hershberger in reply to Kim says:

          99% of people driving are not paying full attention at any point in time.

          I should hope not. The vast majority of driving does not require full attention. Few people have the capacity to single-mindedly focus for long periods of time on a task that does not in fact require such concentration. Trying to force this would be a recipe for wackiness. Listening to the radio or engaging in conversation are good accommodations to this reality.Report

  3. DensityDuck says:

    I’m always annoyed when I see stuff like this.

    “Oh, you can HACK THE LIDAR!” Bro, if that’s what you call “hack”, then I can “hack” your tires by throwing tacks on the road. I can “hack” your engine by sticking a banana up the tailpipe. I can “hack” your window in the most literal sense by smashing it with an axe. What we’re all supposed to be scared about is people using remote-control methods, and that is more along the lines of breathless Nineties stories about how i-criminals would cyber-hack the e-system and destroy all the banks, make dams open up, and make your computer explode just as you finish reading this email, EEEEEEK! (boom)

    One thing you find in a lot of these stories is a little bit about “after sitting in the car for a few minutes” or “with just thirty seconds of access to the car’s onboard computer” or “with a few quick button-presses to enter override mode”. Which means that you can defeat these master hackers by keeping the damn door locked.Report

    • Kim in reply to DensityDuck says:

      Hacking the lidar was silly, I admit.
      Next, I will hack the Drivers! (using pikachu! flashing lights! inducing seizures!)Report

    • Troublesome Frog in reply to DensityDuck says:

      That’s the real issue: a lot of people don’t really think about there being multiple levels of “insecure” that are relevant in different situations. Just about everything is breakable given enough effort. The cheap Kwickset lock on a front door can be picked in seconds by anybody with a few hours’ practice. Barring that, it’s easy to break a window. But locking your doors and windows deters most casual theft and keeps the house reasonably secure.

      Yes, digital security has a new tier of “insecure” that physical security doesn’t have: The remotely exploitable hole that makes it possible for one guy to break into every house in the US while eating a sandwich. We see that failure mode on stuff like web servers and Internet enabled nanny cams, but I don’t think we’ve seen it with car electronics, and it should be avoidable given good design practices. The problem is that that’s the failure mode everybody assumes is happening when a news report says a system was broken.

      Giving a hacker the ability to crash every car in the US from his cell phone is very different from the types of car security flaws that usually make headlines. Typically the “hack” is, “Somebody gets into your car and connects a custom-built physical device to the OBD-II harness that injects bad stuff on the CAN bus.” OK, fine. That’s a real problem that we should mitigate. But it’s not the end of the world. It’s analogous to saying, “Given a few minutes of access, somebody can cut your brake lines.”Report

    • Oscar Gordon in reply to DensityDuck says:

      In the morning, the sun “hacks” every driver heading east, then does it again in the afternoon for drivers heading west.

      Naughty, naughty sun…Report

  4. DensityDuck says:

    “How is a driverless car going to respond in a situation in which hurting someone could be inevitable? ”

    How is that situation going to arise in such a way that a human driver would have been able to not hurt someone?

    Or is it just a matter of having someone to blame for that inevitable occurrence?Report

  5. Saul Degraw says:

    Shoddy infrastructure might be the biggest road block (pun intended) to driverless cars in the United States:

    • Chris in reply to Saul Degraw says:

      Doubt it (by the way, Loomis is a fool; there, I said it). That a Volvo driverless car made for European roads couldn’t quite figure out American roads is irrelevant to the prospect of driverless cars in the U.S.. The existing American driverless cars (at least Google’s) already drive in unmarked lanes, even (famously) go around cars into non-lanes (remember the one that caused the accident), and do just fine in as chaotic a space as a large, busy grocery store parking lot.

      The biggest obstacles to driverless cars will be the legal ones. The practical issues are melting away one by one.

      The parking lot thing: I dunno about in California, but here, driverless SUVs drive around a large HEB and a large Walmart parking lot pretty much constantly. This one is almost certainly stalking me:

      • Richard Hershberger in reply to Chris says:

        That a Volvo driverless car made for European roads couldn’t quite figure out American roads is irrelevant to the prospect of driverless cars in the U.S.

        This is a fair point, but it is pretty funny that the corporate reaction is that we should make our roads consistent with its needs.Report

      • greginak in reply to Chris says:

        I think in some cases bad roads will be an issue. It isn’t odd for us to have large potholes for quite a while. Or snow or gravel covered roads. Or all at the same time. I’d want to see A LOT of video of a driverless car negotiating a standard ( crappy) alaskan road before i would completely trust it.Report

        • Chris in reply to greginak says:

          Bad roads are an issue, for sure. I can’t imagine any company wants to try their driverless cars out in Michigan, for example. But much of what bad-road driving is about is avoiding holes when possible and keeping steady despite bumps and dips and such, and I guarantee you that driverless cars will do that better than most of us.Report

          • Kim in reply to Chris says:

            Big Auto’s testing in michigan (forget whether that’s GM or Ford). They’re not shipping their engineers anywhere, and snow exists for large portions of winter there.Report

        • Burt Likko in reply to greginak says:

          Oh, relax. California’s roads are falling apart, too, and Googlecar is getting tested extensively here.Report

  6. Kazzy says:

    Why are the accident avoidance ethics any different with autonomous cars than with human-directed cars? These decisions are already being made every day. Do we have laws or ethical expectations for how humans approach them? Why do we need some standardized system? Why can’t GM program their cars one way and Ford another way and let consumers choose the one that fits with their value system?Report

    • Kim in reply to Kazzy says:

      The flowcharts would get really, really complicated, for one thing. I’m not certain people would be terribly good at evaluating which one aligns with their ethics…

      [We do have laws and ethical expectations, but they have TONS of fuzziness to them. Killing someone to avoid bumping someone’s bumper is going to get you manslaughter. Hitting one car at 5mph vis hitting another at 10mph? Nobody’s going to care… all that much.]Report

      • Adrian in reply to Kim says:

        At the end of the day, it would require something that is the equivalent length of a house mortgage explaining the decision making processes of the car…

        Which will result in, inevitably, the new owner signing every page and thus taking responsibility for all the cars decisions. I may fault my dog for biting someone else’s dog, but I am certainly the one taking responsibility and suffering the repercussions.Report

      • Kazzy in reply to Kim says:

        All that assumes perfect information. If I choose “Hit the person” over “Scrape my rims”, I’m a monster. If I choose, “Swerve left!” over “Swerve right and scrape my rims!” with no knowledge that a person is left, what then? Should I have known? What if that person is where they shouldn’t have been?

        I guess my argument is that faulting cars for not being perfectly ethical ignores that humans are not perfectly ethical. This isn’t a new issue. Just one we need to confront in a different manner.Report

    • Troublesome Frog in reply to Kazzy says:

      What if you have to choose between hitting two kids and one of them is going to grow up to be the next Hitler, but it’s not 100% obvious which one?

      Your move, Google.Report

  7. Damon says:

    The us gov’t has already given off body language that they don’t like encryption for your phone. Ain’t gonna be different for your car. They will want access to your car, by hook or crook, and they’ll get it. Won’t need a warrant, just plug in the vin number and they’ll know where your car is and what it’s doing. The car will pull over itself, and lock you in it. (And you’ll still get shot because of officer safety)

    “By the time automated cars are available to the widespread general public, they should be safer than any vehicle available right now.” And they’ll be BORING.Report

    • Adrian in reply to Damon says:

      And you’ll still get shot because of officer safety

      Haha, well said. And I quite agree: boringness is my biggest issue with this continued march of science trying to ‘figure it all out.’ If that’s what we do anyway, and science figures out the meaning of life… what’s the point. Admittedly, I probably took your point a tad further than was actually intended.Report

      • Damon in reply to Adrian says:


        Safe IS boring. Risk / Danger is what makes you feel alive. If you’re smart, it’s a minor risk, or a bounded risk, like sky diving or scuba diving. What’s next, safety rails on the grand canyon? *sigh*Report

        • Adrian in reply to Damon says:

          Sometimes when I let my cynicism run rampant, I am inclined to believe that people want to just be able to hand their children a manual on the objective meaning of life. Of course the handing of this book down is in the instruction manual itself…

          This is the Enlightenment’s fault.Report

  8. Michael Drew says:

    I’d be interested in the OT Community’s reaction to the potential problem (or “problem”) with self-driving auto technology (“SDT” for me) envisioned here, or to most any of the related questions raised in the comments to that post.

    Including but not limited to such questions as:

    – Is SDT, for all the talk, per Atrios, really going to turn out to be a non-starter in the U.S.?

    – To what extent will/would SDT shrink the number of surface shipping driver jobs assuming the above is not the case?

    – To what extent will the gains in road safety (fewer injuries, deaths, lower costs) be realized?

    – How much of a time-saving technology will SDT turn out to be in practice? Will we all (those of us who manage to adopt SDT) get more sleep than before? Get more reading done? Time with kids improved due to full engagement? Get more work done? Etc.

    – Will employers capture a large share of the driving time that’s saved, making the time savings not a valuable reason to pursue SDT? (To me this is a much broader social question, so that applying the argument to this one particular case and accepting it would imply a conclusion that we should basically just abandon all time-saving or even just all new technologies of any kind. (The author cites email and smart phones as examples of where this has happened with other technologies, but to my recollection neither of those technologies were much heralded as potential time-savers, but much more as capability-enhancers, and ones largely developed and pioneered explicitly (though not exclusively) for the purpose of making people more productive in their work. SDT to my understanding isn’t presented as having that intended application.) The author of the post places considerable weight on that view in comments, though ,so I’m curious to get other perspectives on it.)Report

    • Michael Drew in reply to Michael Drew says:

      Atrios clarified his prediction. To my ear, there is a bit of “we shouldn’t make them work because we should do other things instead” mixed in here (well, especially since he says that explicitly), and a bit of realization that maybe his preference colored his predictive analysis a bit. But I for one sort of agree with him: I do expect it to happen, but I don’t know if I expect the roads to be filled with a fleet of which more than 20% of vehicles are SDVs in my lifetime.


      I Could Be Wrong On The Internet
      It wouldn’t be the first time. But the reason I harp on what I consider to be the ridiculousness of self-driving cars is that I think the only way they’ll “work” (and even then, not well enough) in my lifetime is if an absurd amount of money is spent on building public infrastructure to make them work, instead of building more sensible things like mass transit. I also don’t buy the various pro-urbanism arguments for them, which amount to “we won’t need urban parking lots or to own cars” because you still have the commute problem. In a car-centric world, you have to have enough cars so people can go to work by car if you don’t have that lovely mass transit system.

      If they really do work as promised without upgrading the technology of our entire road network, maybe that would be a good thing. I just don’t believe they’ll work well enough. Of course in the future we might all upload our brains into robot bodies and finally get our jetpacks. When I say it won’t work, I mean it won’t work in anything resembling the timeline that boosters say it will. I don’t think it’ll work in my lifetime, and I plan to live at least a few more years.


      • Oscar Gordon in reply to Michael Drew says:

        Go read my post on SDT again. Yes, there will need to be infrastructure upgrades, but they can be minor.

        As for public transit, I will say this again because public transit hardliners always seem to handwave this away – public transit fails at the margins, and as long as it does, it will always be a second fiddle to cars, either POVs, or public Autonomous Vehicles (AV). Mass public transit is great for getting people from area A to area B, and in most cases that is sufficient. For the regular daily commute or casual user, those people can walk the last bit to their destination.

        It fails when it can’t get close enough* so the walk is more than people want to deal with, or when people are moving items that are heavy/bulky/difficult to carry, or managing infants/children, or are disabled, etc. Now of course, people figure out ways to do such things on public transit, but those are people who either really love public transit, or have no choice. Since the funding for public transit is largely approved by people (i.e. middle class and above voters) who are not interested in having their choices deliberately constrained in order to satisfy a minority of advocates, it will be a problem for the long term until those advocates start treating those margins seriously.

        Now, an urban system where extensive mass transit can largely meet the needs of the volume of arterial highway traffic, then you can do things like replace highway lanes with rail lines and we don’t need AV to get from home to work, but we will still need cars to get from the house to the rail station near home, and from the rail station near work to the office, or store, etc. And occasionally a person will need a car/van/truck to handle a task that just can not be done on public transit.

        For those margins, a fleet of on-call AVs, that can be called upon in a moment, or booked in advance, will be instrumental in making the rest of the transit system appealing, and that fleet can be relatively small and can utilize automated parking garages to reduce the space requirements for parking, since once a car drops a person off, it goes to get another rider somewhere, or reports back to the garage for charging or maintenance or idle time.

        *Often, mass transit can get you ‘close enough’, but the path to getting ‘close enough’ has to be convenient. If getting close enough involves multiple transfers, especially if taking those transfers involve lots of waiting or riding longer distances than necessary, then the convenience factors falls off quickly for people who are time sensitive.Report

        • Michael Drew in reply to Oscar Gordon says:

          It sounds to me like you’re kind of envisioning an extension of mass-transit involving a fleet of dedicated road-going AVs. At least, in your vision of how they’re used, it sounds to me like the agencies running them might as likely be existing municipal metro transit agencies as anyone else.

          Also, are you aware of how strident and lecturing you sound? Your first sentence is literally in the command form (imperative mood). Sets a certain… tone. Sorry, I didn’t even recall your AV post.

          I’m just asking a few questions here, Oscar. I said responses to any questions raised in that thread were welcome, but the preference for mass transit wasn’t even one that I listed as of particular interest to me. I don’t have any position on that. I didn’t mean to say I agree with Atrios about his preference for mass transit over AVs; I was just saying that I’m skeptical we’re going to see a big influx (I gave a number) of them into the general fleet in our lifetimes. But I could be wrong, as he also says he could be! We could be wrong!

          Ease up just a touch, would you?Report

          • Oscar Gordon in reply to Michael Drew says:

            Yeah, I was a bit. Sorry, should not have pointed that at you.

            My post is here, if you are interested.

            Thing is, AVs have challenges facing them, but critics (and advocates) tend to focus on the wrong challenges as being key. The control technology is already here, the infrastructure upgrades would be minor and can be phased in gradually (and will certainly cost less that tearing up a road and replacing it with rail), & affordable sensor suites are easily within sight; the hurdles are learning AI (software) & security (what @morat20, @dragonfrog, & I were discussing elsewhere).

            The one thing Atrios said that is probably correct is that such technology will be readily adopted elsewhere first, before finding a home in the US.

            Anyway, I’ve seen that attitude about AVs from transit advocates before, where AVs are dismissed or trashed because it doesn’t gel with their grand vision for public transit. Where that last mile concern is hand waved away or ‘solved’ via walking, or bicycles, or some other solution that would be great for them, but ignores the needs or preferences of anyone not like them. I’m sure it seems like a very minor thing, but that last mile will always make or break a transportation system, be it cargo, people, or data.Report

            • Michael Drew in reply to Oscar Gordon says:

              I’ve seen that attitude about AVs from transit advocates before, where AVs are dismissed or trashed because it doesn’t gel with their grand vision for public transit.

              Now that I’m aware that people (unjustifiably afaict) see these as direct competitors for resources & I’m looking for it, I’m seeing it all over. Sad!

              I overreacted a bit there as well, sorry. Thanks for the response. And the passion!Report

            • Anyway, I’ve seen that attitude about AVs from transit advocates before, where AVs are dismissed or trashed because it doesn’t gel with their grand vision for public transit.

              Small electric cars, including autonomous ones, are large-scale compliments to other transit only if you assume the suburbs, both housing and the jobs, continue. That assumption is anathema to many (most?) public transit advocates. I’ve usually summed up their strategic plan as “If we can just make the suburbs as unpleasant as we’ve always claimed they are, people will move back into the cities.” This is not, IMO, a particularly good plan.

              Just on the basis of energy considerations, the suburbs are going to change over the next 30-50 years. That’s not the same as “go away”.Report

              • Kim in reply to Michael Cain says:

                People LIKE streetcar suburbs, they really, really do. But they’re Expensive, so not everyone can afford them.

                Look to the poor being moved out into the suburbs, which is overall a better strategic allocation of resources (having the “good stuff” centralized means that fewer miles need to be driven).

                The public transit advocates I know are far more focused on “let’s improve the things that are horribly wrong right now” rather than “let’s get rid of suburbs”. Then again, I know logistics experts, so they come up with fun things like “Let’s have gondolas! It’s cheaper than building another road!”(and way more fun, bonus!).Report

    • Is SDT, for all the talk, per Atrios, really going to turn out to be a non-starter in the U.S.?

      I think it will start in niche markets. Long-haul trucking is the one that gets mentioned most often. Another likely one is the elderly. They want to stay in their own homes, and keeping them in their own homes is much less expensive than institutions. SDT is potentially an important piece of allowing more of them to stay in their homes longer.Report

      • Michael Drew in reply to Michael Cain says:

        Trucking as a leading market for SDT is certainly the kind of thing that post was concerned about. Do you share the concern?

        I have a bit of a tough time seeing the entire trucking industry as a “niche” market for SDT, though.Report

    • Kim in reply to Michael Drew says:

      – Is SDT, for all the talk, per Atrios, really going to turn out to be a non-starter in the U.S.?

      No. Uber is betting on it, and they’ve got real reason to want their human employees gone as soon as possible.

      – To what extent will/would SDT shrink the number of surface shipping driver jobs assuming the above is not the case?

      A LOT, but look for a graduated phase-in, and dwindling of money paid before dwindling of people hired.
      1) Expect main routes to be mapped and used first
      2) Expect SDT to become mandatory for certain times of day (thus extending the driving time to 24 hours)
      3) Expect SDT to make long-haul driving into less of a “hazard pay” and more of a “slight amount of supervision” job.
      I know someone who’s designed SDT, so…

      – To what extent will the gains in road safety (fewer injuries, deaths, lower costs) be realized?

      Nearly 100%. This will work, once the legal stuff gets busted through.Report

    • Francis in reply to Michael Drew says:

      My own take is that the first successful deployment will be neighborhood electric vehicles not permitted to access the freeway (like those oversize golf carts I see around my community).

      A. It gets drunks home from bars and the elderly around town.
      B. The crashes will be at low speed and in a low weight vehicle.
      C. If the machine gets confused it can relatively easily find a shoulder and stop.
      D. The capital cost should be pretty low.
      E. It allows for small-scale rollout in communities where the roads and the weather never get too bad.Report

      • Oscar Gordon in reply to Francis says:

        Absolutely, and in residential communities, it’s an excellent last mile solution.

        Something like that, that could get me & my son down to the park & ride, would make taking the bus so much easier.Report

      • Michael Drew in reply to Francis says:

        It seems to me as though (C) needs to be vanishingly rare or people are going to be a little spooked by the technology as long as they don’t feel like they can count on it not having to pull over and get it bearings.Report

        • Francis in reply to Michael Drew says:

          I think that early roll-out days will inevitably have some bugs. I’d much rather have first adopters (and the people around them) being in a quiet light-weight oversized golf cart going 30 in the suburbs than an 18-wheeler getting confused doing 70 on the I-5.

          For myself, there are any number of restaurants my wife and I don’t go to, because I don’t want to deal with the drive getting home. (I’ll be signing up for Uber any day now.) But it would be really kinda cool to participate in the roll-out of truly revolutionary technology.

          So if I were Alphabet and Uber and GM and Solar City and Enterprise Car Rental, I’d be forming a JV which (a) sublease land in Enterprise parking lots, (b) installs solar cells in pop-up frames that cover the parking lot, (c) places an Uber employee in the office to plug in the cars day and night, (d) leases NEVs from GM, (e) leases the self-drive tech from Alphabet, (f) contracts with the City (of Venice, CA, for example) to use street-side and muni. lot parking so the car doesn’t have to drive itself back to the lot after every trip; and (g) puts an Uber driver in each car for at least the first six months to a year, or until people calling for the car want the discount price that excludes the driver.Report

    • Road Scholar in reply to Michael Drew says:

      – To what extent will/would SDT shrink the number of surface shipping driver jobs assuming the above is not the case?

      You mean my job? Short answer: some, not all, and not all at once.

      Longer answer: Driving, per se, is only part of my job. There’s also pre- and post-trip equipment inspections, fueling, loading and unloading, and connecting/disconnecting trailers to/from tractors. And that’s just assuming the trailer is your basic box on wheels. Flat bed work also entails load securement with straps and/or chains plus tarping. And you can’t just strap and tarp and drive; the load can settle or shift causing the chains or straps to loosen, tarps can become discombobulated, and so you’re constantly monitoring that in your mirrors and periodically stopping to check and adjust it all. Then there’s tankers hauling various liquids from milk to truly ghastly chemicals, pneumatics, end dumps, and belt bottoms hauling dry bulk stuff, car haulers, etc. These all entail essential activities well beyond just driving.

      Seriously, driving down the road is the easy part of my job. And of the non-driving stuff only a fraction can be automated, eliminated, or off-loaded to someone other than an onboard operator. And doing so will in many cases require significant alteration of processes, practices, and facilities, as well as g-d only knows how many trailers of various kinds.

      So what’s the upside? In a word, productivity. Truck drivers are legally and practically limited by Hours-of-Service rules and human endurance to about a 50% duty cycle at best. I would estimate that, as an Over-the-road driver, somewhere around 75 to 90% of my miles are on the Interstate system. For example, I just finished a run over the weekend of almost 1800 miles from the SF Bay area to Oklahoma that was entirely on interstate highways except for about a mile or so on either end. At 60 mph that’s about 30 hours of actual driving (apart from getting out of the Bay on a Friday afternoon. Yechh.). If I could just navigate to the big road, put it in full auto mode, and sleep, read, play, masturbate, etc while the truck drives itself, I could make the run in literally half the time.

      Other types of trucking work, like local delivery, would realize much less efficiency gains. There the gains would manifest more in the safety arena. For example, the most common type of trucking accident occurs when backing into a dock or parking space. A truck that could do that autonomously and safely would save the industry millions annually.

      So here’s my prediction: SDT will be, in some ways, the tip of the spear due to the economic potentials. It will be an evolutionary rather than revolutionary process due to the enormous sunk investment in the status quo. The job of truck “driver” will metamorphose into “Heavy Truck Operator” as driving, per se, becomes a smaller portion of the job and employment in the sector will gradually decrease as each truck/operator pair becomes more productive. There will be very few — if any! — trucks running around with no human on board; keep in mind, trains run on tracks and they still require operators.

      Finally, and this applies to cars as well, I predict the standard controls we have now — steering wheel, pedals, etc — will give way to something like a joystick as “driving” becomes more a matter of telling the car’s computer system where you want to go.Report

      • Michael Drew in reply to Road Scholar says:

        RS, thanks so much for weighing in! (Haven’t actually digested it yet; if I have thoughts I’ll follow up.) I was hoping to get your take, but wasn’t sure how much we were seeing you around here recently. I’m glad you saw this.Report

        • Michael Drew in reply to Michael Drew says:

          …I was thinking along these lines as well: it seems like SDT might shrink the size of your job, and this may shrink the number of you needed somewhat, but not that it will eliminate you on a one-for-one basis, where each time a driving computer goes into a semi cab, a driver will be put out of work. It almost seems like your job could shift to something more like a train engineer from being the moment-to-moment physical driver of the vehicle to managing the various on-board technologies that will take over that task.Report

          • Kim in reply to Michael Drew says:

            We also may be able to get bigger trucks on the road…Report

            • Autolukos in reply to Kim says:

              Also maybe convoy multiple trucks with one operator? Both would depend on the regulatory environment, of course.Report

              • Morat20 in reply to Autolukos says:

                IIRC, they’re already working on semi-self-driving trucks that convoy. Most of the hardware is there, and since “follow the truck in front of me” is MUCH easier than “drive on your own” you can swap which guy is ‘leading’ the convoy.

                It’s effectively road trains — we’re talking trucks less than a foot from each other. Surprising fuel savings, you don’t need to add much to the trucks (compared to SDT) and a smaller problem set to solve (stay this distance away, and of course some method of instantly communicating speed changes — especially sudden brake applications — from the lead truck).Report

              • Michael Cain in reply to Morat20 says:

                Certainly such would solve one of the problems I find particularly irritating on rural interstates — one slow big truck taking ten minutes to finish passing another, while the traffic backs up behind them (slow meaning the trucks want to go 10 mph slower than the cars do).

                Too-close following by too-long a convoy on the interstates creates a different problem that will have to be addressed. Someone pulls out to pass the line, and half-way through discovers that their exit is coming up right now.

                I’m always — amused? — by various proposals to make the interstates into closer approximations of railroads, rather than simply requiring the Class 1 railroads to provide scheduled freight service.Report

              • Kim in reply to Michael Cain says:

                You only see that out west. Out east, the trucks are driving fast enough (and the speed limit’s slow enough) to not matter.

                [Except on steep hills, where you hope to have a third lane, or in the rain, in which case the trucks are going at “safe speed”]Report

              • Road Scholar in reply to Michael Cain says:

                Michael Cain,

                Certainly such would solve one of the problems I find particularly irritating on rural interstates — one slow big truck taking ten minutes to finish passing another, while the traffic backs up behind them (slow meaning the trucks want to go 10 mph slower than the cars do).

                Ugh. That’s just some truck drivers being assholes. I don’t bother trying to pass another truck unless there’s no car traffic coming up behind me in the left lane and/or the guy in front of me is going considerably slower than me, which doesn’t happen much since I’m governed at 62 mph. And if I have another truck passing me slowly like that and I see traffic backing up behind him I’ll back off so he can get around faster.

                It’s basically just a matter of being courteous, which seems to be really difficult for some folks.Report

              • Oscar Gordon in reply to Road Scholar says:

                The removal of ego from the operation of motor vehicles will probably do more to improve road safety than anything else.Report

              • Kim in reply to Oscar Gordon says:

                I think the panicky people do more damage,on the whole. The egoists pass you and continue along the road.Report

              • Road Scholar in reply to Morat20 says:


                IIRC, they’re already working on semi-self-driving trucks that convoy. (snip)…

                IDK, there’s some unavoidable physics that militates against those sort of following distances even with instantaneous computer reaction times. A fully loaded semi, at the current standard weight limit of 80K lbs, requires approximately a football field length — 250 – 300 m — to come to a complete stop from 60 mph under ideal conditions, exclusive of reaction time.

                If that lead truck were to suddenly encounter an obstacle, like a drunk human driver crossing the median, or have some mechanical failure… well, it might work out as long as the heaviest truck(s) are in the lead. Otherwise, boom bang boom…Report

            • Road Scholar in reply to Kim says:

              We also may be able to get bigger trucks on the road…

              Why? The current limits are mainly a function of physical infrastructure constraints. Weight limits on bridges, length limits for curves and cornering, height of overpasses, etc. Putting a robot in the driver’s seat doesn’t change any of that.Report

              • Kim in reply to Road Scholar says:

                Weight limits remain an issue,of course, but length limits may not be so. I know buses in the city that bend — why shouldn’t trucks have the ability to have multiple pieces that can negotiate a proper turn?Report

              • Road Scholar in reply to Kim says:

                They already do; they’re called doubles and triples. But it’s just unavoidable geometry. The trailing axle will always cut the corner a bit — hell, it happens with your car but you just don’t notice — and every added trailer will cut a little bit further.

                I mean… you could make trailers steerable I suppose. They already do that for the seriously over-length rigs hauling windmill blades and such but it’s hella complicated and expensive.Report

              • Oscar Gordon in reply to Road Scholar says:

                Road Scholar: I mean… you could make trailers steerable I suppose. They already do that for the seriously over-length rigs hauling windmill blades and such but it’s hella complicated and expensive.

                My understanding is such is getting easier & cheaper, but isn’t something that lends itself to retrofit. So what is the normal service life for a trailer?Report

              • Road Scholar in reply to Oscar Gordon says:

                Boy… I don’t know exactly. It would depend on a lot of factors. My company generally tries to retire them at around 12 years but then they go on the used market. I would guess on average around 20 years or so.Report

              • Oscar Gordon in reply to Road Scholar says:

                Seems reasonable, and means that even if independant steering becomes standard, you are looking at 20+ years before it’s everywhere.Report

              • They already do that for the seriously over-length rigs hauling windmill blades and such…

                On one of my treks across the Great Plains a few years ago, I had to pass a series of trucks carrying windmill blades, on a day with 25-30 mph cross winds. Spooky.

                I saw what looked like a more interesting problem the other day when I was going over one of the big rail yards here in Denver and they had the same sort of blades mounted on train cars. I assume that the rail routes are scouted rather carefully for those.Report

              • One of the more amusing legislative hearings is always the three-way one between the farmers who want higher limits on the weight of grain trucks on rural highways, the county and school district governments who worry about putting school buses and such out on rural highways that will be beat to pieces by those heavier trucks, and the state DOT whom the others are trying to stick with the bill for the excess wear.

                Lately a fourth party has been added in some parts of the state: the oil/gas companies that want to run even heavier trucks on those roads.Report

    • Autolukos in reply to Michael Drew says:

      That explains the “Loomis is a fool” above.

      My two cents:

      1. Bad lane markings are not going to be a deal breaker.

      2. Given that some people currently work on mass transit or in a car while someone else drives, I have no doubt that some people will work in self-driving cars. To some extent, for some people, this may be compensated by reduced work in other settings (either shorter hours in the office or less work at home). In general, people navel gazing about changes from SDT need to spend more time looking at similar existing means of transportation.Report