A Futurist Take On Personal Transportation

Oscar Gordon

A Navy Turbine Tech who learned to spin wrenches on old cars, Oscar has since been trained as an Engineer & Software Developer & now writes tools for other engineers. When not in his shop or at work, he can be found spending time with his family, gardening, hiking, kayaking, gaming, or whatever strikes his fancy & fits in the budget.

Related Post Roulette

102 Responses

  1. Damon says:

    Oscar,

    So, everyone is saying that self driving cars are a thing of the near future. I’ve got a fundamental question. Really? Who says so. I’ve not polled anyone but, for my own personal life, I don’t see this as a “gotta have”. I drive a lot and I’d be one of the folks you’d think would want something like this, but frankly, I’m not seeing this as something critical to have.

    I’d be interested in folks commenting on the disruption this is going to have on the economy. Taxi fleets are politically protected. Fossil fuels companies have lots of leverage. Is it really going to be so simple to convert all the gas stations to battery swap stations, all the cars to electric, etc.? What about all those people in job jeopardy?

    Seems to me the “conversion” to auto cars will take a lot longer and be much more bumpy than a lot of people think.Report

    • Oscar Gordon in reply to Damon says:

      @damon

      Assuming we get the software aspect figured out to the point where cars have a better safety record than people, your insurance company will handle the incentive to switch.Report

      • Damon in reply to Oscar Gordon says:

        Interesting question. But “who” would be insured? Why would I, as an owner of a autonomous vehicle, need liability insurance, which is the only mandatory policy you currently needed? The car drives itself. I can carry comprehensive only and or maybe collision, but that should be unnecessary since autonomous vehicles won’t hit other cars.

        I’m not talking about in the short term, I’m talking when the pie in the sky “batter swap stations” exist and all the cars are EV and no one actually “drives”.Report

        • Oscar Gordon in reply to Damon says:

          Well, if you are driving the car yourself because it has no autodrive capability, then you have to be insured, and that will cost you money. If you own a self-driving car, I suspect Comp & Collision will be the default, and depending on the actuarial data, might not be very much.

          But the insurance question is a very interesting one. What do those policies look like? If the car is dispatched, I suspect the dispatcher carries the insurance. If it is POV, then the owner does. What liability does the manufacturer have?Report

          • Stillwater in reply to Oscar Gordon says:

            What liability does the manufacturer have?

            In the automated utopia Damon is referring to, manufacturer liability for “mechanical error” would broaden to include all operational errors (since those only result from mechanical error), yes?Report

          • Damon in reply to Oscar Gordon says:

            Truly, if I’m driving, I’d need liability insurance. Since the state has seen fit to require DRIVERS to have liability insurance, and if I’m NOT driving the car, regardless of ownership, I find it hard to believe that no one would need liability insurance. The state’s gonna want someone on the hook. I can see a future where a car is in a collision and the rider says “Not my fault, I wasn’t driving”. So who is? How could the rider/owner be liable for a collision if he wasn’t driving? Remember, liability insurance covers the driver.Report

            • Stillwater in reply to Damon says:

              Since the state has seen fit to require DRIVERS to have liability insurance

              Not where I live. Here in Colorado vehicles (not drivers) are insured. Certainly gummnt could continue with that policy wrt driverless vehicles, but actual liability would certainly be contested in the relevant situations, most likely between insurance companies and the manufacturer.Report

          • Kim in reply to Oscar Gordon says:

            Manufacturer probably has liability for software mistakes (“car given these inputs crashes into other car”), but probably doesn’t have liability for hardware issues unless they’re extreme (your uv sensor got dirty isn’t the manufacturer’s fault)Report

        • Richard Hershberger in reply to Damon says:

          Presumably whoever owns the vehicle will be obligated to carry liability insurance. If the safety record is better than human drivers, the premiums will be pretty low. Where it gets interesting is when a big-ticket crash occurs, that blows past the liability limits of the policy.

          Right now, if you are injured in a crash that other guy caused, and he has a minimum limits policy and no assets, there isn’t anything you can really do against him. My advice is to check your own policy’s uninsured/underinsured driver limits. This is something people don’t think about, and often lowball to save a few bucks. It isn’t all that expensive, and this is a risk well worth spreading around a bit.

          You know that old military joke about how amateurs discuss strategy while professionals discuss logistics? The personal injury version is that amateurs discuss who is at fault. Professionals discuss insurance policy limits. This is why PI attorneys get a warm fuzzy feeling when they guy who hit you was driving a commercial vehicle.

          How this applies to autonomous cars is the interesting legal question. If the guy that hit you can prove that his brakes were faulty, that is good. You no longer are limited to his insurance. You now have the manufacturers of the car and of the brakes. (At least if it is a big case: auto manufacturers lawyer up expensively and don’t roll over easily, so it isn’t worthy it for a small case.) Arguably, any accident caused by an autonomous car is the manufacturer’s fault, so when something goes awry and the car rolls through a bus stop the litigation will be a lot more complicated than if it were merely a drunk driver doing the same thing.Report

          • Morat20 in reply to Richard Hershberger says:

            Just anecdotally — someone I know was sued for a very, very, very minor fender bender. Somehow this ballooned into a 6 to 7 figure personal injury lawsuit (including six months of PT for what seemed a very unlikely injury for a car crash, but a very likely repetitive stress injury from the plaintiff’s day job), and included such fun as the plaintiff taking a year before filing suit (after all the PT and treatment), getting hit by a vehicle weeks before filing the suit (making an independent medical exam…interesting, due to tje new injuries), and in general it was a lovely mess of hilarity.

            The insurance company eventually settled for JUST under the Texas minimum liability limit. The minimum required liability being, of course, the plaintiff’s initial request weeks after the accident (via the lawyer hired within an hour of the accident). Coincidentally, I am sure.Report

          • Damon in reply to Richard Hershberger says:

            “Presumably whoever owns the vehicle will be obligated to carry liability insurance.”

            Yes, presumably, but I as the owner of a car do not need to have insurance on the car currently. If that were the case, all my kids would be insured to drive my car by the fact that I owned it. It’s the DRIVER that is insured. That’s why 16 year old boys have higher liability rates and 50 year old dudes who are married have lower rates. So, a change in the laws are necessary. How do you think that will be received by buyers of autonomous vehicles? I’d expect one big selling point, and one you mentioned, is that insurance should be lower as the human factor is removed.Report

      • Do you agree with me that no-fault is a necessity for insuring a driverless/autonomous vehicle? Failure of the product winds up being a products liability rather than driver negligence issue, so there’s no human being to point to as the careless one. Rather, there’s going to be Google for the software AI, some other Alphabet subordinate entity for the ultimate chassis assembly, Nakamoto Medium Industrial Concern for the sensor chip, the wire supplier from Mozambique, the connector supplier from Pakistan, etc. etc. all dealing with a strict liability issue about why the product, which is supposed to be totally safe, somehow wasn’t.

        The strict liability system of products liability doesn’t play nicely with the fault-based system of negligence. I agree that insurance companies will want no fault, but they’ve wanted no fault for a generation now and only got it in a few states.Report

        • Oscar Gordon in reply to Burt Likko says:

          It probably would fall largely to a product liability question, which means the vehicles, should they fault, will almost always fault to a no-go state, to minimize risk to passengers as well as manufacturers.Report

        • Will Truman in reply to Burt Likko says:

          So what do you see happening?Report

          • Mo in reply to Will Truman says:

            I see something akin to the vaccine liability pool. All manufacturers that meet certain levels of quality and passing testing pay into a fund and the fund pays out fixed damages based on the damage caused.Report

        • Troublesome Frog in reply to Burt Likko says:

          One interesting thing about assessing fault is that the cars will have ridiculous amounts of sensor data to review if the question of fault has to go to the courts. No need to rely on shaken up drivers’ first hand accounts. The current incarnation of the Google car should be able to give you a 3D replay of exactly what happened and account for exactly why it did the things it did. If we do have to decide whether a driver (human or not) acted appropriately, we’ll have more good data in our hands than we’ve ever had before.Report

          • True! We will be able to reconstruct the collision much better than we can now.

            We may have telemetry or the equivalent of a “black box,” too, which may make it easier to figure out which mechanical part (if any) failed.

            My prejudice is that, as @oscar-gordon suggests above, equipment failures should (at some point) trigger safeties that cause the vehicle to slow and stop. But bear in mind that we are talking by definition about equipment failures, and indeed we may well be talking about software failures and data corruption. My computer here at the office hiccoughed on me pretty good today, which left the IT dude saying many bad words and asking me what the hell I was doing to cause this. And then it turns out that no, I really was using the thing normally, there was just data corruption somewhere and things went all fubar.

            Stipulated that cars will have lots of failsafes and redundancies to minimize the effects of this happening. Still gonna happen. Equipment failures = strict liability; operator error = negligence. Not an equal playing field, and the ones on the short end of the liability stick are the ones with deeper pockets.

            So to answer @will-truman ‘s question, I see manufacturers and their various subcontractors and suppliers and programmers and the rest of them amalgamating their liability for indemnity pooling purposes and then buying some massive policies, if they can’t get the various states to switch to no fault regimes.Report

            • Kim in reply to Burt Likko says:

              YOUR computer is allowed to fail. Airport control computers get the good (slow) memory that alarms at the first instance of “this math isn’t working right anymore”. Ditto with the hard drives…

              If you want to bulletproof hardware, you can (note: if you aren’t the military, you probably don’t care, as bullets are a strange failure mode of a computer).Report

    • dragonfrog in reply to Damon says:

      Do you own a car? If so, does the annualized expense of keeping a car figure seriously in your household budget?

      If you could instead own only 1/8th of a car’s worth of a fleet of hundreds, at say 1/6th of the single-car expense, (and, presumably, have “your car” being in the shop a thing of the past) might it be worth it to you?

      One of the promises of self-driving automobiles is supposedly that auto co-ops could work better, as you wouldn’t have to walk to the nearest co-op car (which once in a while could end up being quite a long way), but instead have the nearest one drive itself to you. Also, insurance for auto co-ops can be prohibitively expensive to negotiate, as there are so many drivers of highly uneven skill involved. In a self-driving auto co-op, there would effectively be only one (perpetually wakeful and sober) driver – whatever driving software the fleet uses.Report

      • Damon in reply to dragonfrog says:

        Do you own a car? Yes
        If so, does the annualized expense of keeping a car figure seriously in your household budget? No, not currently. It might in 5-6 years. It’s fairly new and paid for. I spend more in insurance, on an annual basis, than I do in car maintenance.

        “-whatever driving software the fleet uses.” See that’s where I have a problem. You don’t think that manuf of auto cars are going to try and avoid any liability and stick it to the rider? Seems to me legislation is going to be needed.Report

        • dragonfrog in reply to Damon says:

          That’s what I meant with annualized cost – if you buy a new car ever 10 years, then annualized cost includes insurance, maintenance, and 1/10 of purchase price.

          Manufacturers probably will put the burden of purchasing insurance on the owner of the car. My point is that right now (in my area anyway) you go to an insurer and get insurance for the car, based on the insured driver. One adult driver who’s had a license for over a decade without accidents? $ You want to add a second adult who had a DUI suspension in the last five years? $$$ You want to add two 16 year olds? $$$$ You want to add every current and future member of a thousand member co-op? $ uncountable.

          But when the car will only ever be driven by the 2027 Honda driving algorithm, then it doesn’t matter how many people are allowed to request that the car go to a particular destination, there is only one driver, and that driver’s competence is knowable to a much higher degree than any individual human’s, because by 2029 you will have a full year’s safety record of Honda’s entire 2027 fleet to populate your actuarial tables.

          As the owner, you still probably have to buy insurance, but your insurance company knows the single driver you’re insuring is never going to get drunk, drag race, or text its friends.Report

          • Damon in reply to dragonfrog says:

            I guess it comes down to “how’s ultimately “in charge” of a driver-less car?” The manufacturer, the owner, the rider, all of them in part, none of them?Report

            • Chip Daniels in reply to Damon says:

              Which is essentially a political question of how we define property and contract.

              Whose interests will be defended, where will the lines be drawn?Report

              • Damon in reply to Chip Daniels says:

                Well, even if I own a auto car but I’m not driving, I see no reason why I would need liability insurance. I am NOT the driver and I’m not responsible for it’s actions and have no control over it’s operation. If that’s not the case, no reason to have one.Report

            • Lyle in reply to Damon says:

              Actually there is a nice precedent here in General Aviation. Modern planes have lots of automation, and can essentially fly themselves, all though some of the automation does not make it into GA planes. However in Aviation the Pilot in Command is ultimately responsible. So even if the autopilot flies the plane into a house with no human intervention, it would still be a liability claim against the Pilots estate. (assuming the pilot was killed in the crash).
              The nice thing here is the courts have worked out a lot of the details over of what various things mean if you take the precedent.
              I suspect you may have 2 prices for such a car a low one where you buy insurance, and a high one where the insurance is bundled in the car price.Report

          • Brandon Berg in reply to dragonfrog says:

            Manufacturers probably will put the burden of purchasing insurance on the owner of the car.

            I’m not so sure about this. If consumers underestimate how safe the cars are, then manufacturers might want to provide insurance as a selling point. Not only do consumers then not have to buy insurance, but it’s a strong signal of the manufacturer’s confidence in the cars’ safety. It’s also more efficient, since the insurers can sell policies to manufacturers in bulk, rather than selling to each customer individually. The consumer pays either way, of course, but bundling seems more efficient, and there’s no particular reason to expect it not to happen.

            That said, a system of strict liability for the manufacturers, conditional on proper maintenance, seems fine to me, as well. This would, of course, be priced into the cars.Report

  2. Darth Binks says:

    Problems. Security. Privacy.

    a. Smart cars are already hackable, and its only getting worse.

    https://duckduckgo.com/?q=car+vehicle+automobile+hack+site%3Awww.schneier.com%2Fblog&kh=1

    b. Autonomous cars can be used as weapons.

    http://www.start.umd.edu/news/smart-bomb-every-garage-driverless-cars-and-future-terrorist-attacks

    c.Report

    • Oscar Gordon in reply to Darth Binks says:

      I addressed the security question at the end. Software & network security is not my thing, so I can’t expand too much on that.

      As for the terrorist aspect of it, that is part of why I say the police will insist on an override function. Still, it’s a thing to prepare for, not a deal breaker. A car can be made RC right now, as many episodes of Mythbusters have shown us.Report

      • Glyph in reply to Oscar Gordon says:

        “The police will insist on an override function”.

        And like LE insisting on legally-mandated back doors to your computer and encryption schemes – that is, your devices and software, that you own, and use to conduct your business and life, and count upon to protect yourself – this power will never be abused or fall into the wrong hands.

        If only the police had the legally-designated power to remotely-disable other tools you might have; why, the Corey Maye incident might have had a “happy” ending for them, with Maye dead instead.

        I hope the ACLU and EFF are all over this, and you are wrong.Report

        • Oscar Gordon in reply to Glyph says:

          @glyph

          I’m not a fan of the idea of a Law Enforcement Kill Switch, but I expect LE will lobby hard for such a thing unless we have a sea change of the attitudes of LE in the US.Report

        • Chip Daniels in reply to Glyph says:

          This is the point I find interesting.
          Not just a kill switch, but the changing concept of ownership in the Internet Of Things.

          Increasingly, you no longer actually own, fee simple, the things we use and rely upon. Everything is a form of subscription lease, where you gain low cost and service in exchange for surrendering control.

          Marrying this concept to the concept of universal interoperability, that is, where all your devices need to have a common platform and common language in order to be efficient which results in an effective oligarchy of providers;

          Then add the increasingly strong definition of property rights, strengthening the hand of the makers and service providers in the EULAs that everything you “own” will come with;

          Finally, lay on top of this the fear and paranoia of crime and terror that makes government collusion with the makers a priority.

          These things don’t guarantee a dystopian outcome, but do make it more likely, depending on how we collectively decide to respond to them.Report

      • Lyle in reply to Oscar Gordon says:

        I see two buttons 1 pull over to the side of the road safely and stop. 2 emit alarm to nearby vehicles and stop right now. These would be the override buttons of a self driving car. Since remote stop is on all GM cars now its not to hard to extend it (ON Star can cause your car to shut down remotely)Report

  3. Richard Hershberger says:

    And what does every road and highway have painted all along it’s length? Why, highly visible lines!

    This is laughably far from being true. Even apart from 25 mph residential streets, there are many parts of the country where rural roads lack this amenity. Some fifteen or twenty years ago I lived here,
    https://www.google.com/maps/@40.6829903,-79.1938982,3a,75y,74.39h,72.33t/data=!3m6!1e1!3m4!1shfpmk7mQiU-rPjdaSnz9dQ!2e0!7i3328!8i1664, renting the apartment above the garage in the building on the right. Note the road, with its absence of lines either along the side or down the middle. This is not a mere alley (though a fully autonomous car should be able to handle those, too). It is a state highway, with cars going around 50 mph or so, that runs from one place to another place, albeit neither place being heavily populated. I’m not suggesting that this is a deal killer, but it reinforces the “able to handle anything Mountain View can throw at it” skepticism.

    More broadly, much of this piece describes a mature system, with RFID chips embedded in the roadways. When is that road I lived on going to get those chips? The people there are lucky it has gotten a pavement repair since I lived there. The picture of where thinking is headed is interesting, but we need a pathway from here to there.

    From footnote 1, regarding inclement weather:

    Cars lose control on the road because they lose traction.

    I don’t question the losing traction part. It is the sensory degradation I wonder about:

    you could place a small sensor node at each corner of a car and have 360 degree coverage of the surrounding space

    Yabbut, what happens when each corner of the car is covered with road crap? Road crap is found in amazing quantities when in salt-laden snow-and-slush mode. Perhaps the miracle of chemistry can repel the road crap such that it falls off harmlessly, but perhaps not. I have yet to see a description of the testing of this question.

    It seems to me that the security issue tends to get mentioned then hand-waved away. This has vast potential for political fallout. Just imagine the first time someone hacks a car and drives it into a group, then some terrorist group takes credit. The history of tech and hacking does not suggest that this is an issue that industry will give top priority or will hold back releasing product before it is fully worked out.

    (get on the highway, turn it on, and relax until it’s time to leave the highway – truck drivers will love this!)

    How close are we to this, as contrasted with get on the highway, turn it on, and struggle to maintain full situational awareness while prepared to take over on a second’s notice? Because we all know damned well that people are going to treat it as the former, regardless. No system is ready for prime time if it depends on pretending that people will treat it the latter way.

    My prediction is that autonomous freeway driving will be here sooner rather than later: get on the freeway, turn on the system, and watch a video on your tablet. When the car approaches the off-ramp it will give increasingly annoying notices for the human to take back control. This, however, is the low-hanging fruit.Report

    • Oscar Gordon in reply to Richard Hershberger says:

      @richard-hershberger

      I did offer some measure of a progression. I did describe what I expect a mature system will look like, but I don’t pretend that it won’t require time or be a serious of advances. It’ll start with major freeways, tollways, & interstates, then move to lesser roadways. I expect cars will have a manual over-ride for a long time, so drivers can navigate roads the car can’t. So I’m not sure why you are being critical of this?

      Still, as the sensors improve, and the software improves, the car will be able to “see” the road as well as, or even better than we will, so lane markers will be most important on multilane, high speed roadways.

      Sensor degradation is important, but as I explained earlier, I expect hydrophobic coatings and a minor amount of heating will handle most of it (enough heating to cause a layer of water to form so the building up slides off before it causes a problem), as well as sensor redundancy and warning systems, so the driver/owner knows when the sensors need attention or a fresh coating of NeverWet. Also, as you get more and more cars on the road that broadcast their presence on radio (and this could be a required aftermarket thing older cars have to mount at some point), sensor degradation become less of an issue, since the cars will be talking to each other all the time.

      We start at Mountain View and get the basics figured out, then you start working on the rest. Think about autopilot for airplanes. The first autopilots just kept the plane level and on a constant bearing. Then they started being able to make course corrections, then course & altitude changes. Now they can pretty much land the aircraft at airfields with Landing Assist systems.

      I don’t mean to hand-wave away the security aspect of, it’s just not my area, so I can’t comment on it properly. I expect it will be an issue and dispatchable vehicles will not be a thing until it is. Auto driving still could, if the control system was completely disconnected from any wireless system that can accept commands.Report

      • dragonfrog in reply to Oscar Gordon says:

        It seems to me that the security issue tends to get mentioned then hand-waved away.

        This.

        I thought that SciAm article was pretty silly. “Pshaw! It took highly skilled researchers a year of work to hack that car!” – as if it would take them another year of work to hack another one of the same make and model. It didn’t take them a year of work to hack that car, it took them a year of work to hack that model of car – hacking the next car of that model would take about a second of work. And the research they did on the way to hacking 2015 Jeep SUVs laid considerable groundwork toward the next car hacking work they do.

        So far, the published car hacking research has come direct from academia. The same is true of the early research on chip-and-pin debit card security. But now that chip-and-pin has been around for two decades, a lot of the published research is actually based on forensic work describing the MO’s of criminal rings that get busted – and new vulnerabilities are being discovered that way, only after the police bust a gang and their in-house research gets analyzed and published.

        The same logic as the SciAm article used was probably applied in the 90’s to dismiss concerns about chip-and-pin security, and it’s just as wrong-headed now as it was then.Report

        • Oscar Gordon in reply to dragonfrog says:

          The security issue is the other big thing (aside from the control software being smart enough to handle humans piloting other cars, and piloting them badly).

          I am not entirely confident that automakers will give it proper attention (until after enough people meet bad ends because of it). The article I linked to showed the reason why, they had linked the wireless system of the car to the control systems, which had me asking, why in the bloody hell were those two systems linked such that there was any kind of path hackers could take to get at the control system.

          I mean, like all things, there will never be perfect security. If a person can gain physical access to the car, they will be able to subvert it to their needs, given enough time & know-how. I’d just be happy if they could make a remote hack be difficult enough that any potential gain would be so little as to not make it worth the effort. There is considerable reward in being able to crack credit card systems, and very little risk. What is the payoff versus risk in cracking a remote car?Report

          • dragonfrog in reply to Oscar Gordon says:

            If you can steal a car and have it drive itself to a chop shop, without showing up on a security camera at the site of the theft (or, potentially, show up as someone standing on the sidewalk fiddling with their phone), then later go to the chop shop to collect payment, but never appear along with a stolen car – that seems like a reduction of the thief’s risk.

            In a crime-ridden place, where kidnappings are frequent, car hacking might be a viable kidnapping method – granted that’s getting a bit movie-plot.Report

            • Oscar Gordon in reply to dragonfrog says:

              I’m much more concerned with hacks that involve cars with passengers, or using a hacked car as a weapon as per Darth Binks up above. Hacks that just steal cars are a possibility, but aren’t a major public safety risk.Report

              • dragonfrog in reply to Oscar Gordon says:

                I suspect remote car crashes intended to cause bodily harm would be a small minority of such hacks, at least in first world countries with low homicide rates. I’d expect it might be as common as the use of car bombings as an assassination method already is in your part of the world.

                The biggest potential for bodily harm might be unintentional thefts of cars with people in them – what do the people working the chop shop do when a car rolls up with a terrified passenger in it?

                I mean there’s all kinds of far fetched movie-plot threats. I just thought of this one – in a high speed chase (where the people being pursued by police are in a human-driven car, since self-driving ones refuse to speed), the baddies keep a scanner running for hackable cars they might pass, and when they get one, they use the occupied vehicle as a weapon to ram the police off the road.Report

              • Oscar Gordon in reply to dragonfrog says:

                That assumes a vehicle can be hacked that quickly. Hollywood plot devices aside, most hacking takes time. It can only go quickly if it’s a zero-day exploit that has been automated, so while that might be possible, it would only be possible once using that particular attack vectorReport

              • Morat20 in reply to Oscar Gordon says:

                Also, you can generally get rid of a number of hack possibilities by simply hard-wiring in a big, red button. “Make this car stop RIGHT NOW”.

                Short of breaking into your car, digging into the cuts, cutting cords and replacing hardware components, that will…stop your car. And probably automatically send up a big flag for cops to come by and check out. After all, with auto-driving cars they’re not going to be issuing tickets. Now there’s a wrinkle…half the PDs in the US would go bankrupt.

                Also, I still don’t get the whole “automate trucks and have them convoy on autopilot”. We have those. They’re called “trains” and they’ll get you like 95% of the way to anywhere. Odds are you won’t be able to convoy past there anyways.Report

              • Oscar Gordon in reply to Morat20 says:

                I.E. The cabin has a main breaker panel.Report

              • Morat20 in reply to Oscar Gordon says:

                In terms of customer usage, it’d probably be on the dash under a clear shield. The usual “flip up shield, depress button”. You want it somewhere that it can’t be hit on accident, but pretty easy to do on purpose.

                (I say that as someone whose kid has, somehow disabled the smart key system AND altered my steering column with just his clumsy knees.Took me a bit of googling to get the smart key system turned back on. Small, easy to miss button).

                Maybe a fairly stiff button that needs a good, hard, push.Report

              • dragonfrog in reply to Oscar Gordon says:

                Most compromises of computers nowadays happen because someone didn’t update their Java or Flash player, and stumbled on a malicious website that has a fully-scripted exploit of a vulnerability from half a year ago. Some hacker spent time putting together the exploit kit, but the actual execution of the exploit takes a second or two from the victim clicking a link to malware payload setting itself up to survive reboot.

                I’m thinking about the same sort of scenario – a bluetooth scanner in the getaway car with a library of outdated car OS’s, with canned detection signatures and exploit routines. As they drive into bluetooth range of a 2021 Mazda, the scanner detects this, and fires off exploits for vulnerabilities that were patched in 2025 and 2026. Turns out the owner didn’t take their car in for the 2026 software patch, and suddenly the Mazda has pulled perpendicular across the road in front of the oncoming police cars.Report

              • Kim in reply to dragonfrog says:

                Flash is dead, if you haven’t uninstalled it, you really ought to.
                (critical security vulnerabilities that can’t be patched).Report

              • Oscar Gordon in reply to dragonfrog says:

                Possible, if autocars were largely POVs, that would potentially be a significant issue. Dispatched/Fleet vehicles would likely be patched regularly. So it would depend on how many POVs remain on the road in 2026.Report

              • dragonfrog in reply to Oscar Gordon says:

                Would they though?

                What’s the utility companies’ record for patching smart meters like? I know ISPs are often poor-to-abysmal at patching customers’ routers (just google ‘home router botnet’)Report

              • Oscar Gordon in reply to dragonfrog says:

                Smart Meters involve a person going out and updating the meter at the house (IIRC, or are these meters remotely manageable?), and are ISPs patching routers they own and can remotely manage?

                It’s one thing to broadcast a patch to things you have root on, it’s something else if you don’t own it, or if you physically have to go to each distant location to enact a patch.

                Dispatched cars would return home regularly and could easily be updated.

                Also, failing to patch would potentially shift liability from manufacturer to dispatch service. If a missed patch caused a bad end, the lawsuit and/or fines would do a good job of reminding other services to keep up on patches.Report

              • dragonfrog in reply to Oscar Gordon says:

                In many cases yes, the routers they are failing to patch are rented not sold to customers, as a selling feature of their service – get our DSL service, our modem has a WiFi router built right in. My previous ISP had one like that – some of the pages in the configuration interface were inaccessible to me as the customer. Presumably there was an ISP password that would have made them available.Report

              • Oscar Gordon in reply to dragonfrog says:

                OK, I know what you are talking about now. I always treated the ISP router as simply a modem and put a router I controlled between it and all my devices. I’d forgotten that ISPs actually handed out routers.Report

              • Lyle in reply to dragonfrog says:

                Agreed the rental car companies are bad at doing recalls today, so much that there has been discussion of making it illegal to rent a car that has not had recalls fixed. Some companies do make that a requirement such as Enterprise (I heard them talk about it)Report

              • Kim in reply to dragonfrog says:

                Depends on who’s using the CPUs and for what.
                It doesn’t really cost a home user much if someone else steals CPU time — they don’t really notice, so there’s little incentive for good security.

                If the idiot hacker decides to steal the car? Well, that’s a different story.Report

    • Kim in reply to Richard Hershberger says:

      Richard,
      People have already hacked the googlemobile. It got a software upgrade.
      Please do not underestimate google’s ability to pay attention to hackers.

      The last time someone sent a plane into a building in downtown Austin, nobody did a goddamn thing about General Aviation…

      People accept that there is risk with cars, dunno why, but they do.

      Expect that Truckers get autonomous driving trucks first, and that they stick to the main routes. Insurance, insurance, insurance.Report

  4. Brandon Berg says:

    Richard Hershberger: The history of tech and hacking does not suggest that this is an issue that industry will give top priority or will hold back releasing product before it is fully worked out.

    Should they? In the US, human drivers kill a bit over one person for every hundred million miles driven, for a total death toll of over 30,000 per year. It seems to me that the question we should be asking is not whether self-driving cars are perfectly safe, but whether they’re safer than this. It’s not clear to me why computers should be held to a dramatically higher standard than the highly imperfect humans they would be replacing.Report

    • Richard Hershberger in reply to Brandon Berg says:

      Should they?

      This is the wrong question. Political realities are wonderfully immune from rational risk analysis. Currently, automotive fatalities are background noise that people accept as part of normal existence. We as a society did not respond to the demonstration that an airliner can be hijacked and flown into a building by observing how the number of deaths resulting from this was tiny in comparison with overall traffic fatalities.

      If this discussion is predicated on the notion that said political realities will somehow not operate here, then this discussion is hopelessly flawed.Report

      • Troublesome Frog in reply to Richard Hershberger says:

        True, we should never underestimate the human ability to be bad at estimating things. If we switched to autonomous cars tomorrow, went 3 years without a single fatality, and then had one fatal crash, the discussions all over the news would be and asking if autonomous cars are just too darned dangerous.Report

        • Oscar Gordon in reply to Troublesome Frog says:

          For example, see Anti-VaxxersReport

        • In the first documented crash since automated cars became mandatory 3 years ago, we now can release the name of the victim of the accident: Joe Schmoe was a beloved husband and a father of 3 children. Married for 16 years, his wife now has to raise the children alone. Ironically, Joe Schmoe survived being hit by a drunk driver back in the twenty-teens.

          A life tragically cut short by automated driving.Report

      • Kim in reply to Richard Hershberger says:

        No, but we goddamn did when someone from General Aviation flew a plane into downtown Austin. I think we didn’t change a damn fucking thing.Report

  5. Richard Hershberger says:

    Road crap is not necessarily liquid, even if it when through a liquid phase. In the snow version of road crap it is water saturated with sodium chloride or some other chemical that lowers the freezing point. It is then kicked up by cars into spray form, which in turn is evaporating rapidly. It might hit your windshield as salty water, which then dries, but it also might hit your windshield as salt particles that rapidly form a beautiful, opaque coating on a windshield. The dry form is actually worse. Salt water is pretty transparent, so you can just wipe it off. The dry form requires the application of water. Will a hydrophobic coating help with the dry form? Maybe, but this isn’t obvious. I spent my teen years in the Mojave Desert. There the equivalent is a layer of fine dust coating everything from your windshield to the inside of your underwear. Oh, and “a minor amount of heating” suggests unfamiliarity with Minnesota winters.

    A standard narrative is how you ride to work in an autonomous car, which you might or might not own. It drops you off at your office and goes off somewhere else, then it (if you own it) or another one like it (if you don’t) appears at your office at five o’clock to take you home. Where does it spend the intervening hours? Is it parked under cover, or might it have a foot of snow on it when it is time to pick you up? Yes, there are solutions to this problem. I can envision commercial parking garages located in some outlying spot not actually connected to anything, serving autonomous cars in their down time. But I notice how no one talks about new and exciting parking garage opportunities resulting from this technology.Report

    • Oscar Gordon in reply to Richard Hershberger says:

      Remember, I grew up in WI, lived there until 2006 (excepting my time in the Navy). I am intimately familiar with winter road crap (ice/snow/slush/salt/sand), and the joys of cleaning it off my car. I am also an engineer with a strong background in fluid dynamics & heat transfer, so when I tell you that a slightly warm hydrophobic coating will cause winter road crap to slide off, it will, and cleanly, and I won’t even need a lot of heat to do it, just enough to cause the layer immediately against the surface to change phase to water*. The concern will be keeping the coating in good shape, which means owners will have to keep a can of hydrophobic coating handy during the winter, much like you keep lock de-icer handy (or how I kept starting fluid handy for my old carbureted engines). And again, sensor redundancy can help. A sensor in every corner at bumper level, and another set at the roof line gives considerable overlap, so should a sensor become obscured, the other sensors can cover until the obstruction can be cleared.

      Now, dry, dusty environments is a different question altogether. The primary cause of dust sticking is (IIRC) static electricity. How that can be mitigated is a good question. It could involve the car applying a counter charge to the sensor surface to keep it clear, or perhaps a quick water wash much like headlight wipers, etc. It’s a minor engineering issue that, if it’s important for the safe operation of the vehicle, will be tackled.

      As for what happens to an idle car during a blizzard, let’s get the cars driving themselves first, not every potential problem has to be tackled at the start.

      *Current theory holds that a thin, molecular layer of liquid water is what actually causes most winter ice to be slippery. A minor amount of heat will increase that layer such that it will be unable to adhere to a hydrophobic surface.Report

      • I note that dust — in the form of tiny bits of dirt, tiny bits of animal and plant material, tiny bits of ground-up tire, and perhaps as fine as particulate material in the air — is ubiquitous. But it’s not equal from place to place. Living in the desert as I do, dust and sand in the air is a not-uncommon occurrence.

        Grounding or shielding the area outside the sensor from electricity now strikes me as a very significant problem. Or maybe I just need to learn to wipe down my autonomous vehicle before riding it, if the sensors can work with a minimal layer of the stuff encrusted on them. Or perhaps a robot does that for me.Report

        • Oscar Gordon in reply to Burt Likko says:

          I’ve only ever driven through a desert, never lived in one. However, as you may recall, my Navy service was spent as a technician on Hovercraft, which means lots of blowing sea spray & sand/dust. The vehicle design was far more concerned with keeping the spray/dust out of the turbine intakes than it was keeping the sensor suite clean. In other words, the thin layer of dust or salt film we’d get never degraded sensor performance enough to make it a serious concern.

          I expect that the problem will be addressed with either a material coating, or regular application of electric charge to repel particles, or just a regular requirement to wash the damn car.Report

        • I live in high semi-arid country. If I leave my car in the garage for three days, door closed, when I back out there’s a noticeable film of dust on the windshield. If it’s pollen season for the various conifers, the film is more than noticeable. Seriously, try sitting on one side of a small valley watching sheets of conifer pollen blow along every time the breeze picks up. In some species, the pollen is slightly tacky; enough to overcome any static charge you’d be willing to have build up.

          Not an insurmountable problem, just one that has to be considered.Report

          • Kim in reply to Michael Cain says:

            pollen is indeed bad news.Report

          • Oscar Gordon in reply to Michael Cain says:

            Of all the challenges in front of autonomous cars, keeping the sensor node clean is on the list, but not exactly near the top. Once someone sits down and actually designs the sensor suite for integration (instead of being essentially bolted onto the roof), this issue will be most likely be addressed at that time.

            I think people not understanding the engineering process is a big source of criticism for the idea. People see the obvious flaws and assume no one is thinking about them, when the reality is that dealing with those flaws just isn’t worth the time or effort until the keystone challenges are overcome.

            If you can’t get the car intelligence to handle human drivers, who gives a crap about if the sensor node can stay clean?Report

            • DensityDuck in reply to Oscar Gordon says:

              There’s also the assumption that the flaws must be fixed at the level of the failing item.

              Cars were known to have trouble driving at high speed on unpaved roads. The response was not to design cars with high-speed off-road capability; the response was to make more paved roads.

              Maybe the answer to “sensors can’t see people” is “make it a crime to walk on the highway”.Report

              • Oscar Gordon in reply to DensityDuck says:

                Suddenly Jaywalking gets a whole lot more serious.Report

              • dragonfrog in reply to Oscar Gordon says:

                That’s basically where the whole concept of jaywalking came from -blame shifting campaign on the part of the auto industry. Previously walking on the road was what everyone did, and motorists were expected to drive in a way compatible with that view of what streets were. The idea of jaywalking as opposed to just walking was introduced as a way of coopting the definition of what a street was and was forReport

              • Chris in reply to dragonfrog says:

                The 1905 video of San Francisco’s market street from the front of a cable car is a revelation in this regard, because it shows a world in which the whole system, to the extent that there is one, favors pedestrians and the cable cars (public transportation). Most importantly, everything is slow enough to be safe for everyone.

                https://youtu.be/pChbWHdbkvQReport

  6. KatherineMW says:

    If this does happen, it’s worth considering the economic and labour impacts, which go far beyond the end of taxis. Driving transport trucks is the second-most common occupation among men in Canada, and one of the higher-paying ones for men without postsecondary education (although the hours are brutal). Auto repair is the 7th-most-common. Bus/transit drivers are another large group. All of these are substantially larger than the number of taxi drivers (there are around 50,000 taxi drivers, 85,000 transit/bus drivers, and 260,000 transport truck drivers).

    On second thought, though, transport truck drivers might be likely to stay around even in a world of self-driving cars, because to some extent they’re security as well as drivers. If self-driving trucks were carrying anything of relatively high value, robbers would consider them a windfall.Report

    • Kim in reply to KatherineMW says:

      Security gets paid a lot less than “high risk” driving.Report

    • North in reply to KatherineMW says:

      Yeah I think that needs expanding. Pure transportation service jobs would take it in the gut; agreed, but delivery and goods transportation jobs would potentially -increase- not decrease because you’d still need someone to keep an eye on the goods and walk them up to the delivery site. A self driving car isn’t going to figure out which mailbox to leave your parcel at and it’s not going to keep an eye on your tractor trailer full of product. Also the entire delivery sector is growing enormously and can be expected to continue to expand.

      Auto repair potentially also expands. If cars become things that are running nearly all the time then they’d need enormouosly more maintenance and repair.Report

      • Kim in reply to North says:

        “A self driving car isn’t going to figure out which mailbox to leave your parcel”
        If a drone will figure it out, why not a car?
        (Note: google, stop flying your drones into our potholes and getting them infinitely lost in there – that’s probably a bug).Report

    • Oscar Gordon in reply to KatherineMW says:

      I doubt we’ll be in danger of losing commercial truck or bus drivers anytime soon, because they do more than just drive & provide some measure of security. Hell, trains are automated for the most part and even they still have a person at the wheel. The more damage a vehicle can do should it lose control, the more likely it’ll have a human at hand to take over. Keeping said humans awake and alert will be a trick, so I imagine such positions will be two person teams for long haul/long shift positions.Report

  7. North says:

    My question would be: if you’re powering these things electrically what the heck do you do to heat them? Combustion engines produce gobs of heat as a waste product so heat is easy. Electrical engines don’t (or not in any serious amount) and electrical heating chomps through batteries like crazy. It may not matter to Arizona or California but not every state has balmy temperatures in February.Report

    • Oscar Gordon in reply to North says:

      @north

      Look at the Chevy Volt

      Electronics do produce a lot of waste heat (my Onkyo stereo receiver is evidence of that!). All electric cars would just be more serious about capturing and retaining that heat for passenger comfort in cold climates (better insulation, etc.).Report

      • North in reply to Oscar Gordon says:

        I sit corrected, thanks.Report

        • Oscar Gordon in reply to North says:

          On a heated seat, I hope.

          (I do love that about my Subaru, heated seats in cold places is one of the best ideas to come out of the auto industry in a long time).Report

          • North in reply to Oscar Gordon says:

            No alas, my lil ford fushion doesn’t have heated seats. I don’t mind much being Canadian and cold inclined. MN is pretty balmy and pleasant by my standards (it’s so delisciously dry!!)Report

          • Michael Drew in reply to Oscar Gordon says:

            +1.

            We drive my girlfriend’s parents 2001 Avalon. It’s falling apart – we breathe exhaust & no interior lights work (dash, yes, but no dome or visor-area lights). Makes it tough to retrieve your stuff after 5 pm in MN in the winter.

            But the seats still heat, so I’m good. (Also, it still locomotes, so that’s a plus too.)Report

          • Chris in reply to Oscar Gordon says:

            Average number of days per year on which I would need heated seats: 0.2.

            I exaggerate. The last car I owned had heated seats. My son was in elementary school at the time, and on cold mornings I’d drive up the block to the bus stop and we’d sit in the car, with several of his friends, waiting for the bus. We’d turn on the heated seats and warm up nicely while waiting, and the kids (all in K-2nd grade) loved them. However, by the time I drove to campus (if I drove), it’d be warm enough outside that heated seats were unnecessary.Report

    • Kim in reply to North says:

      My husband’s computer heats his home-office to an alarming degree in the summer and winter. We’ve built the entire heating/cooling system with it in mind.Report

    • Michael Cain in reply to North says:

      Same answer as cooling for summer in Arizona: high-efficiency heat pumps, limited outside air mix, and better use of the resource (ie, heat/cool the seats).Report

  8. Oscar Gordon says:

    Remember what I was saying about tractors?Report

  9. Oscar Gordon says:

    Even if we never get to driverless cars, we will certainly get to much, much safer cars.Report