Sentences about autonomous vehicles

Posted on by

Joshua Gans writes,

Consider, for a moment, the notion that human error could be completely eliminated by having autonomous vehicles. Would we need car airbags? Would we need seat belts? Would we need the car to be reinforced to withstand high impact collisions? Would we need crumple zones? I could go on. Would we need road barriers? Would we need steet lights? Now think about the costs of all of those things and you can see how they add up.

Today, we think that human driving is normal and autonomous vehicles are frightening. At some time in the future we will transition rapidly to the opposite point of view. Gans alludes to that in his conclusion.

16 thoughts on “Sentences about autonomous vehicles”

  1. “Would we need car airbags? Would we need seat belts? Would we need the car to be reinforced to withstand high impact collisions?”

    Yes we would. There are plenty of other unpredictable hazards that might cause crashes besides bad drivers. Deer and other animals. Rock-slides. Blown tires. Lost cargo, truck treads and other debris. Flash floods. Motorcycles (are they going to be autonomous too or banned?) Cyclists or pedestrians who suddenly dart out into the road (is the autonomous vehicle going to be programmed to mow down pedestrians or swerve and possibly crash into something else?) Even if 100% of the vehicle fleet was autonomous, the roads would remain a complex, unpredictable environment.

    Also, the transition to a hypothetical all-autonomous fleet couldn’t be quick. The average vehicle on the road in the U.S. is nearly 12 years old, and obviously half are older than that (and the average age is still growing). After the first fully autonomous vehicles hit the market, we’d have to expect at least a 20 year transition period.

    • +1

      Additionally, right now autonomous cars can drive in 90% of driving conditions but in terms of programming it is best to assume the 10% of programming is probably 50% of the work. So they can drive in good weather conditions but not snow yet. Also accidents will continue to happen as there will be lots of issues here.

      It will happen in ~15 – 20 years.

  2. The ideal case for robot drivers is a subway, or any “rail” system with negligible overlap with other ground traffic. Indeed, these have been around – usually in demonstration project mode – for a long time. Still, most systems employ human drivers, when even in an emergency, all a human driver can really do is call for help, which a robot could also easily do, even giving people “Uber updates” on where the help is and when it will arrive.

    It’s going to feel a little absurd if we get autonomous cars before autonomous trains.

  3. That’s only is true when the crash rate of autonomous vehicles is 0%. If all cars on the road are autonomous, then that should be possible to achieve. However, if some vehicles have drivers, then the crash rate might be some amount lower than it is now, but not necessarily 0% unless the technology significantly improves (so that the car can avoid a crash even when someone purposefully tries to crash in it).

  5. “we will transition rapidly”

    I hope so. I’m not getting any younger and the thought of getting old and being trapped in my house watching The Price Is Right everyday makes me ill.

    I wonder if anyone predicted such a rapid transition to automobiles. If you look through the photos on Shorpy.com, you’ll see the years where horses are the dominant mode of transportation. Then you see a very short period where there’s a mix of cars and horses. Then boom — no horses.

    • I wonder if anyone predicted such a rapid transition to automobiles.

      What do you mean by rapid. The first crude automobiles were failing in 1880s and 1896 was the first internal combustion vehicle. There were 8M automobiles registered in the US in 1920 and 26m in 1930 so it was around 1925ish when half the US households had an automobile.

      It was a ‘rapid’ change but most technology we way underestimate the time it takes to grow.

  6. We have no airbags on buses. Buses already have autonomous roadways for robotic control. We rarely have airbags in the back seat of a car, possibly sidebags. My suggestion is to put your van on autopilot to find the local beach. Then crawl in the back and have a drink while watching a movie.

    • Buses not having airbags or seat belts has more to do with their sheer mass than anything else, meaning less chance of sudden deceleration on impact. IOW, in any accident between a bus and another vehicle, it’s probably going to be the bus that ‘wins.’

  7. We will still have crumple zones, road barriers, just about everything we have now. Because people are scared. Look at how so many people react to GMOs.

    And people don’t want to “make things less safe.” Damn few politicians are willing to disagree with, “If it saves one life …”

  8. We’ll see how this goes, eh?
    https://medium.com/waymo/michigan-is-waymos-winter-wonderland-9b3cffbb9bab

    Also, re: speed of conversion. IIRC from my reading of “The Box,” by Marc Levinson (and I’m cheating by referring to the back cover), Malcom McLean’s first attempt to ship cargo via container happened in 1956. Because of the vast infrastructure supporting break-bulk shipping that had to change (new & converted ships, labor union buy-in, totally re-jiggered port facilities & tracking software systems, railroad boxcars, OTR trucking, etc.), it took decades to convert the industry to container shipping. (And you will still see a few traditional boxcars interspersed among the containers on a freight train, and the coal trains and liquid tanker cars are still old-style.)

    So, when thinking about autopiloted vehicles, let’s also give a thought to all the vehicular infrastructure that may be impacted by conversion to autopiloted vehicles of all sorts, and the tipping point at which each may be convinced that autopilot is here-to-stay, including trailers (boat, cargo, horse, etc.), RVs, utility & business, law enforcement, fire fighting (urban and wildland), etc., etc., etc.

  9. If autonomous drone transportation takes off, the question might become “Why do we need roads?” Considering the trillions spent on road building and maintenance, the impact would dwarf that of autonomous road vehicles. The implications for the forecasted robot job apocalypse would also be enormous, considering all the jobs associated with roadway transport.

    It seems to me the real reason we have not had flying cars previously was not because the technology was unavailable, but rather because the average human could not fly the vehicles (check out the requirements for a helicopter license). With newer technology, flying cars will be cheaper, but the key ingredient will be technology that allows the general public to fly them safely.

  11. I like the optimism displayed by autonomous car enthusiasts. Unfortunately, I think this optimism is unjustified given where we are in tech and software development.

    Let’s take voice recognition technlogy from companies like Dragon. Their medical product transcribes text much faster than I can type and with none of the typographic errors I commonly make (e.g., “teh” vs. “the”).

    The problem is that it fails in very very very bad ways. It has a habit of missing the “no” in “no tumor recurrence,” for example. While any human reading a “teh” can tell what it’s supposed to mean, the missing “no” is a huge deal. You still have to double check every word.

    We are still in early days with autonomous vehicles. If the relatively mature AI behind voice recognition is any guide, we won’t be getting rid of humans driving cars in any of our lifetimes.

    Plus, you have the social/human side to contend with: When a human makes a mistake and runs over a kid wearing dark clothes on a rainy night with poor visibility, we understand. We know what it’s like to be that driver. When an autonomous vehicle runs over a kid on a bright, sunny day because of a software glitch, we are less likely to understand. Add the morally bankrupt media and their need for controversy, and autonomous car deaths will be all over the news, while the much more common human-driver car deaths will be ignored.

    Finally, you have the issue of competence of the software industry. Look at all the security breaches because of poor coding (they always blame users for not keeping up with updates, never themselves for putting out bad code). Do you want these people designing your AI?

    Let’s look at Google in particular. As great as it is with search, their non-search work has been crap. We’re too enamoured of the company and its search prowess to notice, but look at how crappy chromecast functions compared to a box with a remote. Try to locate your android when their code forces you to verify your identity USING YOUR PHONE. These are simple, tractable problems that have Google beat. Why do we think they’ll manage a life-and-death situation like autonomous cars any differently?

Comments are closed.