August 02, 2015
Self-Driving Cars To Cause Car Insurance Industry Revenue Plunge
In 15 years average premiums could be less than half of today's car insurance costs.
That is even more dramatic than it sounds because 15 years from now some of the cars on the road will be pre-autonomous. Just how far autonomous vehicles cut accident rates will depend on how many people insist up controlling their cars at least some of the time. It seems likely that at some point in the future pedestrians, bicyclists, and motorcyclists will be much bigger causes of accidents than cars.
Insurance rates on autonomous vehicles will be set by such factors as how many non-autonomous vehicles operate in the area where each person lives and how many pedestrians, bicyclists, motorcyclists are on the roads the driver travels. Another factor: different companies will have different accident rates due to variations in the quality of their hardware and software. Car companies with the lowest accident rates will advertise the lower accident rates and the resulting lower insurance rates.
Since autonomous driving equipment will come to upscale cars first the safety advantage of premium cars will be much greater than it is today.
Autonomous vehicle introduction could cause a large winnowing in the auto industry. Autonomous capability will entice a lot of people to upgrade to a new car once it becomes clear that certain models from certain brands are doing a great job avoiding accidents.
Right now the autonomous highway driving option for a Tesla costs $2500. In just one year Cadillac will offer their Super Cruise autonomous highway driving feature in a new Cadillac model. Autonomous driving on surface roads in towns with pedestrians and parked cars is a much harder problem
Various car companies are predicting fully autonomous vehicles on the market between 2018 and 2024 with with 2044 for mandatory autonomous operation. In the latter forecast autonomous operation without driver back-up doesn't come to market until 2030. We do not know the exact timing but full autonomous operation will come. In that future I predict adventurous groups of peoplewill go on organized trips to countries that still allow autonomous vehicle operation. Then the adventurous groups will go on driving marathons across a country's rural roads to meet up at night and tell stories about accidents and surprises.
Randall Parker, 2015 August 02 12:41 PM
A potential problem that I see with autonomous cars, which no one has really touched on, is ability for someone to hack them and perform actions that the owner doesn't want performed. There's the obvious problem of the hacker purposely causing the car to crash, but the hacker could also do something like lock the doors and then drive the car to a predetermined location, where the hacker or their buddies then kidnap the occupant. Or hack a bunch of cars and cause them to all drive to a certain location, in order to purposely cause a traffic jam, for whatever nefarious purpose. With the recent Dodge/Chrysler/Fiat hack where it was demonstrated that it was possible to gain control of the car from the driver (to a certain extent), it's only going to be worse with a more autonomous car.
There's no technical reason that an autonomous car has to be easily hackable. Humans manage to navigate the roads based on sensory data, an autonomous car could, too, with the advantage of being 'familiar' with the entire road system. You don't even need a reliable gps, if the car has the map built into it.
The real problem is that, up until now, the automotive industry hasn't taken seriously the threat of hacking. They've done remarkably stupid things like putting critical systems on the same buss as the entertainment system, when the two systems should be air gapped, or at least have a one-way connection.
And industry in general tends to install obvious security gaps for it's own convenience, like retaining the ability to "push" software updates.
I'm kind of hoping industry is going to be taking hacking seriously enough to stop doing those sorts of stupid things, before self-driving cars catch on. Maybe I'm just too optimistic about that.
I hope automation and safety mandates for cars has the side effect where it eventually enables "platooning" which could reduce congestion in our cities. Traffic congestion is a major problem in the US that costs a lot of money.
How will we deal with uninsured and underinsured motorists in the future? They should pay a $2/gallon surcharge on gas unless they have a policy that can cover $1million liability.
The big problem with the software in current vehicles is that the entertainment software which must have internet access, is not separated from the vehicle control software which should not have internet access. Future vehicles could completely separate these functions with two different computer systems. That would 'airgap' the control software from access via the internet. It would make hacking the system much more difficult and it could be hardened further with other well known security features.
"Autonomous cars..." Yes, and brick buildings, sprinklers, fire departments all cut homeowners' rates.
That is a social good.
Meanwhile, there are new methods to combine physical and financial risks so that they are appropriate for actuarial methods. We will always have insurers as well as hedge counter-parties.
As long as there are risks, there will always be market mechanisms to manage (or at least price) them for the timid.
The control system does need real-time traffic data, which is standard feature on dedicated and phone based navigation systems, and that means wireless communications. But there is no reason this info cannot be sandboxed and message between it and the control system severely constrained. Also, if the traffic data is done through a 1-way system, similar to GPS itself, it will be difficult to exploit unless you want to affect every vehicle in a region. This mostly happens over FM to dedicated devices now.
A perfectly unhackable system with a good feature is almost certainly impossible. The task is to make the opportunity cost too high to be of value.
Require every vehicle to have an EMergency Off (EMO) push-button that is hard wired (no firmware req'd, no software req'd) to cut power:
(a) to fuel pump
(b) from alternator
(c) to spark plugs
Require Emergency Brake to always be a direct mechanical linkage (again no firmware req'd, no software req'd)
I'm still waiting for an autonomous vehicle to drive as many accident-free, moving-violation-free miles as I have.
Google's entire fleet has a long ways to go - much less an individual vehicle.
Here is the real question - can we really trust the same lazy/incompetent/corrupt coders allowing modern vehicles to be hacked, to write code that will safely drive a vehicle with no driver input?
There is already the problem of, your self-driving car will be programmed to intentionally kill you to avoid hitting a school bus full of kids.
If you want to put your life under that sort of of control - be my guest - just stay the hell off the roads.
Autonomous cars are a pipe dream, a thinking human being does a lot of things while driving a vehicle that can't be broken down into rules which is necessary for a computer to do the job (at least with the current hardware we have). For a simple example imagine driving down a forested highway, you see deer in the forest... you slow down and prepare to stop if necessary.. you are aware of these deer and how they are likely to act (if you live in an area with a lot of deer). Video imaging can't even reliably detect the deer, anyone who's hit one know that they move fast and 2 tons of steel driving down the road at 55mph can only stop/swerve so fast.
Same with people jaywalking, you as a driver see someone on the sidewalk.. there are subtle clues (often not even consciously noticed) that tell you someone is about to start walking across the road... these clues could possibly be coded however try to use video to detect when someone shifts their weight to the street side foot preparing to cross....
Maybe some day in the future we will have hardware that can properly be called intelligence, however until then all we have are rule based systems that cannot think or learn outside of the rules defined.
Autonomous vehicles don't have to be better than the very best drivers...just significantly better than the average. That's not hard at all.
For instance, a supremely alert driver might notice that deer in the forest, but 99% of drivers won't start braking until the deer is a foot from the bumper, and most won't brake at all (deer move pretty fast). An autonomous system will do much, much better than the average driver.
Self driving cars. What a brilliant idea. Because voice recognition software works so very perfect all the time, and so do commercial mapping programs like Google Maps, Tom-Tom, and Mapquest. Never an error in those programs. And humans can quickly evolve in the next 20 years to know the exact address they are going to, without even looking or thinking. There has never once been an internet outage that could possibly take down those mapping and navigation systems. And not once in living memory has a computer glitched, or failed on its own. Furthermore, every car owner takes meticulous care of their vehicles at all times, changing their oil and checking their safety equipment as per the manufacturer's recommendations. I cannot recall the last time I have seen anything but a pristine, well maintained car on the road. And even if someone did actually forget to maintain something on the car, the self-driving program would easily compensate for it all.
As for vehicle "hacking", I am sure if it were even possible to hack one of these vehicles, that the car companies would immediately, and efficiently, rectify the problem. and there is no reason to EVER fear your government commandeering a vehicle to say, serve a warrant, or remove someone they don't like. The badge is the equivalent of a halo, you know. Nothing could possibly go wrong with this system; we have nothing to fear. (Sarcasm off)
Your sarcasm will not save your job in 10 years time. Better start learning to do something else.