Fully automated self-driving cars, including being able to fully communicate with other similar cars and with the infrastructure, will happen over time. As various posters have said, we’ll get the various pieces, like automated braking, over the next 5-10 years, yet people will still be required to be in the driver’s seat for the foreseeable future. My view is that total automation will take many decades to fully happen, maybe 50-100 years or longer, and by extension, the legal stuff will get worked out along the way.
The biggest hurdle (IMHO) will be the grandfathered self drive cars. If ALL cars were automated, and you didn’t have to account for the vagaries of stupid drivers, I’d hazard a guess that we are already very very close.
Question based on the OP -
Is the accident CAUSED by the software (as in, wouldn’t otherwise have happened) or is the accident because the software didn’t deal with a developing bad situation as well as a human could have?
Amusingly enough, I am a computer tech too, eagerly awaiting my self driving car.
There are plenty of ways to enhance reliability in computer systems. Embedded systems with limited software and limited inputs can be rock solid. Do you worry about your injection computer? Like amusement park rides, you make the “if in doubt stop” the default state, the system has to say “it is safe to proceed” to not stop.
That any but probably the top 5% of drivers might be able to outdrive computers with another 5 years to refine further is laughable. Systems that are 360 degree aware, react in thousandths of a second, able to independently brake or apply power to individual wheels. No human driver can do that. Not to mention things like stopping the vehicle in the event of a collision minimizing or even avoiding secondary impacts will take a huge bite out of injuries.
The stuff already exists, it already works, this is tech that is 5 years out not 50.
Hopefully the dope is still here in 5 years so y’all can mock me, but I predict that by 2018 SDC’s will be as common as hybrids/electrics are now and will be making human drivers who insist on driving themselves look like people with a death wish.
Broadly I agree with you.
And like many (much as I do enjoy driving) I’m looking forward to the day when cars can do it themselves - mostly for the drop in travel time it promises and the lack of jams and accidents.
But I still think there’s some things that humans will sometimes decide better than software (for example - crashing into the delivery truck instead of the bus), and that it will still be possible for a human to have more “experience” or “predictive anticipation” than a computer. Of course - as more and more cars go self drive - “predictive anticipation” will mean less and less because car behaviour will be totally predictable.
The law requires that all owners have insurance. My insurance company is going to be much happier to insure my car computer than it is me. We’ve agreed it is 10 times a better driver. And there are already product liability laws for when a car itself causes the accident that find the manufacturer liable. 10 times safer, is a lot of savings for everyone except lawyers and doctors and their vendors.
Pretty much the same thing. These kinds of vehicles will have event recorders onboard that record all data inputs and control outputs for some large time window. Upon reviewing the event records, the accidents in question are ones that a good human driver could have avoided. Some are from bugs in the software, some are from design flaws in the software, some are from the software’s inability to recognize what is happening because it does not have as sophisticated a vision system as a human.
If that’s the case - I think the liability will be limited - there is no current requirement in tort law that says a driver must be “good” right?
It’s more about establishing which party is more at fault for the cause of the accident - and failure to avoid is very different to cause.
I think you could retroactively equip the grandfathered non-SDCs with basic transmitters that at least convey their basic travel data to the more sophisticated cars. The new SDCs should be able to detect cars without them as well and identify them as safety hazards to be monitored.
The first widespread trial of this technology may be when Google automates its fleet for its new delivery service on the west coast. Transporting packages will be a good test – no humans aboard to injure and lightweight, slow-moving delivery bots that are unlikely to cause any real damage to existing traffic.
Overall a fascinating thread - solid speculative discussion about near-future scenarios.
I see a lot of this playing out, JohnT - with Ravenman’s thinking about self-driving cards leading to a lot less personal car ownership over time, as well as all of the legal and economic implications that other Dopers are suggesting.
Jeez is this going to be a transformative next few decades.
I wonder which countries will be transformed the most via self-driving cars? I suspect it will be China, India, Brazil, Russia - i.e., countries with huge populations, large size and under-developed infrastructure. The U.S. will be focused on sorting out all of the issues being discussed in this thread - those countries will be using self-driving cars to unlock more territories and economic potential…???
I just read an article in Bloomberg speculating that the Japanese will be early adopters because of (1) their rapidly aging population (the elderly are involved in a disproportionate number of their traffic accidents and there will only be more of this in the future) and (2) young people are less inclined to drive. Japanese in their 20s now account for only 13% of total license-holders, down from 26% 30 years ago, a faster decline than the group’s share of the population.
The same demographics are occurring here as well, just not as dramatically. As always, youth will be the first adapters of new technology but this is one that will have strong appeal to the elderly, too. SDCs mean that we won’t have to take grandma’s keys away from her even if it might mean that she keeps punching in the wrong destination code on her car’s computer and winding up in Albuquerque instead of Al’s Grocery Mart.
I would tend to argue that if a human driver had time to make that kind of decision the computer would not have hit either one or slowed down enough to make the resulting impact far less serious. The question is how did you get into a situation to have to make that choice? Would the computer have backed off on speed before you realized you were in trouble?
Unless it really happens in a fraction of a second where neither machine nor human stands a chance of overcoming the laws of physics, the computer will have the advantage over the human. Things that are happening out of sight and unbeknownst to the human driver will have been transmitted to your car in plenty of time for evasive maneuvering. As soon as anything affecting traffic occurs within your car’s transmitting/receiving range, it will change its speed, stop or change course if possible and necessary.
In the case of the pedestrian stepping into traffic unexpectedly, the computer will be faster to react, as will the computers in the cars following so those nasty fender benders can be headed off. If Mr. Pedestrian insists on getting hit, however, onboard recording devices will reveal that human error was once again to blame.
I could see the cars organizing themselves into groups traveling at the same speed, with some distance between groups. That seems to be the most efficient way to organize traffic; it leaves spaces for new traffic to enter and would help to pass everyone through intersections in the least amount of time. The southbound group of 10 cars would time themselves to hit the intersection five seconds after the westbound group of 8 cars has cleared it. No need for anybody to stop.
Tangental, but another obstacle to the widespread adoption of SDCs.
[Conspiracy Mode]It’s none of the government’s damned business where I go and when I go there![/Conspiracy Mode]
And if we can keep Apple from getting the mapping/navigation contract we should be able to overcome the technical issues.
Apple Maps’ Six Most Epic Fails
Those are some examples where I would most assuredly not want the responsible company shielded from responsibility even if the overall number of incidents had been significantly reduced.
Yeah, that was a good point. I bet this: that if we get to a “car rental” society, it will be via a subscription model for most people… and, surprise! That subscription model will somehow price itself out to about the same as paying for a new car and insurance (adjusted for inflation, of course.) So if this were available today, if I were fine with a bunch of Nissan Versa’s and Ford Focus’s, I would pay about $250-350/month for the service, but if I wanted something in the Lexus/Infiniti range, I would pay $500-1,000/month.
So the consumers get better cars, but in reality we would have fewer of them and don’t really own anything… but we end up paying the same.
Bastards! I’m already pissing myself off about this. 
Regardless, I also would not be surprised if the more efficient driving has a measurable (positive) impact on fuel economy.
Of course it will. Cruise control already does.
In fifty years, self-driving cars will probably ‘hook up’ into caravans for cars that are going the same route for a while, gaining the same sort of economies of scale that trains do.
I admit that I don’t get this line of thinking. Seems to me that automated car usage rates will be a factor of costs and competition like everything else. If it’s the same as current ownership costs, that would seem to be completely coincidental.
My concern about car ownership dropping because people are sharing has been that we’ll have a lot of empty cars traveling the highways burning up power. That may still be cheaper for the family than owning two or more cars but it still has a cost nationally.
I’m being a bit cynical there, true. But remember, while the subscription pricing has to start below the cost of owning a car to make it more attractive to consumers than owning one, there’s nothing that says the pricing has to stay that way for the long haul. We’ll end up with a situation like your cell phone provider - you get an introductory rate of $69.99/month, then a year later you look at your bill and it’s $130.
In addition, the subscription companies will still need to keep vast fleets of vehicles available - over 200,000,000 people need to move between home and school/work between 6am-9am every day, and they’ll need vehicles. These cars represent a vast capital outlay in the first place, an outlay that will need to be recouped, and they require fuel, maintenance, insurance, etc, for the life of the vehicles.
Lastly, many pricing models are based on what the customer will bear, and currently we are bearing a situation where we spend from $200-3,000/month on automobiles (including gas, maintenance, insurance, car payments, storage (my garage is part of the mortgage, right?) etc). And car subscription companies will want to, need to, capture as much of that as possible.
I was picturing the types of cars changing entirely. Say you put in a request for a car each time you need one based on what you’re doing with it. If you’re going to work, you just need a seat and an engine. If you’re doing grocery shopping, the car that shows up will have a trunk. If you’re picking up a piece of furniture, maybe something more truck-like shows up. But the times where people need a vehicle with 5+ seats and the ability to haul cargo probably do not cover the majority of the miles driven. If we can cut down on the materials to construct them and the fuel needed to move those cars around, there might be some savings there.
Costs could also be brought down if people are willing to share a ride with strangers. If you’re using a car that’s taking you to work half an hour away and there’s three other people within half a mile of you going to the same general area at the same time, there’s no reason beyond it being a bit weird that you can’t all get taken to work in the same automatic car.
I believe that it could be managed so that there’s not too much waste due to empty cars driving around. The car likely won’t go back to its garage after it takes you wherever you’re going. Instead, it’ll be sent to pick up someone nearby.
Parking of cars in off-peak times and automatic refueling are probably going to be interesting things to see.
A company is not going to suddenly emerge of the size to handle all of America. If a shared service autonomous driving vehicle market develops it will likely emerge in the same growth pattern as the current car share services have - they will just start offering plans that includes ordering the vehicle that meets your needs of the moment to come to you and to deliver itself to its next client when you are done, or, if it is an EV, to a charging station (plugless induction technology will be of big help there). And as there is competition now between car share services and between them and the costs of ownership, so there will remain.
The caravan (car train) approach has already been tested on highways, albeit first runs have been with a professional driver leading the train. The idea of being able to pack cars together all linked as a virtual train with minimal spacing is often put forth as a major part of solving traffic congestion and saving energy.