I think we see driver assist AI become standard on cars for about a decade before full autonomous cars become a thing. Keep in mind right now most cars have a button that effectively lets you set a speed and the car will then hurtle forward in a straight line, blissfully revving the engine to keep you at 75 mph without being aware of anything around you. It would be nice if cruise control could follow a lane and press the brake if there was traffic. That alone would save many lives a year, and would be a feature I would use. Sure I would still drive in cities and when its raining, but next time I'm on an open road, I'd be happy to let the computer take over.
edit: Also, the latest model S update does what you describe, it has lane change assist and follow during cruise control, maintaining a certain distance from the car in front of you.
I nerds who live in a city and don't really use cars other than to move them and a little shopping around forget about farmers, post men, telecom engineers, water engineers... if I get to a job and need some tools I can get them delivered from the depot while I work on something else! If I need a mini-digger I can order it and get it. Or if I need to take a fallen tree to the dump - easy. All these things I do in my little car at the moment, I have to compromise.
Right now I have _three_ cars -- one I use for commuting, one my wife uses all day. and one (an old van) that I use for all my dirty projects. I would love to replace my commuter and project van with _more affordable_ versions of the same that also have the same flexibility. The last part is the limiting factor now -- but won't be for long (and I'm not even talking autonomous cars yet :) )
To take this to an extreme, humans driving cars manually might just become a sport.
Now the only thing I really want is being able to follow a lane and take over steering and simple line changing. Then I can sit in my car at night, take a nap and end up in san fran from seattle.
Highway driving can be severely automated since its long and tedious with a simple task of going on a straight road for ages.
Granted I don't think it would work as well in driving conditions like those found in other parts of the world, but I still think they're rather advanced in their abilities. I was impressed at least.
Much more fun would be jumping in front of the beer truck while your buddies empty the thing out the back. I would imagine crime would be a huge issue with these things. On a par with the ratio of car burglary smash and grab vs carjacking crime rates.
I would imagine those things would have elaborate and easy to use nationwide systems to shut down in case of an accident or runaway vehicle, such that no matter what antennas were knocked off or radios burned out a fireman anywhere in the nation could just pull a lever... once the criminals learn, it'll be open season. Water softener salt delivery truck is probably safe, but the beer truck is in serious danger...
Replacing licensed drivers with rentacops wouldn't necessarily save all that much money as rentacops aren't that much cheaper than drivers, and if the rentacop is sitting there doing nothing while the truck drives, some beancounter is going to demand they work, perhaps by driving the truck...
Something that's interesting to think about is people who've never worked retail have some peculiar ideas. I worked retail as a student and got to know some delivery guys and they spent surprisingly little time driving. They'd go from the warehouse to the store to gas station to bar to restaurant wheeling around kegs and 6packs, stocking store shelves by hand, and selling to the businessmen. Data entry filling out little forms and receipts for each custom delivery, etc. Stopping them from driving would have about as much effect on their employment as the deployment of hand trucks to wheel around beer. It would make the job a little easier, but not much.
Of course you could outsource the labor to the recipients, but then you go from 10 trucks and 10 drivers delivering to 100 sites, to 0 drivers and 100 reception clerks at each site to do the manual labor... Given two competitors the guy with the drivers will provide a better cheaper service than the high tech competitor, so ...
It seems to me that that the first, lonely, autonomous cars will/do have a harder time technologically than subsequent versions because they'll have to comprehend/compute more by themselves.
Once 1000s of cars are on the road all communicating with one another ..and pedestrian mobile devices, signs, nearby buildings etc things could get easier. An economy of scale that is a result of ubiquitous mobile devices and sensors.
Snowy roads are actually much better in many ways for cars (theoretically) than humans, because cars have much shorter reaction time. As for obscured road signs, if self-driving cars become the norm, road signs will likely disappear, because (A) they will use wireless communication to transmit information to cars, rather than visual scanning, and (B) a lot of information contained on road signs would simply be irrelevant or redundant for self-driving cars.
Remember that modern cars are already very autonomous, even though they "feel" manual. Recently, I drove a car[0] that will actually automatically steer the car back into a lane if it senses that the driver is veering on the highway (ie, if he/she falls asleep or is drunk).
We don't think of this as "self-driving" because the improvements are incremental and the driver is still present and feels like he/she is in control, but drivers have been slowly giving up control over cars for several decades now. Self-driving cars are actually much less of a leap from the status quo than people realize; it's simply[1] the final step in a long series of incremental automation of automobiles[2].
[0] Subaru, I believe.
[1] Not that it isn't still an impressive feat, of course.
[2] Let's not forget the prefix or etymology of the world automobile - it's not accidental!
My current car has a lot of automation, and all of that automation falls apart in the winter. The automatic windshield wipers don't know how to handle snow or freezing rain, the automatic brights try to go bright when it's snowing heavily, I would never dare try to use the automatic cruise control on anything but bone-dry roads... even the rear-view camera doesn't work without frequent cleaning due to the aerodynamics of the tail of the car coating it with sand and snow in almost any conditions.
The only two automated parts which work as intended during the winter are the anti-lock brakes and automatic four wheel drive.
This is what control systems get very good at and apparently the traction control firmware in many cars that controls anti-lock brakes can include code that detects incorrect steering in icy scenarios and compensate for it. (sorry no reference - I heard this on NPR a while back)
More likely a pedestrian in the road, a cop pulled someone over, a bicyclist is on the shoulder and you feel the need to give him some space, there's an accident, a crate with a refrigerator in it just fell off the truck in front of you, it sounds super creepy. Set off an alarm of some kind, sure. However, having a "haunted car" take control away from you and mow down a little kid because it thought it was helping you drive, sounds like a recipe for PTSD, and the related gigantic lawsuits, in addition to the lawsuits for the dead kid.
And of course the legal liability protection that driverless cars will require will result in car designs like the bad old days, so now when your gas tank explodes in a minor fender bender your heirs will be legally unable to sue because driverless cars have blanket immunity.
We already have park and ride buses, I think autonomous cars (buses) will fit in just fine. No one cares if we need to carve out some 'auto-bus lanes' and traffic lights for them, sort of like trams used to be. If it saves money, then countries in Europe that are still in recession will be all over it.
I'm in India, and there are so many times an hour that driving here makes me wonder what autonomous cars is all about.
For example, that guy driving on a two wheeler on the wrong side of the road, with 4 kids somehow balanced on the rear seat? How will an autonomous car parse that, or how would it parse a road where someone decides to take an illegal U-turn just because they can?
Even today, I saw something that surprised me, despite having see just about everything. A bus playing chicken with another bus, so that it could overtake from the wrong side and barrel through an intersection. While I fully understand, that a car will be able to respond through a shared data base of reactions, its still a situation which requires far more of a car than what is being promised right now.
I think the promise of a fully automated city transportation grid is actually what is being discussed. When all traffic obeys the same rules, the cost for automation and the expected degrees of freedom get drastically simplified.
"that guy driving on a two wheeler on the wrong side of the road, with 4 kids somehow balanced on the rear seat" ... what's to parse? A moving object, try to avoid. Stop to avoid danger.
In montreal, trains user rubber wheels. It totally makes sense to have auto-bus lanes and auto-highway lanes where every car on the lane has to be in automatic advanced cruise control mode.
The only problem is a legal one - it's inevitable that at some point a driverless car will hit someone. Even if the total number of pedestrian collisions drops dramatically (which I'm sure it would) the questions of responsibility and culpability become a problem.
So, I really don't think it's as much of a problem as you think it is.
And the vehicle manufacture would only be considered "at fault" if the vehicle fails safety regulations.
The only time the consumer (driver) would be at fault is if they tampered with the vehicle, or entered it knowing it was damaged.
Thats how insurance works in the state that I live in.
One minor problem is I'm not sure exactly what the average driver experience age is, but I know new teen drivers are incredibly expensive to insure and after 10-20 years of primary driver experience cars start getting pretty cheap to insure (until your kids are teens, of course).
Changing a cars primary driver without informing the insurance company will get you cancelled. I can't basically gift my old car to my teen cousin while paying "safe old man rates". They'd be more than happy to insure my old car, at "crazy teen boy rates". So I suspect a software upgrade of the driving system would require insurance company contact.
Observationally, today's github release will be insured at a small integer multiple of "crazy teen boy" insurance rates.
This develops into other problem areas. At some crossover point its going to be cheaper to hire a human driver... My gut level engineering guess is it'll never be cheaper to run an autonomous car on shared roads than to run an autonomous train on dedicated (elevated?) rails, so throwing capital at the problem is a losing strategy, you'd come out way ahead financially building elevated monorails or just going with modern intermodal trains.
> ...
> I can't basically gift my old car to my teen cousin while paying "safe old man rates"
I don't think you understood what I meant by "insuring the vehicle, not the person". If cars are autonomous, then the car owner ("driver") becomes completely irrelevant to the insurance price calculation. There is now no such thing as "crazy teen boy" prices.
So in other words, to contradict your final paragraph, at NO POINT will it be cheaper to insure a human driver. And given how empty roads will become once cars are autonomous, it seems crazy to build a whole new transportation infrastructure.
Stage 1. Limited self-parking cars. These will only be self-driving in large parking lots that have been designated as allowing self-parking. These parking lots will be at shopping malls, stadiums, concert halls, large grocery stores, home improvement stores, and the like.
When shopping at a place with a designated self-parking lot, you manually drive to the front of the store, get out, and your car drives to a parking spot at the far end of the lot (leaving the spots closer to the door for people who do not have self-parking cars). When you are done shopping, you summon your car via your smartphone, and it comes pick you up.
With a little planning, such as a little separation between where people enter/exit their self-driving cars and where manually parking people will walk, it should be possible to arrange it so that people who are skittish about self-driving cars don't come to close to them.
Stage 2. Limited street access for self-parking cars. Some parking lots and garages in cities will be designated as allowing self-parking, and some street routes leading to those lots will be designated as allowing self-driving cars to travel between designated points and the parking lots.
Essentially these designated spots will be like bus stops, except for self-driving cars. You manually drive to one, get out, and your car self-drives to a nearby parking lot. When you want your car, you go to one of these designated spots and summon it.
The participating lots, the drop-off/pickup spots, and the routes between the former and the latter would be chosen to keep the self-driving confined to a few well labeled streets, so that people who do not yet trust self-driving cars can be extra alert on those streets or avoid them.
Also there would be freeway access for self-driving cars during stage 2.
Stage 3. Expansion of stage 2. Designated routes expand, and when there are enough of them we flip to allowing all routes except designated forbidden routes. Drop-off/pickup can happen anywhere near a designated route (and later, anywhere that doesn't require using a prohibited route). Parking flips from only allowed at designated lots to allowed anywhere except designated forbidden lots.
The focus in stage 1 and stage 2 is use in limited areas at low speed, while the public gets accustomed to the presence of self-driving cars. The reason for the emphasis on parking is that for most suburban or city driving (outside of rush hour commuting) the driving to your destination is usually not overly annoying. It's dealing with parking, or carrying your purchases a long distance to your car, that is the most annoying part.
In stage 3, the network of available routes has expanded enough to span from your home to your destination, so you can let the car self-drive the whole trip.
Maybe it can drive around some of the roads in the US after it has been fed with detailed 3D maps of the whole area along with road signs. But how about in other places ? Could it ever drive in a place like India ? I don't think so.
Autonomous vehicles could completely take over in closed areas like Open cast mines , vast factories and on farms where companies could profit by cutting the driver out of the loop. From there we would see the technology slowly improve to start getting traction in the real world. Even then the vehicle wouldn't be fully automated , it would be a more advanced form of the cruise control we have today with it falling back to the manual driver especially in crazy places like India.
Also we'd need laws forbidding people from jumping in front of self driving cars etc just to f@#$ with them.
This is also a trend with automation stories in general , with several publications already predicting doom and gloom and claiming that most people would end up unemployed because of automation of their jobs. The fact is we don't even have an AI good enough to impersonate a tele marketer yet and it's decades away at best.
I'm not denying that thesis in the extremely long term but AI and machine learning are in a very primitive state for any of this to materialize any time in the next two or three decades.