That's how we see it. I love having fun cars to drive, but for my 15 mile commute sitting in traffic or 22mi for the Tweeze? No point. I'm commuting in a Corolla(5spd, ya I'm a badass;)) We are looking at getting her a prius in the next two years. If there was a viable self driving car sub 30k available at that time, she'd get one. Hell even the Model 3 is on her radar, and they aren't even real yet and pushing the budget! :D

Well around here that would mean, for example, the vehicle would have to know it's safer to plow straight into a deer because the shoulder is icy. Other times there might be room to sneak by on the shoulder when there is good surface available. Can it distinguish between patchy surface conditions so easily?

Always protect the drive and give the owner ultimate control in every situation?
Or sometimes?
Do you know positively?
Or is this some kind of vague concept to be argued by lawyers in the future?

Is this boldly stated in the EULA, cough, excuse me, corporate sales contract?
i.e.: "You bought our physical product. Our product has been tested to work for you. This is what we guarantee it will do."

Let's review the re-affirming facts which give the purchaser(micro-ton or multi-ton vehicle manufacture)l rights as stated for vehicles which operate within the public sphere:
1). There aren't any.
2). There still aren't any. Sorry.
3). No guarantees for you.

Sign the blank checkbook already, you senator sucker.
On the NHTSA level involving millions..

What I'm getting at is that the road will look very different. Imagine a long train of cars, inches off of each other's bumper. ...and there's that exit you need to take.

AI may clog roads like truckers do - with their speed limiting systems. Ever watch a truck over-take another with a 0.1mph speed delta? --expect more of that type of cloggage.

Guys, until the legal hurdles are addressed, probably by government regulation, true self driving cars will never be a reality. Ultimately the technology is the easy part. When the driver causes an accident, they are liable. When the car causes an accident, the manufacturer is liable. THAT will be the primary hurdle.

You'll simply merge into the lane full of AI cars, the AI cars will hit brakes to avoid collision, faster and more attentively than human drivers. AI cars will be easy to bully, I reckon.

I strongly doubt a programmer would allow a car to mow down a pedestrian to save the the driver. I picture young guys prodding some sort of mannequin into the road to cause a crash just because.

Inner city commuter a big maybe.

And whose going to be the first to hack one these?

The liability question is also immense.

Not yet I don't think. I did a research project / presentation on this a few months ago. Level 4 autonomy is only a few years away. The CEO of Toyota shared an interesting perspective on it.

I believe the cars will use environmental indicators like temperature and rain sensors. If below freezing, reduce speed etc.

The manufacturer has already said they will be covering the costs. Insurance companies are hand wringing over this.

Is it really?

AI cars, when the technology is well developed, should be involved in far fewer accidents than human driven cars.

The accidents will be completely documented with sensors, video, black box. When the accident was caused by the other driver, that will be easily proven.

There may well still be an insurance requirement for AI cars.

What should the premiums be for an AI that is never tired, distracted, drunk, texting, hurrying and has inhumanly fast reaction time, perfect vision in all directions, and an accident rate a small fraction of human drivers? Might be pretty affordable insurance. The auto manufacturers can provide it profitably through captive insurers, with no marketing or agents to pay or offices to rent. Gives them a recurring revenue stream.

Want that insurance to cover you if you take manual control?. That might cost extra . . .

The cars can determine road conditions through the stability control system. Measure wheel slip during steering and acceleration. Just like an expert human driver can sense when the road is slick.

Passive safety in modern cars is so good, this really isn't an issue. If the AI car is driving along a dark city street and a hipster dressed in black clothes suddenly darts into the car's path, the AI will have been tracking the pedestrian on the sidewalk, tracked him as he started moving toward the street, detected that he was moving on an intercept course, started slowing before the pedestrian even stepped into the lane, triggered horn and flashed lights and taken evasive action . . . and if all that fails, the AI could steer into a parked car at 25 mph, pre tensioning seat belts and triggering air bags . . . the occupant of the car won't be hurt and the pedestrian will be alive.

The transition from 10% self driving cars-90% human drivers to 90% self driving cars-10% human driving cars is going to be "interesting".


And for any new technology:
The first 90% percent is easy, it's the last 10% that is going to cause the problems.

Hm, as for who the car will save, the driver or the girl with the stroller, I am sure we will see a lot of quickly changed laws very soon. As always when we have to support new technology.

As for "I want to have the AV for every day and take my Porsche when I want to have fun" I hope the AV-technology will NOT lead to that. Because I would not want thousands of unexperienced drivers all over the place, driving their non-AV two times a year.

That is my sense.

There are no purely technical issues with AI cars. Integrating the human as well as anticipation of events beyond the range of the sensors on the car will be, as mentioned, "interesting".

Protocols in terms of en route rerouting, in situ weather impacts, transition from AI to driving the car, failure modes, licensing, etc. reliance on other systems (GPS) will need to be addressed.

Humans will be the major deterrent. I know I will be. I would never buy a car that was AI all the time. I want, as I suspect most would, the ability to take over driving duties.

Going to be 'interesting'.

Surely AI cars will have to be programmed to obey the speed limits?

How many people are going to buy one once they realize that they will be the slowest car on the Interstate?

Really? That will be short lived at least in the USA.

The wife and I just picked up our new cars, both with Lane Keep Assist and Adaptive cruise control. Needless to say, very very nervous the first few times you are on the highway and engage these options. The car does for the most part drive itself. Sometimes when nearing an exit or entrance they will become a little confused but I have booked maybe 1500 miles without touching the steering wheel.
Its happening!

And how long before the aftermarket "turner" chips and programs are out and allow you to set the limit and 5 or 10 or 50 over the limit? Hacking will be a real issue. If someone has possession of the car, it can be hacked.

Of your 1000 inexperienced drivers, on any day only 7 will actually be manually driving ("two times a year", you said).