Those features will be ubiquitous in cars long before they are autonomous. However I’m not sure I’d trust a roadmap defined in the '90s all that much. The desert challenge was not that long ago, and we’ve come a long way since then. We’ll be even further in 20 years. I suspect that plenty of people will be sleeping in the front seat - especially those who grew up with the cars and trust them. Which none of us posting here will ever do, I bet.
The hell it’s not!
Before I die, I want to sit in my car with a coffee and a book, tell the car where to take me, and look up again when I’m at my destination. Then I’ll tell the car to go park itself, and call it on a futuristic phone like device the size of a peanut to come pick me up when I’m leaving.
This is not just a minor problem . It seems like a major issue to me.
People own a car because it gives you freedom to go where you want. If a robo-car can’t be told to pull into the McDonald’s drive thru window, or stop off at the convenience store , you’re requiring a huge change in people’s psychology,and most Americans won’t buy one.
Voyager,
I agree. It will be a slow transition.
The system described in the 90s is very close to what is happening. Automatic tolling was top priority and that is now common. The information systems were overtaken by cell phones and the internet. So the data is the same but the method of delivery is better than anything we imagined.
The goal of the system is to increase the efficiency of the road surface. The number of cars in the US will double in the next fifty years. They will all have to use the existing road surface.
Crane
I’m sorry, but you’re contradicting yourself here. This hypothetical car can or can’t get you there. I don’t think you can have it both ways.
But you don’t need a network for communication at the level of turn signals. If the car can avoid a pedestrian, it can see turn signals and brake lights. The network might be useful for things such as platooning, but you are getting into unsolved territory.
Cars don’t have well vetted code at the present time. It’s not even close. The have a long way to go before they reach the level you’re talking about.
Having a static, unflashable ROM isn’t very likely due to software not being perfect, and needing to be updated. At any rate, if its ubiquitous enough tech that I can get the work done at the local auto shop, someone will figure out a way to flash custom ROMs.
Traffic jams aren’t a big problem now, because everyone stuck in the jam can drive creatively out of the problem. Once you get to the point where you’re putting people who can’t drive alone in cars, a bug in a great deal of the driving code could get them into a traffic jam that they can’t get out of. To avoid this, the AI will have to be able to navigate creatively at some point. We are getting closer to that, but I don’t think we’re quite there yet.
Well, this isn’t a dig at you, but most of the problems on the internet have arisen from people not fathoming how their system will be attacked. In the case of using a network to platoon the cars, breaking/overloading the network to break the platoon might be enough for some mischievous fool. On the extreme end, maybe they take over the platoon by spawning or co-opting enough IDs of the platoon that they control it. I dunno, it’s hard to think up attacks for an unspecified network.
If the car is going to transfer control back to the driver, it can’t be used for many (most?) of the hypothetical ideas in this thread. If it might transfer control back to me at any moment; I can’t be unattentive, drunk, or a 12 year old. I might as well drive myself.
I am thinking more on the scale of driving the car onto a government site, or down someone’s secluded private drive that they want to keep private, while its still making its detailed maps of the world. In the proposed scheme it then is going to upload that data to a central mapping server. I’m not sure how great that’s going to go over with some people.
Scab,
Driving is public behavior.
Crane
If a car can get you 95% of the way home, but your house is on an unmarked gravel road, it can still probably drive the rest of the way. It might need to slow waaay down, but if it can creep along the last mile of potholed gravel at 2 miles an hour, then fine.
I know it is infuriating as a driver to be stuck and have to slow way down in heavy traffic. But that’s because you’re stuck in the driver’s seat, you’re bored but you have to be constantly alert at the same time. If you’re in the back seat watching sports center you just care about the total time time takes to get from point A to point B.
And if you’re going home to your own house along the same gravel road you take every day, then your car can learn the road. Maybe the first 10 times you have to take the wheel, maybe you have to put up a transponder or two along the route, but you adapt.
If you’re headed out to a strange place along a deserted gravel road in the middle of a snowstorm while you watch porn in the back seat, well, sorry your self-driving car can’t handle it. I guess self-driving cars are useless.
And of course your car should be able to handle, “Pull over to that McDonalds!” Or “Take me to the nearest McDonalds! How far is that? 20 miles? Forget it. Wait, what else is nearer? Taco Hole is only 5 minutes away? Make it so.”
Our relationship to driving will certainly change as more and more driving can be done autonomously, just like the change from horse-drawn to self-propelled carriages changed the way people used transportation.
But there’s a difference between a driver needing to stand by ready to take over at a moment’s notice when the automated driver fails, and a system that will warn you that in X minutes you’ll need to take control or the car will have to pull itself over to the side and wait for further input.
You can drive your car into private places. I’ve driven my car on the roval at Texas Motor Speedway, and would have been indignant if I’d gotten cited for the speed I was driving at (private, but not really a privacy issue). I have a relative that’s driven onto top secret, (possibly black, I couldn’t be sure) sites while working for the aerospace industry (private, and really a privacy issue if it’s being mapped by the car).
I have thought about that. In most of the examples I gave, what would be the advantage of a five minute warning? I’m still going to be drunk or 12 years old in five minutes. If I’m inattentive, stopping at the side of the road or in a parking lot somewhere is a solution. But if a car that is billed as self-driving knows it can’t reach a destination that its set to drive to, I think it should refuse to self drive. If it took off thinking it could get somewhere and ended up not being able to, I think that’s almost as severe a fault as the AI spontaneously giving up while moving down the freeway. If it wasn’t, then it’s probably going to regularly strand people in a place they can’t get themselves out of easily.
Again, I think self-driving cars will be neat and useful once we get to a point where we can build one, but for most of the useful applications it’s going to need a pretty robust AI. It’s at least going to have to do some limited version of creative problem solving where driving is concerned. I think by the time we have that worked out well, my cyborg body may be available, too*. My guess is 2030-ish, maybe?
*Cars? Screw it, I’m jogging to work!. J/K, I seriously doubt my cyborg body will be able to maintain 70mph without conking out.
Oh, they will get hacked, and the network attack you mention is very plausible. However this is going to happen long before we get driverless cars, so with any luck they’ll figure out how to deal with it. For cars with drivers you just fail over to independent mode, which will have to happen for a host of other reasons in any case. Ditto for the driverless car which had better be able to deal with the case of being the only networked car in the area. Anyhow, a network attack will look like chaos. I don’t think we need to worry about hordes of hackers driving for us. The car isn’t going to transmit enough information for one thing.
But they will get attacked. A society with people so stupid that they don’t have strong passwords for cameras which watch them in the shower (who puts a camera in their bathroom anyway?) is not going to be smart enough to put strong passwords on their cars.
Hehehehe, ok. If you agree the autonomous car will have fits+starts leading well into the the 2020’s, we seem to be on the same page. 
The poor security on the current generation of cars is already being hacked. I hope they will have much better security before they begin to unleash autonomous cars on us en masse.
OTOH, I imagine the cyborg body I hope and pray for at the advantageous year (for me) of 2030 will also probably be hackable. Have you seen the Ghost in The Shell series? No more eyes for you, Mr. Scabpicker! Even so, becoming a government built killbot might be appealing at that advanced age. I just hope I can out-think the current generation of AI on the battlefield with the cooperation of my on-board AI assistants. 
Scab,
It is driver behavior that is public, not the surface.
In the 2020s many of the components needed for smart cars will be standard, but that will be a long way from autonomy.
The next major step will be vehicle to vehicle communication and vehicle to monitor communication. This is being pioneered by commercial vehicles today.
In addition, the most controversial technologies will be Dynamic Tolling and automatic infraction billing.
Crane
I would guess that automated cars will rely on a “ping” network as opposed to server-management. They will broadcast/listen in the immediate vicinity, which is a little more difficult for script-kiddies to break. Transceivers will also measure dynamic signal doppler to confirm that the other vehicle’s heisenberg is in fact what it is reporting. These safeguards are not that difficult to implement and not all that easy to fake. This is not sophisticated AI, you do not need AI to drive a car from there to here, just some elaborate, adaptive programming, not too far off what we already have.
And cars today do not have user-set passwords, they use something akin to PGP key pairs for the remotes. Those are kind of hard to break, especially when they are in constant rotation. I think your security concerns are wildly misplaced.
‘Driving’ is done within the car. It does not require active communication with any central authority. In that case there is nothing to hack.
Crane
The scenario in the OP, like most ethical dilemmas, is unrealistic. If the car retains enough control to be able to choose between veering into the oncoming traffic and veering off the side of the cliff, then it also has enough control to remain in its own lane.
Agreed, assigning ethical dilemmas to machines is unrealistic at best.
I don’t follow. If I take my private car onto private property, how is that public?
Your behavior is public because it can be observed by anyone exercising due diligence.
That is the justification for using automatic license plate readers to track cars in traffic.
Crane
Scab,
Sorry, I believe we have diverged. An autonomous vehicle will function on private property. It does it visually so does not require any mapping or central control. No rights are violated.
The fact that you have turned onto private property, and it’s location, is public behavior.
Crane
Why would there be traffic jams? People are the ones that cause traffic jams. Cars working with a central controller won’t. The speeds may be slower, but cars will keep flowing.
Your car could be sending out erroneous information. But, other cars don’t just rely on the information you send to decide what to do. They would also be monitoring the roadway with their active systems. If they see you doing one thing and saying you are doing another, then a flag should be sent to the local constabulary to come and arrest your ass. They should do the same if you are driving the car yourself. Send a flag to the police that you passed them at 120mph with a snapshot of your license plate and the appropriate data.