Slashdot Mirror
How Do You Give a Ticket To a Driverless Car?
FatLittleMonkey writes "New Scientist asks Bryant Walker Smith, from the Center for Internet and Society at Stanford Law School, whether the law is able to keep up with recent advances in automated vehicles. Even states which have allowed self-driving cars require the vehicles to have a 'driver,' who is nominally in control and who must comply with the same restrictions as any driver such as not being drunk. What's the point of having a robot car if it can't drive you home from the pub while you go to sleep in the back?"
to ticket a driverless car. The car, by design and foregoing any human intervention, will obey the law exactly as it is programmed to. It will not speed, it will not swerve, it will not disobey traffic signs nor will it deviate from its programmed course unless directed to by human intervention.
Ergo, if the driverless car fails to function as specified, then the manufacturer is to receive a citation for the vehicle's failure, or otherwise the human who was in control at the time of the infraction will receive the ticket. The car itself is irrelevant.
We are at the early stages. Look at the laws from the first few years of automobiles. You had to walk in front waving a lantern. And go slow enough that the cop on horseback could give you a ticket. What's the point of a car with laws like that?
Peter predicted that you would "deliberately forget" creation 2000 years ago...
Have you ever played with sensors before? They aren't perfect and can give incorrect readings (depending on type they may not even be very accurate under the best of conditions.) Which means software must be written to take those conditions into account, and usually coordinate among different types of sensors. But that software is written by people, and may have bugs in it (in fact certainly will) plus may simply not cover all the real world situations perfectly.
In some cases where people would have a tough time driving, the cars may do awesomely, but in cases where people would have little trouble the cars may behave strangely as sensors give odd readings, etc.
My phone has an incredible processor in it and can handle millions of calculations per second, but it still locks up sometimes, occasionally responding seconds later to all the stored input. Isn't that pretty close to being distracted?
Don't get me wrong, I want a self driving car so badly it hurts sometimes, but I don't expect it to be perfect. And if that's what people expect they are in for a world of disappointment and pain. And my fear is that will mean people panic at the first accident and push back against allowing them at all.
The human component is just there in case something unexpected happen on the road that self-driving cars may not be able to react to in time. While such disaster scenario may be rare, the possibility isn't 0%, which is why you need someone who is able to drive.
It's also possible that relieving the driver of the drudgery of driving during the vast majority of uneventful rides will actually deprive him of the instinctual familiarity that would allow him to react correctly in those marginal cases. That is, the purpose of keeping a human being in the loop just for disaster scenarios might be self-defeating if the driver does not possess the experience to best resolve the situation.
If by "gets distracted" you mean "is an entitled narcissist" then I agree. Robots will take the deadliest thing out of the driving equation, ego.
http://www.rootstrikers.org/
Perhaps, but a 9 YO that is paying attention is probably a better driver than most people out there.
On the other hand, with humans, each human learns how to correctly deal with situations. With computer drivers, they ALL learn from one mistake.
You must never have played a game with AI.
Here is a hint, we are actually very bad are creating smart machines. A 9 YO would be a more intelligent driver any most super computers.
In relation to the present discussion, I'd have to say that Google's driver-less cars pretty much put the lie to that statement.
In August 2012, Google announced that they have completed over 300,000 autonomous-driving miles accident-free, typically have about a dozen cars on the road at any given time. Not explicitly stated in their announcement was how often the driver had to take command.
Further, the summary above may be wrong, because the Nevada law also acknowledges that the operator will not need to pay attention while the car is operating itself, which implies the State has no reasonable expectation of holding the driver responsible for accidents.
Sig Battery depleted. Reverting to safe mode.
Phones are not realtime operating systems. Embedded vehicle components are, and are based on operating systems designed from the ground up to ensure that when a process is going to get its quantum of time, it gets it. No app in the background hogging CPU, no blocked I/O from a long write to flash, no fooling around like on a general purpose OS.
Of course, automotive/marine critical stuff is expensive, but you get what you are paying for. This isn't just another x86 processor that is running Windows and autostarting a "run engine" app.
A self-driving car will have the circuitry most likely on a dedicated processor using CANbus.
I really don't know why this is a difficult question, I see a simple law that solves the problem:
1. If there is an occupant in the car who holds a local drivers license, they are required by law to sit in the drivers seat, and they are responsible if the car is on autopilot or not.
2. If there is an occupant in the car who is unlicensed or incapable of driving they must not sit in the drivers seat and rule 3 applies.
(ie. this is what you do when you are drunk)
3. If there is no occupant in the car (eg. the car is driving its self to pick you up), the owner of the car is responsible as if they were driving.
(ie. If your car kills someone because Sergey programmed it wrong, you go to jail. You knew this was the law when you purchased the car and sent it off on it's own so don't bitch about it.)
4. For civil claims (that is, if someone is seeking money from you in damages), and it is proven that the software was at fault, then the liability is joint and several. (ie. the person who is suing can take you for what you are worth, and take google for what they are worth).
This is easy for lawmakers because there is always someone in their jurisdiction who is liable for the car, and as the owner, you need to trust that the software works. If you don't trust it, don't buy one.