Slashdot Mirror
NHTSA Gives Green Light To Self-Driving Cars
New submitter tyme writes: Reuters reports that the U.S. National Highway Traffic Safety Administration (NHTSA) told Google that it would recognize the artificial intelligence in a self-driving car as the "driver" (rather than any of the occupants). The letter also says that NHTSA will write safety rules for self-driving cars in the next six months, paving the way for deployment of self-driving cars in large numbers.
So is each individual instance of an AI a driver? Each version of the software? Each combination of hardware and software?
If a single car is found to be doing something that would have its license revoked, does that car lose its license, or are all Google cars immediately banned from driving? Would a version tweak cause that license to be reinstated, or would Google be out of the self-driving-car business?
++ Say to Elrond "Hello.".
Elrond says "No.". Elrond gives you some lunch.
If I don't have to lose an hour each way on maintaining moderate concentration, moving out of the suburbs into the country suddenly becomes feasible. Sweet! NHTSA approval is a major milestone in this becoming a reality.
Suborbital [spaceflight] is the special olympics of spaceflight. - Rei
I'm suggesting if Google is driving, and the passengers are passengers, then why the hell would anybody pay for things like liability insurance for an AI?
Could it be because it's still going to have a "fuck it, you drive" mode which passes responsibility to the human so Google can claim they're not responsible?
A self driving car becomes useful when I can have no controls, and be asleep in the back. I don't pay liability insurance on a bus, train or taxi ... why the hell would I pay it when something created by Google is in charge of driving it?
Lost at C:>. Found at C.
Until they can flip a bird and show unambiguous road rage
Phase 1: limit AI driven cars to say 35mph or under "network" control (in either case Hazard Lights GO ON)
Phase 2: increase speed by 10mph and put laws in place that a car in AI mode is exempted from DWI (as long as the car is driving directly HOME or to the nearest medical facility)
Phase 3: increase speed by another 10 MPH (or to current speed limit)
Phase 4: AI cars allowed to not have Hazard lights ON unless otherwise needed ....
Phase N: AI cars allowed to travel without somebody in the car and to pickup children (note we had better have KITT level AIs in cars at this point)
In the UK, most tiny karate clubs have a GBP 1m public liability insurance, and it costs a pittance each year.
The fact of the number makes no difference, it's what's covered. I imagine they have to cover a lot more, but even the WORST of these may be better than human drivers on average, so it will quickly re-balance once the risk statistics are apparent, even if companies only pay at first for their testing cars.
Honestly, $100k+ liability insurance is pretty low. Even a school will have GBP 5-10 million and it get claimed on all the time and they handle care of children, including activities, trips, sports, staffing, etc.
...until the first time AI kills someone.
1. AI Manufacturer pays millions and millions in damages?
2. AI Manufacturer finds a way to pawn off responsibility on to the owner.
3. AI Manufacturer passes a law capping damages and maybe even some kind of limited indemnity for the AI Manufacturers
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
But with the consequence that the first accident of note will result in all kinds of problems for EVERY instance of that model running in EVERY model of that self-driving car, rather than just a single driver being an idiot.
Assuming that accident is the fault of the AI, then you can reasonably expect a patch within a week, a month if the issue is extremely complicated. Good luck fixing human drivers this efficiently.
'The Great Race' was a 1965 movie and also an American tradition. Competitors race from one side of the country to the other in various vehicles with various rules.
Self driving cars will surely do the same. They will be judged on safety and speed and technicalities like choosing the best route and handling obstacles. Car buyers will want this information and car makers will struggle to optimize their software to win the next race.
...omphaloskepsis often...
Google "expresses concern that providing human occupants of the vehicle with mechanisms to control things like steering, acceleration, braking... could be detrimental to safety because the human occupants could attempt to override the (self-driving system's) decisions," the NHTSA letter stated.
Bullshit. Vehicles must have a full set of manual controls available to the human operator at all times, and furhermore they must be fully educated, trained, licensed, and insured, just like always. To do otherwise is what will put people's lives at risk. Google is smoking crack and needs to be put in their place.
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
Tickets / civil liability / criminal liability?
And with tickets you have
parking ticket is issued by a private, non-governmental parking authority patrolling an office parking lot or shopping area
private security guards issuing live speeding / moving tickets (non state tickets) (some HOA's and some parking lots or shopping area)
moving tickets from a real cop
parking tickets
red light tickets (parking like)
red light tickets (moving like)
speed camera tickets (moving like)
speed camera tickets (parking like)
toll violations (I can see an auto driver car with bad DB info getting one in some settings)
That's my major problem with this technology: there's an awful lot vague answers to specific questions.
A "self driving car" means you put little Timmy in it, send him to school, and monitor it on your cell phone to confirm he gets out in the right place and a teacher has collected him ... or it means you come out of a bar, fall into the backseat, and say "home, James" ... or it means grandpa who has lost his vision and his driver's license can get in and say "take me to my doctor's appointment".
No driver's license or legal responsibility for operating the vehicle at all. You are livestock being transported. You're not driving or operating, you simply told it your destination.
This bizarre model in which the car drives, except when it doesn't, and with no clear demarcation between is damned near impossible to make sense of.
If the car decides it's got no idea what to do, and it just says "you're in charge", and before you even know what's happening you're in an accident .. and the logs say "human was driving, his fault", you're screwed. Or, worse, someone builds in code which lies and just says "human was driving" 5 minute before any crash is triggered (so they can avoid liability).
There can't be a gray area between who is in charge and who isn't. And paying for liability insurance when the computer is in charge sounds moronic to me, why would you do that? Are you accepting liability on behalf of the computer or something?
Self-driving-ish cars? Autonom-ish cars? It just seems like everybody is pretending this is a solved issue, and I don't believe it is.
Lost at C:>. Found at C.
You're arguing a situation that won't happen.
It's simple, if the car runs into a situation it doesn't know how to handle it will come to a stop. At that time the human operator can take control. The car won't just hand off control without warning, in fact the car won't hand off control at all -- the human would have to take control.
"Grab them by the pussy" -- President of the United States of America
Wonder how much Google public liability insurance premium just increased by.
Nothing. It's called Alphabet now. The self driving car subsidiary has probably been left holding zero assets to handle the possibility that a horrible accident occurs and the victims try suing the company.
Have gnu, will travel.
Seriously... I can't think of any way shape or form that the "AI" behind a "self-driving car" is anywhere near ready for full legal responsibility for this.
An AI cannot have legal liability; it is a machine. Depending on how this shakes out, either the auto seller (Google, Toyota) or the auto owner will provide insurance and have legal liability.
And since human drivers are almost universally incompetent, as long as the AI driver is more competent than the average human (a low bar), the insurance will be cheaper than insurance for human drivers.