Slashdot Mirror
California Legalizes Self Driving Cars
Hugh Pickens writes writes "The Seattle PI reports that California has become the third state to explicitly legalize driverless vehicles, setting the stage for computers to take the wheel along the state's highways and roads ... 'Today we're looking at science fiction becoming tomorrow's reality,' said Gov. Brown. 'This self-driving car is another step forward in this long, march of California pioneering the future and leading not just the country, but the whole world.' The law immediately allows for testing of the vehicles on public roadways, so long as properly licensed drivers are seated at the wheel and able to take over. It also lays out a roadmap for manufacturers to seek permits from the DMV to build and sell driverless cars to consumers. Bryant Walker Smith, a fellow at Stanford's Center for Automotive Research points to a statistical basis for safety that the DMV might consider as it begins to develop standards: 'Google's cars would need to drive themselves (by themselves) more than 725,000 representative miles without incident for us to say with 99 percent confidence that they crash less frequently than conventional cars. If we look only at fatal crashes, this minimum skyrockets to 300 million miles. To my knowledge, Google has yet to reach these milestones.'"
Here is a scenario where if a self-driving car can pass 100% of the time, then I would deem it safe to get into.
Driving on a mountain road around a sharp corner where there is a steep cliff on the right side. Auto-car is passed on the left by some *sshole "manual" driver, but then the *sshat driver cuts in short because of oncoming traffic at the last second. Robo-driver identifies there is suddenly a car intruding into its safe-T-zone (TM) and does what its programming tells it to do, avoid hitting other vehicles. So the self-driving wonder swerves right to avoid the other car and zooms off the cliff.
A human driver would recognize that hitting the other car in this instance is the safer solution then to go careening off the steep cliff.
I agree that a self-driving car can work, and 99% of the time will perform adequately to protect its occupants from disaster. But since we have not mastered true AI yet, all self-driven cars will be built with flaws in their logic that will fail catastrophically. "Avoid hitting all cars", for instance, is not a good enough directive to ensure the safety of the occupants in 100% of all situations.
Someone mentioned that the deaths caused by self-driven cars would be far less then manual drivers, but then I would disagree that any technology introduced on the highways would be adequate to allow any fatality, especially in scenarios where a human driver may have been able to avoid death.
Basically what I am waiting for is the inevitable 100 car pile up with massive fatalities that WILL occur at some point in time where investigation will identify that a self-driven car, or cars, was the cause of it. Any company involved in programming or manufacturing that self-driven car will be sued out of existence and the "love affair" everyone seems to have about auto-driving cars will end quickly.
I am amazed at how delusional governments are into so quickly allowing this technology on the roads, sounds to me like there is some massive lobbying going on to short-cut the necessary amount of time to test auto-driven cars under all senarios, not just ones in controlled and predictable setups like we have seen. 5 years ago robo-cars could not drive around a dirt track, now they are quickly being allowed on our highways. That just is irresponsible.
Having lived in CA and driven on the freeways, I can say that you don't need "self driving" cars for the freeways.
All you need is a car that can self park and you are good to go...or...not go.
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
I have been thinking about driverless cars and I'd love to ask the people at Google (or where ever) how they cope with several real life issues
* Emergency vehicles in general
* Vehicles on the side of the road. In general you move over to the other side (road,next lane etc) to give them some room. But where I am (VA) its an offense if you fail to move over when passing a cop car on the side of the road.
* Temporary speed limits posted during road works
* School zones
* Really bad weather where you can't even see 20 feet ahed of you
* Looking down the road and predicting that there will be an issue and doing your best to avoid it (ie slowing down/lane changing to avoid the person on the phone who is weaving from side to side)
* Crap lying all over the road (saw lots of rocks on a mountain road yesterday)
I'm sure there are lots of other "interesting" situations that human drivers have to deal with day to day that would be difficult to encode into hueristics for the self driving cars.
I am Slashdot. Are you Slashdot as well?
... does what its programming tells it to do, avoid hitting other vehicles ...
Its a bit of an assumption to believe that the driving software has that single goal. Staying on the road seems to be something the software is already considering. I wouldn't be surprised if existing software already has "prepare for crash" code that tightens seat belts, unlocks doors, ... maybe even sends an "oh shit" text message to the road side assistance service.
Driving is enjoyable?
Since when?
Sure a race track is enjoyable, twisty deserted roads can be fun, but 99% of driving is mind numbing boredom.
As far as I'm concerned, letting humans drive is putting trust in the other human drivers around me, and frankly, I don't trust them at all. I'd feel much safer if manual driving was illegal.
So if I come out of the grocery store and my car's not there, it might not be stolen?
Tell this to Capt. Sully. When you are travelling over 200 MPH the closing speed is so fast you can't really react fast enough. Um, tell it to the Geese too!
An airline pilots job is hours of boring flights punctuated by moments of shear terror.
Heck, modern planes even try to fix the problems itself. In the famous case of Colgan Air 3407(crashed near Buffalo NY) after shaking the yoke to alert the pilot the autopilot attempted to trade altitude for speed to get out of a stall. The human pilot overrode this safety feature and killed everyone on board by attempting to gain altitude and thus turned a recoverable stall into a crash.
punctuated by moments of being deathly afraid of scissors?
There's a lot of weirdness here.
For example, 725k miles for any incident, but if we look at only "fatal" crashes it skyrockets to 300M? There's a disconnect here: if we look at only "fatal" crashes, I'm pretty sure we can smash up Google cars every 30k without killing anyone and make it to 300M.
If you say, "well it has to be 1 in 300M because that's how often a fatal crash occurs and we want to reduce fatal crashes," you're talking about something completely different. 1 crash in 300M miles isn't likely to be 100% fatal; beating 1 in 725k miles reduces the number of actual incidents, which reduces the number of incidents with a likelihood of being fatal, which theoretically should reduce fatal crashes.
Further, if the car handles better than a human, it may reduce damage and injury in crashes. That means if it crashes more often than a human, it may still cause less total injury and less economic costs--a $300 scuffed bumper rather than a $3000 smashed in side panel, a gashed and mashed-in door on a sideswipe rather than a car's front end smashing straight through and crippling the driver on a T-bone, plain old reduced impact speeds due to better braking and faster brake reaction, etc.
I'm sure Mercedes-Benz, Volvo, and Google can all work something out. Benz has those pre-crash systems (best in the industry); Volvo has reactive braking (alerts the driver, pre-charges the brakes, but doesn't apply braking force until a crash is unavoidable; in a self-driving car, it'd react by suggesting braking to avoid collision to the driving program, which in theory should acknowledge that a likely collision is about to occur and apply brakes along with all other proper maneuvering). If the car is reasonably capable of driving itself, there's no reason it can't reduce incidence and reduce damage in incidents with all these collision prediction systems and automatic pre-crash systems that can brake, steer, and do all kinds of other shit for you.
Support my political activism on Patreon.
I believe on public roads you do need a human available to take over for legal reasons.
And that worked so well for AF447.
Aviation autopilots should have proven by now that relying on a human to take over when the situation is so bad the autopilot can't handle it is a recipe for disaster. Besides, what's the point of a 'driverless car' if I have to be continually ready to take over at a millisecond's notice?
Car: 'Warning, warning, kid just jumped out in the road, you are in control'.
Driver: "WTF? I just hit a kid and smeared their insides all over my windshield'
Car manufacturer: 'Not our fault, driver was in control, human error'.
The usual standard for a statistical "proof" is held to be a 95% confidence, or a p value of 0.05 that the hypothesis is wrong.
Using a 99% confidence interval is skewing the numbers away from the usually accepted standard of proof, which makes me suspicious about the motives of the person proposing it.
I agree Colgan 3047 backs that up even further.
The trouble with them is that they'll take the sting out of long commutes. You already have people who think it's a good idea to spend four hours a day driving for the sake of cheaper real estate. What if they up it to six hours a day when they don't have to stare at the road?
Note: cutting a problem (pollution, car-deaths) would do no good if you double the miles.
Why do Luddites chose to come to Slashdot?
I wonder if the aging population will end up pushing this into reality. We will not make mass transit is not going to work on a large enough scale, and for many transportation needs are only met by POVs. It will become yet another device to assist people's independence, and that I believe will push the technology and laws as the need for it increases.
The problem is that most people overestimate their own driving skill - a 1981 study (Svenson) found that 93% of American drivers tested ranked themselves as being above the median in driving ability.
I never get upset at your wife.
May Peace Prevail On Earth