Slashdot Mirror
Fully Driverless Cars Could Be Months Away (arstechnica.com)
An anonymous reader shares a report: Real driverless cars could come to the Phoenix area this year, according to a Monday report from The Information's Amir Efrati. Two anonymous sources have told Efrati that Google's self-driving car unit, Waymo, is preparing to launch "a commercial ride-sharing service powered by self-driving vehicles with no human 'safety' drivers as soon as this fall." Obviously, there's no guarantee that Waymo will hit this ambitious target. But it's a sign that Waymo believes its technology is very close to being ready for commercial use. And it suggests that Waymo is likely to introduce a fully driverless car network in 2018 if it doesn't do so in the remaining months of 2017. [...] According to a report on The Information, Waymo's service is likely to launch first in Chandler, a Phoenix suburb where Waymo has done extensive testing. Waymo chose the Phoenix area for its favorable weather, its wide, well-maintained streets, and the relative lack of pedestrians. Another important factor was the legal climate. Arizona has some of the nation's most permissive laws regarding self-driving vehicles. "Arizona's oversight group has met just twice in the last year, and found no reason to suggest any new rules or restrictions on autonomous vehicles, so long as they follow traffic laws," the Arizona Republic reported in June. "The group found no need to suggest legislation to help the deployment." According to the Arizona Republic, a 2015 executive order from Gov. Doug Ducey "allows universities to test vehicles with no driver on board so long as a licensed driver has responsibility for the cars and can take control remotely if the vehicle needs assistance." Waymo is getting ready to take the same approach.
Let me guess: the "extensive testing" took care of that problem.
...how is this ride-sharing?
Debate is a form of harassment. Do not question my truth.
Check list.
* Favorable weather
* Well-maintained streets
* Lack of pedestrians.
* Everyone driving slow golf carts
* Shopping malls don't always work.
Sounds like not really ready for prime time, just cherry picked locations.
Not any more than pulling the same jack-assery on a meat-bag driver. Don't forget these are literally mobile data gathering systems. It will accurately record how maliciously(and awkwardly) you threw your three hundred pounds of cheeto smeared mountain-dew dribbled slovenly self in front of its nicely maintained chassis.
Only I can judge you.
What all of you luddites are missing is a very simple point - a driverless car only needs to be a little better than the average human driver, and it will take over like wildfire from humans.
Not only is there the obvious savings of not paying every increasing salary and taxes to pay a human driver, but there's the massive hidden savings on insurance costs as driverless cars ratchet up the quality of performance relative to again, the average human.
The demand will be enormous, not only driven by taxi/trucking companies, but by an aging population who no longer have to worry about driving as senses and reflexes deteriorate.
As for the examples you list - most of the deployments would be hard pressed to be confused by much, they will not need ANY network connection because on-board they will have the entire system they might be driving within already stored for comparison with external sensors. And the examples you gave? Why would any *human* driver not be equally confused by a bag or a ladder, plenty of accidents have been caused by people swerving to miss something innocuous - a computer system can classify and react to something WAY faster than human can, while also knowing with certainly if cars are to either side to maneuver - something most humans cannot handle.
I cannot believe how behind the times most Slashdot readers seem to be of modern performance and capabilities of neural networks, to the point where calling this place a technical site anymore seems pretty questionable. I never really bought fully into the idea of the Singularity, but there are facets of life like driving where that concept is obviously valid.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
sarcasm ;), ;) If you do not settle with us legal action :"WILL BE" taken. . zzzzzzzzzzzzzzz! ;) lol
/sarcasm
;)
;) And a central point to this is the degree of risk the CEO's, bureaucrats and politicians are accepting for the public at large as the kinks get worked out of this technology.
I would like to know what happens if one hits me. Does the vehicle just try and drive off and perform a return to base, while I am trying to get the lic number and vendor id before it gets out of sight.
Will it drop a little paper note saying, "Call this number to settle up with us over your accident."
And when you call you get a message saying, "Your call is very important to us! Someone will be with you soon."
Do these vehicles have orange flashing beacons on the roof letting everyone know "Warning, Be Careful, Be Aware, Automated Machine in your midst!"
Do they have any liability protections built in to the laws to protect the companies deploying these vehicles?
Is it assumed the human is wrong and the autonomous vehicle is right?
Do the companies have complete control of all the logs on the vehicle before the authorities? Do they get to choose what to hand over and when?
I am very suspicious about this being fast tracked. About issues and accidents being covered up. What are the liability protections for the public and the riders?
I think it would also be interesting to hear from inidivduals working in one of the Amazon warehouses that have both bots and individuals working in them. What is their experience? Maybe that would be informative
I am not against this, I just think the path is longer than most think
There are five SAE accepted levels of autonomy:
Level 0: No self driving features
Level 1: Some driver assistance
Level 2: More driver assistance
Level 3: Conditional autonomy
Level 4: Nearly autonomous.
Level 5: Completely autonomous.
When will it get here? Dates range from 2017 (Ol' Musky) to 2026 (president of IIHS) and beyond, from people in the know.
Every bit of driver assistance I think is a good thing, but Level 5 - true autonomy - is still a ways off, it seems to me.
In Phoenix we call "warm sunny days" Winter.
Waymo's self-driving cars work by having an extremely detailed 3D map of the area in which they are driving (important details like the height of the curb are included). With such a detailed map, self-driving is easier. You don't have to do object recognition of things like stop-signs, you know where they are. If there is an unknown obstacle detected, the car can stop. It doesn't have to distinguish between pedestrians and trees, because the trees are all mapped out.
If Waymo wants the car to leave that small, well-defined area, then they have two options: map the whole world (at a much, much higher quality than Google maps), or develop new algorithms for self-driving cars.
"First they came for the slanderers and i said nothing."
They said we'd have flying cars when my grandfather was a kid, by the time he was an adult it hadn't happened. He was born in 1913. They also said that when my father was a kid. He was born in 1949. They said that when I was a kid too, that was ~40 years ago. And people have also been saying driverless cars, robots that will do everything and a life of ease that we'll never see.
Serfs might have worked harder, but they had a hell of a lot more leisure time then we do. I doubt we'll see driverless cars within the next 40 years. Hell even the trucking companies who've run trials with driverless trucks are finding that while the workers complain, they still got a better return. If we see a demand for driverless stuff it'll be there first, especially since there's a huge demand for drivers in north america and not many people wanting to do it.
Om, nomnomnom...
Well, the summary hints at it, but in TFA it states it outright:
The company has built a real-time command center that allows self-driving cars to "phone home" and consult human operators about the best way to deal with situations it finds confusing. The ability to remotely monitor vehicles and give timely feedback on tricky situations will be essential if Waymo hopes to eliminate the human driver from its cars.
So they are taking a hybrid approach, at least initially.
According to TFA:
The company has built a real-time command center that allows self-driving cars to "phone home" and consult human operators about the best way to deal with situations it finds confusing. The ability to remotely monitor vehicles and give timely feedback on tricky situations will be essential if Waymo hopes to eliminate the human driver from its cars.
We can fly drones in Afghanistan from Nevada, it's not too far fetched to think they can have a remote driver for tricky situations. One human could supervise multiple vehicles.
Perhaps when it's Waymo/Google's cars that get stuck trying to make a left turn across a four lane road at an uncontrolled non-intersection during rush hour because Google Maps thought that was quicker they'll get around to correcting those suggestions.
This Space Intentionally Left Blank