Waymo recently hosted a number of journalists at its private Castle testing compound, and treated us to rides with no safety driver behind the wheel – now, the former Google self-driving car company is going farther still, however, launching public road tests of its autonomous Chrysler Pacifica minivans with no safety driver on board.
The tests aren’t limited to one or two routes, either; the test area where the truly driverless trials are being conducted is in Chandler, Arizona (part of the greater Phoenix metro area), and the cars are able to go anywhere within this defined space. It’s hard to understate the importance of this milestone: Waymo is operating at full Level 4 autonomy, sharing public roads with human-driven cars and pedestrians, with no one at the wheel able to take over in case things don’t go as planned.
All my friends live at least an hour’s drive away from where I live (assuming no traffic, which is a big assumption in The Netherlands). That’s not a long drive by standards of large countries, but for us, it is, and since it basically comes down to a boring drive over a few boring highways in a boring part of the country, it’s mind-numbingly tedious.
I can’t wait until I can just sit down on the backseat of my car, tell it to drive to Amsterdam or wherever else my friends live, and just chill for an hour with some YouTube or webbrowsing. I know we’re not there yet, but I hope I can at least experience that at one point in my life.
The thing I find interesting about all of this is the number of people who want no part of it, claiming “it’s not safe” or “I don’t trust computers to do this”.
Both claims are equally warranted about human drivers, arguably more so in many cases; especially in the US, where you don’t have to do anything except show up to get a driver’s license renewed (except for commercial ones I think), and it takes some pretty egregious offenses to get it revoked, at which point it’s usually just suspended, not permanently revoked.
They are but you’re no longer in control.
You won’t be owning the car, you won’t be owning the path you want to take (aside from the quickest .. predefined one).
However this does apply to other services as well. We’re just consumers at the end of the day.
Safer, much more profitable, that is the great goal everyone is going to these days.
They are better drivers than humans. Sure no problem there. I totally agree.
We just suck at making computing devices secure:
https://www.schneier.com/blog/archives/2017/08/confusing_self-.html
https://phys.org/news/2017-02-cybersecurity-self-driving-cars.html
I do that already. I take public transport. Why would i want to spend hours of my life each day staring at a strip of tarmac when i could be listening to tunes, having a nap or watching some videos.
I live in the middle of nowhere. Said trip by car is an hour; by public transport it’s well over 2 hours, and far more expensive. No time for that.
Clearly, you don’t live in the US. Take public transport here, particularly where I live, and you’re asking to be jumped and mugged if you show the slightest hint that you have anything expensive on you. Given the nature of my job I’m usually working or taking calls during travel, and the last time I rode the transit around here a knife fight broke out right in front of me. So, no thanks. I’ll stick to Lyft.
That’s not so bad in most places. The WORST parts of US mass transit are: 1 – the hours are inconvenient; unless you work 9-5, you can be screwed in finding a bus/train. 2 – since they have to pick up people everywhere, it takes forever to get anywhere; making routes helps, but that often means you need transfers to get where you want to go. When I was going to college, it took 2 hours and three transfers to get there… 4 hours if traffic was bad. 3 – mass transit is so sparse in the US that it’s badly overcrowded, particularly for the primary hours for going to or from work. Some times the bus is so crowded, they stop picking people up as they can carry no more. THEN you have to wait for the NEXT bus and hope THAT ONE isn’t overcrowded as well.
While I agree with your general sentiment …what’s stopping you from listening to tunes when driving?
(I’d guess it’s one of the primary settings in which people listen to music/radio…)
Edited 2017-11-11 14:50 UTC
There is already a proven autonomous system where humans let a computer “take them for a ride”..
The Space Shuttle.
Now compare the computer infrastructure the shuttle used verses what people are putting in cars.
The Space Shuttle used five AP-101 computers as general-purpose computers (GPCs). Four operated in sync, for redundancy, while the fifth was a backup running software written independently.
So when cars that are drive me down an interstate at 75 MPH and i have no way to override at a minimum need to have that type of framework.
And they DO NOT have it.
Ehh, modern airliners (mass-deployed, non-experimental, safe), with their autopilots and Cat IIIc autoland systems, are probably a better example… and they’re not as extreme WRT computers as the Shuttle was / still no accidents linked to computers.
(besides, the Shuttle gained full autoland capabilities only after Columbia disaster …the idea was that in the event of similar damage to the heatshield, it would at least try to reenter/land without the crew, which would wait on ISS for another Shuttle; plus, the Shuttle docked to space stations manually, while Soyuz, Progress, or ATV dock autonomously; a better example of autonomous system is probably the Buran shuttle, which performed its only flight without a crew …and, in doing it, it sort of demonstrated more fully the whole point behind such winged orbiter (wings giving it crossrange, to land after 1 or 2 orbits at the launch site, while the Earth rotated away) than US Shuttle ever did
)
An even better example would be probably …elevators!
They were once human-operated, too… and now we don’t even stop to think when using an automated one. Cars will be most likely similar.
My disagreement with your points are also in the safety realm.
-) Elevators have fail safe braking, so if the cables snap. It stops moving. Also At no point does the operator of the the elevator have the ability to kill you.
-) planes are a better point, Yet the pilots still retain control and I think most commercial flights require a pilot to monitor the autopilot. And this is really just advanced cruise control. The analogy to fully automated case would be allow the autopilot to land the airplane.
In general the designs of the car “level 5” systems are missing redundancy (sensor, computers,controllers), well documented failure conditions (does it just brake, does it coast??) and 2 independent software stacks to make the decisions.
Maintenance??? What happens when the vehicle is 10 years old limping by on one misaligned sensor, and the idea of fixing it is some JB weld….
As a point, in Louisiana (and other places) traffic lights are mounted on poles and used a pedestrian walk lights. Based off a “How we find traffic light” presentation last year it would render the pedestrian light as a traffic light…. simple things…
And the safety brakes in elevators are basically only “feel good” solution (what you seem to be advocating with cars and how they should have computer redundancies like the Shuttle …which ultimately didn’t make it safe) – when was the last time you heard an elevator cable snapped?
Plus, the operator of the elevator very much has the ability to kill you – if it moves the elevator while you enter or exit it (but it’s telling that you didn’t realise this / it points even more to how you’re obsessed with “feel good” solutions instead of real safety enhancements)
But that’s exactly what Cat IIIc autoland systems do, routinelly / hundreds if not thousands of times daily! (the regulations dictate autolandings even when manual landing would be possible, to train the procedure) Pilots don’t touch the controls untill the plane stops at tarmac. It’s astonishing how many people don’t realise this…
Ultimately, the data we have so far points to autonomous cars being already safer than human drivers. And it will only improve…
Edited 2017-11-14 00:13 UTC