Do you have any objective reason to think these are problems the cars can’t handle right now, or is this just projection?
This is a terrible analogy. Airplanes and the current crop of self driving cars have the ability to override the auto pilot specifically for safety reasons. The AI just isn’t as flexible as a person.
That’s just the thing, as has been covered before, the Google car can’t pass the driving test once it hits bad weather, or as soon as it gets off its map. They’re level 3 cars, and pretending they aren’t doesn’t help anything. By the time you can put your ten year old in a level 4 car by themselves, and send them off across town, the car will pretty much be sentient. We’re nowhere close to that.
I’m not aware of any self driving car that can handle snow (and that’s before it’s plowed over the sensor several inches thick). So if they have a solution to problems such as those, they haven’t implemented them.
And really, the most horrifying instances of planes on autopilot going awry are when the sensors (such as airspeed) are malfunctioning, gets the aircraft in a bad state, and the human doesn’t have the good sense to disregard what the faulty sensor is telling them. Sensors failing are a problem that hasn’t been really addressed well, even before we get to AI piloting the vehicle.
I didn’t say that in any way shape or form. Please show me where you got that idea. I said that it was important for the driver to be able to take control of the car.
Post was ate, but comes down to this. Not everyone is dealing with easy every day driving. Today, May 19th brought 15 inches of snow and a driveway and county roads that few people, let alone computers could handle. What are the car sensors going to do when it comes across a 2 foot snow drift that I know I can blast through?
A button to travel across Kansas. Maybe. Get me and my Wife home? Park the car out of the way of our snow plow and tractor? Move it, and get stuck vehicles ready to get winched out by my plow truck? Hey, I would love it if it could be automated. Good luck.
It depends on price, I suppose. For a pure commuter car, maybe this would work. But a significant number of people use their cars for recreational uses, too, where manual control is paramount. Think launching boats or other watercraft, locating your trailer at a campsite, etc. There will clearly be a demand for manual control.
Although, the idea of a self-driving RV suddenly makes considering that type of retirement a lot more appealing! Pull up camp at 8:00, hit the road, and wake up the next day at another campground!
Hell, even backing into your garage without worrying about pulling loads can be a challenge. Maybe not a technical challenge, but a UI challenge. Today I want to park to the left so I can access the tool bench while the car is parked, but tomorrow I will park on the other side for some other reason.
The fact that you challenged (Do you know that for a fact?) instead of asserting it is not an open problem indicates that is (still) an open problem.
There was a story somewhere that a Google car stopped when it came across a puddle (which was, presumably, not mapped - not that the mapped data is fed to the car or anything) - if that is true, the snowdrift is a test I want to see.
How about a driving rain in which the sensor gets a flashing light source?
There are a whole lot more problems with navigation on the surface than in the air - and $500,000 aircraft navigation systems still screw up Air France 447 is a classic - a damned speed sensor killed that plane.
Instead of spending that last billion dollars to completely and safely automate cars, spend it on rail systems - the old streetcar lines can move more people faster than these million-dollar bubbles you imagine.
I’m having a lot of trouble imagining that it has not occurred to the people at Google that snow happens.
First hit on google with “google car snow”: “Google Says Snow More of Self-Driving Car Snag Than NHTSA”
[QUOTE=The fine article]
The technology giant doesn’t intend to offer a self-driving car to areas where it snows in the near term, Chris Urmson, director of Google’s self-driving car project, told reporters today in Detroit.
[/QUOTE]
Well, it’s not clear it has handled snow, since the tests have been in the Bay Area, and we sometimes see it on the mountains, but that’s about it. Pity, since I still have my snow shovels from NJ in my shed. Rain I rather suspect it has handled. Car doors no problem I’m sure, and there are multiple sensors.
Oddly enough, I see lots if Minis on the road which I somehow suspect would have a problem pulling boat trailers. I know lots of people at work with pure electric cars - they have other cars for long distance travel, but for the bulk of their miles, which is commuting, the electric cars work great.
I own a Prius and a big Chevy truck. The first set of people who will buy these cars are going to have money for them and another car too. Or will never go off road or tow anything big.
If these cars ever get close to 100% market share, it will be after lots of development to make more things self-drivable happens.
BTW, if enough people need multiple garage parking spot capability (who has space in a garage for a car anyhow?) it would be trivial to point where you want to go on the image from your backup camera. Which would solve a lot of the other problems also.
I’m sure that stuff like that happened. Then they fixed the code. That’s how development works. They’ve been constantly updating the code for new situations.
How about if a UFO lands on the highway in front of the car?
You seem to think these cars are going to go on sale next year. When they do start, if they start selling them in the Bay Area and LA they have a big market and no snow. Given the number if wrecks I’ve seen in snow storms from our wonderful human drivers, a self-driving car not perfect in snow might still be better. They’d probably be smart enough not to drive 70 mph on ice that one moron in Princeton did in his final trip.
Renee’s idea of a control stick might be the answer. Using some sort of joystick or button pad you can tell the car where you want to go, and the computer can make it happen. Drive by wire. The computer won’t decide where you are going, but it will decide if it is safe to proceed. I would envision this as a relatively low speed affair, and one where the computer would learn an acceptable path.
Cars can already parallel park on their own, I’d bet a “launch watercraft” subroutine wouldn’t be unobtainable for a self driving car.
So now we have an answer to the OP’s question of when self-driving cars will be mandatory:
Never! (well, at least not in the foreseeable “near term” future.)
I would estimate that 70% or so of Americans, and 50%(?) of Europeans live where it snows.,
That still leaves a viable market for robo-cars…but it’s never going to be a majority. And if a city like Los Angeles wants to restrict certain roads to robo-cars , they’ll have to take into account drivers from out of state, especially commercial trucks.
Maybe there will be restricted robo-lanes, similar to the HOV (multi-passenger) lanes we have today.
I believe a phase in of self driving cars will happen in a few years if they are proven safer on a few major roads during peak times, much like HOV lanes and roads are now. I they keep that track record that will gradually expand to more roads.
Not only increased safety, but increased road capacity (including unique lane usages currently impossible with manual driven cars.), reduced fuel, reduced environmental impact, shorter travel time will be driving factors.
Additionally reduced (even elimination of) liability and elimination of traffic enforcement can be achieved.
The question will self driving be allowed. I believe it will be limited depending on conditions and the computer will be able to take control away if it determines the person to be doing something unsafe.
The state of the art is actually the opposite, and we’re ages and ages away from even considering an unattended vehicle, much less one that can override a human.
I’m all for a car that’s safer, but everyone wondering whether self-driving will be allowed is putting the cart before the horse. That sort of development isn’t happening in our lifetime without a serious breakthrough in AI.
I would disagree with this. Google wanted to remove the steering wheel on some of their prototypes but only backtracked on this because the state mandated it. State of the art is removal of the driver.
Added we already have vehicals that will override the driver and apply the brakes.
No, but it will cost you – in terms of type of car you’ll be able to get, and no doubt insurance.
I think a rough analogy with the self-driving car is the amazing concept of the self-shifting transmission. There’s been lots of whining that automatics make cars less “fun” but they’re pretty much ubiquitous today, mostly because people just want to get from “A” to “B” and manual shifts are really a drag when stuck in traffic. But purists can still get everything from stick shifts to BMW paddle-shifts if you’re willing to get a particular type of car and pay the price for it, both in terms of $$ and impracticality – and also, to complete the analogy, the high cost of insurance because anyone who buys a high-performance sporty car is presumed to be someone who will drive it accordingly.
That’s a terrible analogy. The very first thing the automated systems did when they encountered an anomaly on AF447 was disengage the autopilot and auto-thrust and turn control over the human pilots. The humans then proceeded to do just about everything wrong. They killed that plane, with every possible automated device blaring warnings at them. (I realize this is a gross simplification of a very complex situation, but it’s essentially correct.)
I’m all for more and better public transit – it’s essential for many reasons. However I have to go out to the store later and looking out my window there is no rail system between me and the store, and there never will be. Public transit serves major and minor traffic corridors, cars do everything else, pretty much in inverse proportion to population density.
I completely agree. And we’re already getting there – as you mention, self-parking cars exist, as do cars with various forms of collision avoidance, and low-level stuff like ABS brakes. Cars should be more like the computer-controlled drones that you can buy now – you don’t worry about the fine points of controlling them, but only the higher-level function of directing them.
And that needs to be able to be overridden. See ‘snowdrift’
You yourself say that even proposing it was in the context of a prototype. Since they themselves say that the car can’t handle parking lots, or bad weather, I’m going to say the car without the steering wheel or brakes is marketing or engineers reaching. When you have an actual production vehicle that is more capable than a human driver, you might be on to something.
Yes, and most of them can be turned off, if necessary. The human is still in control.