Google describes its work on robotic cars with its typical tight-lipped optimism but academic experts in robotics are cautious about the prospects. They estimate it will be decades until they can perform as well as human drivers in all situations if they ever do at all.
When surveyed by the organizers of the Automated Vehicles Symposium, the 500 experts in attendance were not optimistic such problems would be solved soon. Asked when they would trust a fully robotic car to take their children to school, more than half said 2030 at the very earliest. A fifth said not until 2040, and roughly one in 10 said "never."
For an alternative viewpoint, consider the first and fourth(!) of Sir Arthur C. Clarke's Three Laws:
- When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
- The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
- Any sufficiently advanced technology is indistinguishable from magic.
- For every expert, there is an equal and opposite expert.(*)
(*) Arthur C. Clarke's Profiles of the Future (new edition, 1999).
(Score: 1) by bitshifter on Sunday July 27 2014, @03:04PM
IMHO The problem is not the machines - it's the people driving around them and making stupid things.
If we can solve the culpability issue, we will see this very soon.
Oh, now that I think of this - who is responsible when a car with automatic traffic jam breaking system causes an accident?
Is this not the same?
(Score: 2) by Nerdfest on Sunday July 27 2014, @04:09PM
I was going to say ... as *well* as human drivers? Please.
(Score: 2) by forsythe on Sunday July 27 2014, @05:52PM
I'm not sure if I've seen enough lay public rallying, nor enough distinguished but elderly scientists denouncing, to apply it in this case, however.
(Score: 2) by c0lo on Monday July 28 2014, @07:27AM
If this is a system that can break traffic jams (and make the traffic fluid again), I'd like one: where can I buy it?
https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
(Score: 2) by Phoenix666 on Monday July 28 2014, @01:56PM
One thing that perplexes me about autonomous cars and the drive to take driving out of car travel, is that we already have several means to get from point A to point B without driving. Planes, trains, buses, boats, bikes, walking. If we had a smartphone based way to knit all those together so you can get to your destination fastest, cheapest, and with the most comfort, it would make a heck of a lot more sense and cost a heck of a lot less than replacing every one of millions of cars with autonomous models and working through all the kinks of autonomous cars en masse (keep in mind that Google has been testing autonomous cars in a sea of human-driven models--what happens when you have a sea of autonomous cars with a few human-driven ones that learn how to game the behavior of the autonomous ones?).
Washington DC delenda est.
(Score: 5, Insightful) by Immerman on Sunday July 27 2014, @03:11PM
>They estimate it will be decades until they can perform as well as human drivers in all situations if they ever do at all.
And why should we demand they do? If an AI can perform 10x better in the situations responsible for 90% of potential accidents, but 10x worse in situations responsible 1% of potential accidents, then the AI can be expected to dramatically decrease the accident rate. Yes, some people will be involved in accidents that a human driver might have avoided, but far, far more will be involved in near misses that would have ended tragically with a human at the wheel.
The only reason a rational person might still prefer the human driver is if those 1% of accidents have far worse potential outcomes than the 90% avoided. After all, making me virtually immune from fender benders is nice, but not enough to compensate for the fact that my new car has acquired a taste for playing chicken with locomotives. I knew I should have emptied the ash trays more often...
(Score: 3, Interesting) by opinionated_science on Sunday July 27 2014, @03:27PM
automated vehicles will a great advance for many drivers, especially those who CANNOT drive for some reason.
Google's test cars have shown that for highway driving they are much safer than humans.
For as long as there is a human required to be able to take control, there is probably no difference in liability since only when the human is control is that contract used. There will be general liability on the cars...
Once it become retrofit , it will happen quickly as the insurance companies will make it a lot cheaper when the machine is control...
(Score: 2) by naubol on Sunday July 27 2014, @03:31PM
Possibly not, it may put the insurance companies out of business.
(Score: 2) by Vanderhoth on Sunday July 27 2014, @04:58PM
I doubt that. Insurance, with the exception of a few states as far as I know, is legally required. I imagine it will continue to be legally required in the majority of places. What this will do is increase the profit margin for insurance companies as well as give them another excuse not to pay out. "What's that, YOU were driving? it states in the insurance contract no human driver is to operate the vehicle on interstate highways, secondary roads and roundabouts, sorry claim denied."
Insurance is ridiculous, My wife has been in several accidents her insurance had to pay out for, I've never been in even one accident. A few close calls, but I've never made a claim. The cost to insure me is almost double what it is to insure my wife and in our contract I'm not suppose to drive the car more than 30% of the time. Like the insurance compnay would know how often I actually drive, but if there was a slip of the tong during an investigation I'd be willing to put money on them most certainly denying a claim on those grounds.
"Now we know", "And knowing is half the battle". -G.I. Joooooe
(Score: 3, Funny) by aristarchus on Sunday July 27 2014, @09:25PM
If there are tongs involved, it is more likely an inquisition than an investigation. Just saying.
(Score: -1, Troll) by Anonymous Coward on Sunday July 27 2014, @11:08PM
You seriously believe what Google claims for it's cars without a shred of real proof? Idiot.
(Score: 3, Insightful) by zocalo on Sunday July 27 2014, @04:03PM
As an intellectual exercise on a long drive try counting up the number of scenarios you see where a computer might struggle, or at least require some very specific programming. Stuff like signs being obscured by a truck along side you, electronic signs not working properly, a toll booth being closed, having to leave the main carriage way, and so on. Even on highways the number can rack up pretty fast, and if you are on urban roads, and having tried it on a day spent driving around in inner city I've pretty much given up on the notion of being able to just jump into a car and tell it where I need to go anytime soon.
UNIX? They're not even circumcised! Savages!
(Score: 1) by GWRedDragon on Sunday July 27 2014, @04:41PM
Absolutely right.
The real application for robotic vehicles is in a better version of cruise control, designed only to handle common situations. People underestimate how difficult it is to design in special cases for everything you encounter in the real world. "Intelligent" software tends to be used in bounded scenarios with a fixed number of variables.
Even in aircraft autopilots, the number of things the pilot has to specify is not trivial, and at critical times* the pilot has to constantly monitor every action of the system ready to take over at any moment. Inattentiveness can easily cause, say, a stall that crashes you into the end of the runway at landing**. Driving a car is like that, except all times are critical times.
*And in fact, in theory they are supposed to do this at all times.
**Don't blame the 777 autothrottle design, the whole point is that stuff like this will always come up when there are enough variables at play.
[Insert witty message here]
(Score: 0) by Anonymous Coward on Monday July 28 2014, @12:39AM
cruise control[...]autopilots
...and the folks who don't know the difference.
http://www.snopes.com/autos/techno/cruise.asp [snopes.com]
-- gewg_
(Score: 2) by Immerman on Monday July 28 2014, @03:30PM
Absolutely not. Some sort of hybrid autopilot/cruise control system, that does most the work but requires the driver to still pay attention at all times, is an invitation to accidents for the simple reason that there's no way most people will maintain situational awareness if they're not doing anything with it except in that occasional 30-second occasion once or twice a year.
Now perhaps if it was a "perfect highway autopilot" that could take over completely on the highway and then ensure the driver was conscious and ready to take over before leaving the highway and handing off control it might work. Anything less though is worse than useless. If the computer can't actually drive the car, then there's no point in having it.
(Score: 3, Informative) by frojack on Sunday July 27 2014, @05:01PM
Until or unless Google opens up on the number of situations that drivers have had to take over the controls of their autonomous cars, I don't think you can assume that these situations aren't handled already. They have posted some suggestions that the cars handle everything they have thrown at them todate.
In fact I find it odd that all of the experts in TFA suggest none of this will work for another 15 years at the soonest to be a reputation risking assessment, since there are cars on the road today which have never had an accident, and by all public reports never have to have control taken over by the human driver.
No, you are mistaken. I've always had this sig.
(Score: 0) by Anonymous Coward on Sunday July 27 2014, @05:42PM
> I don't think you can assume that these situations aren't handled already.
Given that some of those examples are spelled out in the article itself, I think you are doing that thing where you assume absence of proof is proof of absence. It's like all those people who voted for Obama because they projected their own hopes and desires onto the "hope" and "change" slogans.
> since there are cars on the road today which have never had an accident, and by all public reports never have to have control taken over by the human driver.
The assumption that those cars have been put through such scenarios rather than having had their routes carefully pre-choosen so as to never encounter them is pollyannaish.
(Score: 3, Insightful) by zocalo on Sunday July 27 2014, @05:58PM
Even so, "15 years" does seem a bit much. That's an awfully long time in software development and I think the experts are way off base on the technical side of the problem, especially given that Google is already running their cars in live environments so must have the bulk of the routine driving down and be working though the exceptional cases as fast as they can. It's also something that would get easier over time; the more autonomous cars you can get on to the road, even with a human ready to step in at the drop of a hat, the less exceptional cases you are going to run into in the first place.
What is going to take the time is definitely the bickering that's going to take place over liabilities if the worst should happen. No matter what the stated operational requirement for a standby human, you just *know* people are going to sit back and let the car get on with it, no matter what's going around them. I can only imagine the kinds of disclaimers that manufacturers are going to use to try and avoid becoming the losing party in an idiotic legal case because some random customer didn't use sufficient common sense.
UNIX? They're not even circumcised! Savages!
(Score: -1, Flamebait) by Anonymous Coward on Sunday July 27 2014, @04:30PM
I've spent my whole life riding bicycles and --over all these years-- have developed a cynical attitude to cars and car culture. Yeah OK they are useful on occasions, and I own one myself (a pickup/ute) but I hardly ever use it on a day-to day-basis and when I do use it I still feel cynical and can't wait to switch it off; I ALWAYS prefer to use my bike in combination with the train system as much as possible...
...I can never get over the fact that an average human of 70kg needs a mass of 1200kg or more to propel themselves around, and then I think of the millions and millions of people all over the world having the same hard-on for doing this, and then I think of planet earth's limited resources, and climate destruction, and human destruction, and so on.
I don't want to live or ride a bike in a city congested with high-velocity automated 1200kg objects---FUCK THAT! When I'm on a bike I feel totally free, I swap lanes and cut through traffic and mount gutters and my path is random and never predictable!
(Score: 4, Insightful) by TrumpetPower! on Sunday July 27 2014, @05:41PM
I bet the first we'll see of wide-scale adoption of autonomous vehicles on the road is with long-haul trucks. They can go from a depot in an industrial district just off the freeway to another depot in an industrial district just off the freeway. They can drive around the clock. Refueling can be done by contracting with one of the truck stop chains to have people on staff. The trucks can then run at optimum fuel economy, and the trucks don't have to sit idle half the day while the drivers sleep and eat. Even an expensive autopilot would thus pay for itself in very short order by doing the work of three humans. And liability insurance? Pennies compared with the health insurance for the drivers.
The biggest obstacle would be the unions, of course. But a "startup" founded by a current executive of one of the shipping companies wouldn't have to worry about that, and could seriously undercut the parent corporation and still make an huge profit. Once the startup grows big enough, the parent corporation merges with it.
I know there were some truckers who read the Green Site. If any all y'all're reading these words, now is a great time to plan your next career. Personally, I don't expect humans to be driving big rigs in significant proportions a decade from now.
Cheers,
b&
All but God can prove this sentence true.
(Score: 2) by dry on Monday July 28 2014, @01:25AM
The last strike here (Port of Vancouver) involved both the union and non-union truckers so your attempt at union bashing is lost. No one likes their wages dropping below the cost of operating or losing their jobs.
(Score: 1, Interesting) by Anonymous Coward on Monday July 28 2014, @02:13AM
your attempt at union bashing
I note that this group of people doesn't suggest a method that improves the lot of working people better than what they dismiss.
It's Like Rooting for the Lions against the Christians [googleusercontent.com] (orig) [dissidentvoice.org]
-- gewg_
(Score: 0) by Anonymous Coward on Monday July 28 2014, @06:19AM
The premise that unions are the working man's best friend (upon which their arguments are based) is no longer true.
Their statement that opponents of the current union regime have no answers is also false. Their are two general answers:
1. Reform the unions. 2. Expand the concept of minimum wage into a set of minimum standards for various occupations (e.g.,
You can't hire a truck driver at minimum wage, you have to use a different wage scale for truck drivers). I'm partial to
the 2nd idea myself.
As it stands, it's pretty ridiculous to say that the union is the "working man's friend" when your occupation
isn't likely to be unionized, you pay taxes to support public employees who *are* unionized, the union won't do
anything to improve your working conditions unless you fork over the dues, and if you are in a union the contracts
have a way of bleeding the companies dry. Yeah, some post-WW2 workers hit the "sweet spot" of not having to get their
skulls bashed by Pinkertons, and not having the plants shut down. It was unsustainable.
(Score: 0) by Anonymous Coward on Monday July 28 2014, @07:25AM
Reform the unions
I'm all for making things better.
I notice you didn't mention any specifics.
Expand the concept of minimum wage
Reagan (and JFK before him) said "A rising tide lifts all boats".
In an economy that isn't completely fucked up, raising the baseline will float everyone upward.
Since it has only been tried on an extremely limited scale recently, there still aren't enough data points to see *exactly* how well it's working, but the early numbers say states that raised the minimum wage have seen an uptick in their economies.
The Chicago School of Economics has been wrong for over 50 years and remains wrong.
ridiculous to say that the union is the "working man's [best] friend"
You left out an important word.
Now, identify an entity that better fits the description.
N.B. I will accept "Worker Self-Directed Enterprises" or "Worker-Owned Cooperatives" as correct.
when your occupation isn't likely to be unionized
Joe Hill, Mother Jones, et al had to fight like hell to get places unionized.
Capitalists murdered swaths of people including infants in their efforts to break strikes.
The current generation seems to think this stuff just falls off a truck.
the union won't do anything to improve your working conditions unless you fork over the dues
Wow, really are you poorly informed.
The SCOTUS ruling from the other week said just the opposite.
You need to switch off Fox so-called News; it's filling your head with nonsense.
union[...]contracts have a way of bleeding the companies dry
That's swill. There's plenty of cash left over after paying a living wage.
Worker productivity has continued to go up since 1968--but wages flatten then.
The only economic problem in American workplaces is the greed of the Capitalists.
You need to switch off Fox so-called News; it's filling your head with nonsense.
not having to get their skulls bashed by Pinkertons
Nobody said it was going to be easy.
Though, with all the templates from decades ago, and existing organizations|infrastructure, it *should* be easier.
...and putting line breaks at random places in sentences is just goofy.
Not everyone has the same screen width and font size as you.
-- gewg_
(Score: 0) by Anonymous Coward on Monday July 28 2014, @12:45PM
That doesn't help the guy with a house at the beach which will be flooded by the rising tide.
(Score: 2, Informative) by FlatPepsi on Monday July 28 2014, @02:17AM
I totally agree. I have long said that long-haul truckers, UPS and FedEx will be the first major adopters of self-driving vehicles. Yes, I agree that driving is not a good career choice for this generation.
I hadn't considered the union implications - which means that the US Postal System will be the last to adopt this system. (Thus adding to their financial woes...) Small delivery shops doing point-to-point or just-in-time delivery will be the first to cause a major upset in the industry.
The upcoming battles of unions vs Johnny Cab will be long & hard fought. The clashes with Uber are laying the groundwork for how those fights will play out.
(Score: 0) by Anonymous Coward on Monday July 28 2014, @03:36AM
My guess is UPS is last. The post office contracts 90% of it out to other companies.
I have dealt with all 3 groups. Surprisingly the guys doing the contracting work for the post office were the first on board with tech. Mostly because they are not union and the people running the trucks can say do it our way or find a new job. They were mostly tired of the lies truckers told them and the 3 log books they kept. UPS was last being a union shop. It took rescuing a few drivers from hijacks to convince them.
Small delivery shops doing point-to-point or just-in-time delivery will be the first to cause a major upset in the industry. The LTL guys are iffy as they have to change what they do quite quickly. But it probably would work as many times the do not do the load/unload and just drive it.
However, if I were a long haul guy doing 5 by 5 I would be worrying about it.
(Score: 1) by CRCulver on Sunday July 27 2014, @06:13PM
(Score: 1) by CRCulver on Sunday July 27 2014, @06:20PM
(Score: 2) by Immerman on Monday July 28 2014, @03:48PM
Actually, I seem to remember hearing something out of Google that obviously they should be liable for any accidents caused by their AIs. Assuming I'm not misremembering, that should make the regulatory environment quite simple - the usual problem with getting regulations established is all the lobbyists trying to make sure their sponsors get to profit obscenely while dodging any responsibility.
(Score: 1) by islisis on Sunday July 27 2014, @07:25PM
If that's the questioned being answered, never seems quite a rational choice. How far were they expected to stretch their imagination?
(Score: 2) by jdccdevel on Monday July 28 2014, @12:40AM
Current car manufacturers can't even get the Traction Control to work properly in Snow and Ice. I will have absolutely ZERO faith in any diverless car until they can at least get that right.
All the driverless cars I've ever heard of use an expensive, and extensive, set of sensors to get the information they need about current conditions. What happens when they ice up or get covered in hard packed snow? Until they have cars that only need 2 cameras and an accelerometer to drive, I highly doubt we'll ever see them in the snow. There's also all the feedback from the road that an experienced driver gets, that an AI would probably just ignore (for example, the sound of the tires on ice vs bare pavement, which is very useful to detect black ice.) Not to mention, how will it react when a residential street has knee deep snow drifts with ice underneath after a storm?
As zocalo said above, there are just too many unknowns, and snow and ice complicates the situation exponentially.
For that reason, I don't see driverless cars being a viable alternative in colder climates without some significant advances in AI.
Still, something like an autopilot would be nice for those longer drives in the summer :-)
(Score: 2) by quacking duck on Monday July 28 2014, @03:28AM
I'd bet on it doing what a human would do if our primary driving sensors (eyes) can no longer make out our surroundings well: turn hazard lights on and slow down, or even perform the ultimate failsafe: pull over and stop.
But overall I agree with you, it doesn't seem like Google cars have been properly tested in winter, and probably won't until years after certification on dry/rainy conditions. Will it know where the lanes are if it can't see them? Will it ignore lane constraints when a more travelled path straddles lanes, like how our 4-lane (each way) freeway sometimes gets treated by drivers as 3 extra-wide lanes (safety margin for minor, recoverable slips/skids) during heavy snowstorms in between snowplough passes? The latter might be a moot point if *all* cars are driverless, but programmers and the AIs must assume for a long while that humans will be driving most of the cars around them.
(Score: 2, Interesting) by Anonymous Coward on Monday July 28 2014, @03:56AM
> Will it know where the lanes are if it can't see them?
Yes. That's one of the things that is crazy about google's current system - they map out everything in extreme detail, like the height and position of the curbs down to the inch or less. They surely know exactly where the lanes are and have augmented their in-vehicle GPS to use that precision.
(Score: 3, Interesting) by jdccdevel on Monday July 28 2014, @02:12PM
Height and position data don't work very well when everything is covered in blowing snow, and your mapping laser is bouncing back from snowflakes falling from the sky.
(Score: 4, Funny) by LaminatorX on Monday July 28 2014, @03:05AM
...just get this to work by the time I'm to old to drive.