Self-Driving Hype Doesn’t Reflect Reality

[cpzilliacus]

Wall Street Journal: Self-Driving Hype Doesn't Reflect Reality - Auto makers promise "˜autonomous' technology, but it will be years before we get self-driving cars that can go anywhere

To judge by recent claims, "fully autonomous"  self-driving technology is just around the corner. Uber Technologies Inc. is offering Pittsburgh residents rides in autonomous Ford Fusions. Ford Motor Co. , BMW AG , Volvo Car Corp. and Lyft Inc. say they will produce fully autonomous vehicles by 2021 or sooner. Tesla Motors Inc. Chief Executive Elon Musk, rarely topped in hyperbole, says the technology will be here within 24 months.

To many industry insiders, these claims are largely hype. They're not false, but they abuse the terms "autonomous vehicle"  and "self-driving,"  which evoke images of hopping into a car, entering a destination and disappearing into sleep, food or our phones.

That is not what we're going to get by 2021. It won't happen for a long time, maybe decades.

[jeffandnicole]

Wall Street Journal: Self-Driving Hype Doesn’t Reflect Reality - Auto makers promise ‘autonomous’ technology, but it will be years before we get self-driving cars that can go anywhere

To judge by recent claims, “fully autonomous” self-driving technology is just around the corner. Uber Technologies Inc. is offering Pittsburgh residents rides in autonomous Ford Fusions. Ford Motor Co. , BMW AG , Volvo Car Corp. and Lyft Inc. say they will produce fully autonomous vehicles by 2021 or sooner. Tesla Motors Inc. Chief Executive Elon Musk, rarely topped in hyperbole, says the technology will be here within 24 months.

To many industry insiders, these claims are largely hype. They’re not false, but they abuse the terms “autonomous vehicle” and “self-driving,” which evoke images of hopping into a car, entering a destination and disappearing into sleep, food or our phones.

That is not what we’re going to get by 2021. It won’t happen for a long time, maybe decades.

This seems to be more of the news media playing good-cop, bad-cop.  One reporter hypes up the technology, then the guy in the next cube talks about how the auto manufactures are over-hyping it.

I've never read where any car company has said we'll all be driving self-driving cars in just a few years.  It's the people that erroneously use that information to think it's going to happen.  I've seen it on these forums quite often.  The expectations are extremely unrealistic in many cases when some people say we'll all be driving self-driving cars in 5 or 10 years...anyone buying a car today is probably still going to be driving it in that timeframe.

[Max Rockatansky]

Come get me when they start using hover conversions, we've been promised a flying car a lot longer than an autonomous one.

[kurumi]

There may be certain cases that never get addressed (oil fields, tundra, mountain routes, etc.) because the payoff doesn't match the cost and effort. I could see a lot of constrained use cases being solved enough so that we can say autonomous driving has arrived -- possibly farming, urban areas, commute routes; even specialized things like school buses* with no blind spots at all (but a chaperone on board to keep order)

* for parents who drive their kids to school every day in the minivan (and complain about the traffic), school buses were big yellow mass transity things that transported large groups of kids to and from school

[empirestate]

Well, in a way it's true: the actual self-driving technology–the fully automated future most people envision–is already in existence, or nearly so. The hold up, and it's a huge one, is the (supposedly temporary) integration of self-driving and human-driven vehicles.

Were it not for this, we certainly could have self-driving technology implemented by 2021. But because we first have to figure out the crossover period, yeah, it'll take a while–probably not within any of our lifetimes. (Well, unless some autocratic government somewhere decides unilaterally to outlaw human-driven vehicles...)


iPhone

[vdeane]

Looks like they've been predicting self-driving cars for a LONG time: http://spectrum.ieee.org/geek-life/history/selfdriving-cars-were-just-around-the-cornerin-1960

[empirestate]

Looks like they've been predicting self-driving cars for a LONG time: http://spectrum.ieee.org/geek-life/history/selfdriving-cars-were-just-around-the-cornerin-1960

Yep...even back then, the technology was perfectly feasible–except, of course, for the obstacle of human-driven cars. Had self-driving cars been the first type to be invented, we'd have been a driverless society for decades now.


iPhone

[Max Rockatansky]

Well, in a way it's true: the actual self-driving technology–the fully automated future most people envision–is already in existence, or nearly so. The hold up, and it's a huge one, is the (supposedly temporary) integration of self-driving and human-driven vehicles.

Were it not for this, we certainly could have self-driving technology implemented by 2021. But because we first have to figure out the crossover period, yeah, it'll take a while–probably not within any of our lifetimes. (Well, unless some autocratic government somewhere decides unilaterally to outlaw human-driven vehicles...)


iPhone

There is no way the self driving technology is ready for anything but mainstream roadways.  My GPS can't even keep with some of the mountain roads that are on this side of the country, I couldn't fathom how much extra work it would take to get things going on secondary roads.  Besides....flying cars have had the same problem too, the human factor.  That human factor isn't going away any time soon no matter how good the technology gets.  There are still people slogging around in Geo Metros....they ain't going to be willing nor be able to pay for the premium on technology.

[empirestate]

Well, in a way it's true: the actual self-driving technology–the fully automated future most people envision–is already in existence, or nearly so. The hold up, and it's a huge one, is the (supposedly temporary) integration of self-driving and human-driven vehicles.

Were it not for this, we certainly could have self-driving technology implemented by 2021. But because we first have to figure out the crossover period, yeah, it'll take a while–probably not within any of our lifetimes. (Well, unless some autocratic government somewhere decides unilaterally to outlaw human-driven vehicles...)


iPhone

There is no way the self driving technology is ready for anything but mainstream roadways.  My GPS can't even keep with some of the mountain roads that are on this side of the country, I couldn't fathom how much extra work it would take to get things going on secondary roads.

Do you mean for some reason other than the human factor? What do you see as the major obstacle?



iPhone

[Max Rockatansky]

Well, in a way it's true: the actual self-driving technology–the fully automated future most people envision–is already in existence, or nearly so. The hold up, and it's a huge one, is the (supposedly temporary) integration of self-driving and human-driven vehicles.

Were it not for this, we certainly could have self-driving technology implemented by 2021. But because we first have to figure out the crossover period, yeah, it'll take a while–probably not within any of our lifetimes. (Well, unless some autocratic government somewhere decides unilaterally to outlaw human-driven vehicles...)


iPhone

There is no way the self driving technology is ready for anything but mainstream roadways.  My GPS can't even keep with some of the mountain roads that are on this side of the country, I couldn't fathom how much extra work it would take to get things going on secondary roads.

Do you mean for some reason other than the human factor? What do you see as the major obstacle?



iPhone

Human factor goes beyond just drivers on the road, it goes into economics too.  Look at the progression of hybrid and electric cars.  The price of those vehicles is based on supply and demand.  R&D demands that the initial cost is very high and will come down as time progresses due to increased demand.  So basically you have a high entry level price that only that people with wealth can afford as a luxury.  Considering that most people don't fall under that definition that could be a slow jog out of the high price points on the progression down to something that can be affordable to the average car buyer.  Even then you'll have people who still won't be able to afford the technology due to economics not enabling them to buy new cars.  More so you'll always have the people who don't want to change or even a large glut of the population that still hangs onto their current vehicle for a 7-10 years. 

So basically economics by itself is going to force a progression of integration of self driving technology.  That is going to take decades to achieve on any sort of large scale.  Look at the progression of hybrid cars, in the most optimistic scenario the progression could be something like that.  On the flip side....it could be a complete disaster and be a total market failure in general that never really catches on. 

But then you still have the obstacle of infrastructure.  I think there is an assumption that self-driving cars will fare well on ALL types of roads.  How well will the technology hold up on a road that's two-lanes?....one lane?...gravel?...has an bunch of extreme grades?...even dirt?  I would speculate that the technology is going to see an overwhelming entry point from the urban market simply due to the nature of the driving and the ease those roadways will present to integrating the technology. 

Ultimately I see the technology being a thing in some small capacity in the 2020s but from there it will remain to be seen how viable it really can become. 

Also, something that is certainly a human factor that really hasn't been discussed is the nature to distrust a new technology.  Something could have proven reliability to an extremely high degree but it is human nature not to trust something new until it is around for a long time.  That kind of mindset isn't just part of the general populace, it plays a factor in media and government in almost everything.  Don't forget there was a large segment of people who were convinced video games, mobile phones, and the internet were the Devil's work. 

[empirestate]

There is no way the self driving technology is ready for anything but mainstream roadways.  My GPS can't even keep with some of the mountain roads that are on this side of the country, I couldn't fathom how much extra work it would take to get things going on secondary roads.

Do you mean for some reason other than the human factor? What do you see as the major obstacle?

Human factor goes beyond just drivers on the road, it goes into economics too.  Look at the progression of hybrid and electric cars.  The price of those vehicles is based on supply and demand.  R&D demands that the initial cost is very high and will come down as time progresses due to increased demand.  So basically you have a high entry level price that only that people with wealth can afford as a luxury.  Considering that most people don't fall under that definition that could be a slow jog out of the high price points on the progression down to something that can be affordable to the average car buyer.

Also true of regular cars when they were first introduced. Had self-driving cars been the first type to be offered, the same would surely have happened.

But then you still have the obstacle of infrastructure.  I think there is an assumption that self-driving cars will fare well on ALL types of roads.  How well will the technology hold up on a road that's two-lanes?....one lane?...gravel?...has an bunch of extreme grades?...even dirt?  I would speculate that the technology is going to see an overwhelming entry point from the urban market simply due to the nature of the driving and the ease those roadways will present to integrating the technology.

Infrastructure would have developed quite differently without human-driven cars. Roads would have been designed to accommodate the existing technology, which in this hypothesis is self-driving vehicles. So that, too, isn't an obstacle to the technology, other than the fact that it's the current status quo.

Also, something that is certainly a human factor that really hasn't been discussed is the nature to distrust a new technology.  Something could have proven reliability to an extremely high degree but it is human nature not to trust something new until it is around for a long time.

Yes; that's part of the reason it won't happen right away. But those people will naturally diminish in number, and that will probably happen sooner than the technological integration is complete; or more likely, the two eventualities will occur hand-in-hand.

So yeah, as we both seem to agree, the only thing holding up the technology is the fact that there's another technology already in place that must be overcome, or at least temporarily accommodated. Without human-driven cars, self-driving cars could be deployed instantly. The infrastructure would be tailored to them instead of to the cars we have now, and there wouldn't be an inherent distrust of automated driving since the populace wouldn't have any experience to the contrary.

[Max Rockatansky]

Right but let's be real here, the computing power necessary to automate really anything of substance didn't begin to emerge until after the World Wars.  That's a hell of hypothetical situation to even envision a scenario where computing technology could have advanced while transportation technology somehow remained train and wagon based.  One could argue that advances in automotive technology was one of the biggest pushes of over all technological advancements through the 20th century.  It would be interesting though to see a write up on a hypothetical 2200 with steam powered computers and pneumatic tube carts pushing people through Art Deco sky scrapper laden city scapes. 

Basically if the automated technology can become affordable enough, have enough reliability, and is something that makes life more easier for a driver, then it will likely will succeed.  Don't forget there was a time when steam and electric power were in competition with the early automobile.  The internal combustion engine won out in the end because it was cheap, had acceptable reliability, and definitely made things way easier for the driver.  Something fun to look up is Jay Leno going through the start up procedure on several variations of steam cars.  Even the hand crank procedure used on internal combustion engines at the time was infinitely easier by comparison.  The only reason steam stuck around as long as it did was because it was at it's technological pinnacle from the previous century.

But as far as road development goes, even that was a rough change when cars became an increasing presence on the road.  Roads were basically just dirt paths that would had be best considering a walking trail by modern standards.  That was a hell of a leap just to get something like grading progressing to dirt and even asphalt.  Not to mention even organizing roads into actual routes like the Auto Trails or US Route system....they all took time even with the quick rise of the automobile.  So in turn automation is going to have share the road with human controlled vehicles....for what I would suspect would be a large portion of the rest of my life if not likely all of it.

One piece of good news though if automation ever became the sole technology on the road then flying cars suddenly become realistic again.  Funny to think there was actually a time when people really thought a flying car was viable to operate for the everyday person much less that some people though helicopters were also.

[empirestate]

Right but let's be real here...

Well, we can do that, of course, but then we're not really discussing my assertion anymore. My point is that, had we skipped over human-driven cars and gone right to the self-driving kind, we would have easily been able to roll out the technology by now, or very soon. Obviously, that didn't happen (for the reasons that come up when we start being real), but that's why it's a hypothetical.

I'll put it this way: assume that we could realistically eliminate all human-driven vehicles tomorrow. Could we have a fully-automated fleet of vehicles on the roads by 2021?

[Max Rockatansky]

Well no...say if somehow all the self driving cars disappeared overnight the economic and societal costs would be a disaster.  I'm assuming that self driving trucks would go as well or pretty much anything that drives on a road.  Basically freight and goods would be limited back to ships, trains, and in some cases cargo planes.  Given that we have billions more people today than back at the start of the 20th century it would be basically anarchy overnight.  Not to mention people would be back to moving by foot, bike, and even train....which would be substantially harder to do opposed 100 years ago.

But that's looking at something literally which wouldn't ever actually happen.    The problem with that hypothetical is this....basically if somehow millions of people could agree on anything, engineering projects would get done hundreds of times faster than they would today.  The problem is that you have interest groups, environmental regulations, businesses, ect that all have a hand in things that are going to prevent any infrastructure development from becoming a fast reality.  In the case of self driving cars it's main enemy right now is cost and reliability....somehow it's managed to dodge any really significant governmental attempts to slow it downs....but that's so far.  That US 27 crash with the Tesla pretty much showed that one incident can get things turning bad in the other direction though...now imagine something similar happens half a dozen times?...who's stepping in to say "no" then?

Anyways, thought this was interesting since the "no cars" hypothetical came up:

[empirestate]

Hmm, yeah, it does seem as though you're thinking about this too literally and thus missing my underlying point.

I'm saying that the reason self-driving cars haven't been fully deployed yet, and won't be in the immediate future, is because we first have to figure out how to integrate them with human-driven cars. Were that not the case–and perhaps I over-complicated things by trying imagine a world where it isn't–the existing technology would be easily sufficient to be put into use immediately. Do you think that's not so? You say cost and reliability are the biggest obstacle, but isn't that because of the costs and reliability issues associated with incorporating automated vehicles into a non-automated environment?


iPhone

[SP Cook]

CGP Grey's latest video is on this idea of self-driving cars.  I normally like his work very much, but the idea that we will ever have a car that is truly self-driving is science fiction.  And not just because of this "human factor" .  Even if every car could be made self-driving overnight, it just will not work.

Ever had a car lose a timing belt?  Ever had a computer freeze up?  We are talking 100000000 decisions every second, and if it fails one out of 20 million times you are dead.  Within a few months everyone's short straw would come up. 

[Max Rockatansky]

CGP Grey's latest video is on this idea of self-driving cars.  I normally like his work very much, but the idea that we will ever have a car that is truly self-driving is science fiction.  And not just because of this "human factor" .  Even if every car could be made self-driving overnight, it just will not work.

Ever had a car lose a timing belt?  Ever had a computer freeze up?  We are talking 100000000 decisions every second, and if it fails one out of 20 million times you are dead.  Within a few months everyone's short straw would come up.

I recall someone once saying on Dateline or one or those shows something to the effect of "if your car malfunctioned like your computer did you wouldn't drive it."   Basically it was something in reference to computer reliability had a long way to go despite advances.  I still get my infotainment system to sync to the correct time on my 2014 Sonic or will lose Bluetooth randomly.  Those are all factors in what I was talking about with reliability above.  Besides there is no way that there will ever be a car that doesn't require maintenance nor drivers who neglect them.

But in reference to above, the hypothetical above was more or less a poor attempt humor. 

[jeffandnicole]

I'm saying that the reason self-driving cars haven't been fully deployed yet, and won't be in the immediate future, is because we first have to figure out how to integrate them with human-driven cars.

Isn't that intragration occurring today, with Google, Uber and other companies having self-driving cars out there on the same roadways as human-driven vehicles?

Any product that comes to market tends to have a long R&D factor.  TVs, electronics, appliances, etc with features new on the market today weren't thought of in August.  They've been research and testing those products for years in many cases.  Cars are no different, with the exception that the testing phase is occurring alongside of us on the roads, not in a warehouse seen by few people that are told to keep their mouths shut.

Even cars - things that are becoming commonplace today, such as rearview cameras, were the stuff that was found on very high-end, expensive cars not too long ago.  Video cameras themselves have been around for decades, of course.  Cameras that were small enough to fit into very small spaces and wired to batteries are much more recent.  No one expected to be using a phone as their primary means of taking pictures and shooting video, for instance.  Hell, no one from the 1980's and prior realistically expected for their primary phone to be portable.  Even in the 90's, there were severe restrictions on how portable phones really were.

CGP Grey's latest video is on this idea of self-driving cars.  I normally like his work very much, but the idea that we will ever have a car that is truly self-driving is science fiction.  And not just because of this "human factor" .  Even if every car could be made self-driving overnight, it just will not work.

Ever had a car lose a timing belt?  Ever had a computer freeze up?  We are talking 100000000 decisions every second, and if it fails one out of 20 million times you are dead.  Within a few months everyone's short straw would come up.

I recall someone once saying on Dateline or one or those shows something to the effect of "if your car malfunctioned like your computer did you wouldn't drive it."   Basically it was something in reference to computer reliability had a long way to go despite advances.

Traffic lights are computerized, yet no one seems faxed that a few lines of programming is what's preventing them from displaying green in every direction.  If this happened once in a billion times, people would be complaining up the wazoo about it.

Most people associate computers with what's on their desk or lap.  You are paying for Chinese-made cheap crap, and the life-expectancy of it isn't much more than a few years; several years if you know what you're doing.  You can expect the computer in your car to cost a bit more...and be a bit more reliable...than someone most people use primarily to send out pictures of their dog eating from a dinner plate at a kitchen table.

[SP Cook]

Most traffic lights are not really "computerized", but why knitpick.  Self-driving cars are making 100s of 1000s of decisions every second.  One in billion means it only kills somebody every few minutes. 

The complexity of driving is vastly beyond the ability of a computer, no matter where it is made.  Self-driving cars are the stuff of science fiction, like the old Popular Mechanics mag that had everybody in a flying car by 10 years from now for 60 years or so. 

[empirestate]

I recall someone once saying on Dateline or one or those shows something to the effect of "if your car malfunctioned like your computer did you wouldn't drive it."   Basically it was something in reference to computer reliability had a long way to go despite advances.  I still get my infotainment system to sync to the correct time on my 2014 Sonic or will lose Bluetooth randomly.  Those are all factors in what I was talking about with reliability above.  Besides there is no way that there will ever be a car that doesn't require maintenance nor drivers who neglect them.

No, and I don't predict that self-driving cars will be any different. But just because they won't be error-free or maintenance-free doesn't mean they won't (or can't) be deployed. I can't think of any other technology that's in widespread use and is error-free. But computers can be made somewhat more failsafe and reliable than the typical consumer product, and so they will be with these vehicles–they'll simply have to for it work.

I'm saying that the reason self-driving cars haven't been fully deployed yet, and won't be in the immediate future, is because we first have to figure out how to integrate them with human-driven cars.

Isn't that intragration occurring today, with Google, Uber and other companies having self-driving cars out there on the same roadways as human-driven vehicles?

Yes, in a very limited way.

Any product that comes to market tends to have a long R&D factor.  TVs, electronics, appliances, etc with features new on the market today weren't thought of in August.  They've been research and testing those products for years in many cases.  Cars are no different, with the exception that the testing phase is occurring alongside of us on the roads, not in a warehouse seen by few people that are told to keep their mouths shut.

Right. Leaving aside the "alongside of us" factor, I believe the R&D for self-driving vehicles is at a phase where release of the technology is imminently feasible.

Most people associate computers with what's on their desk or lap.  You are paying for Chinese-made cheap crap, and the life-expectancy of it isn't much more than a few years; several years if you know what you're doing.  You can expect the computer in your car to cost a bit more...and be a bit more reliable...than someone most people use primarily to send out pictures of their dog eating from a dinner plate at a kitchen table.

I think car computers already are more reliable, but it's largely because they have more limited functions that your general-purpose laptop or smartphone. They don't need to be bothered by provided a user interface, or providing a comprehensive range of functions like graphics, audio, networking and so forth. Self-driving cars will need networking capability, of course, but processing power will still be far more dedicated to the specific function of driving the car and communicating with other traffic.