Unexpected Consequences of Self-Driving Cars | Hacker News Reader

Unexpected Consequences of Self-Driving Cars

rodneybrooks.com

183 pointsghosh209 comments

Whenever people jump up enthusiastically about autonomous cars being right around the corner, I remember taking a cab in Lima, Peru, in the middle of some horrendously chaotic traffic, where drivers were navigating with little regard for lanes, negotiation for merging is done eye-to-eye, and impatient pedestrians would throw themselves into the street, tired of waiting for minutes to cross and trusting that the driver ahead wasn't suffering from some murderous rage that day.

I also remember run-down suburbs where massive potholes and ditches make navigations of certain streets a puzzle by itself, compounded by garbage trucks ahead just doing their thing, improperly placed signage ahead, and those badly-designed intersections where you can totally count on drivers misjudging distances.

As the designers move away from the relatively trivial confines of well-maintained urban spaces and into different countries and cultures, I seriously question their ability to keep things reasonable for the rest of the human drivers. So far nothing in the near-future horizons of this technology indicates that the software would be able to navigate in many of these adverse situations without incurring huge time penalties or causing unacceptable delays.

I really do think that a lot of the hype happens to be because certain actors (Uber comes to mind as the most significant) have basically placed all their bets in these edge cases being small enough that a small number of human drivers should suffice to get around them. I remain skeptical.

Sure, those scenarios you pointed out are difficult, but what about highway driving?

Replacing highway driving with autonomous vehicles is still completely game changing for the transportation industry. It means that 90% of truck drivers are out of a job. (Truck drivers of the future will just handle the last mile).

That is certainly something that is going to be disruptive.

I'd love this as a private vehicle too. Imagine if you could turn a 12 hour drive into an hour on each end (from highway exists to local destination) with the ability to just relax or sleep in the middle? I don't think I'd fly anywhere within 10 hours of me, because an overnight drive suddenly isn't a burden!

Yes. I don't know why more time isn't spent on this aspect. It would totally start competing with air travel at that point for sure. I travel all the time for work between cities 4 hours by car or 45 minutes by air. So car is slightly longer all in given fixed costs at airport on either end. But I would do it ALL the time if I didn't have to drive it myself!

well, there can be 'autonomous-car-friendly-zones' which would automatically entail ridiculously high rent prices : )

I'm not sure those "well-maintained urban spaces" are as simple as you seem to be implying. They have plenty of quickly-changing construction in the middle of active roads, out of date / invisible signs, and far more crazy drivers and pedestrians to interact with per day than in Middleonowhere Pennciltucky.

As to garbage trucks: ever seen people pull U-turns randomly in the middle of the road in a busy city, or dive across 4 lanes to make an exit? Garbage trucks seem simple by comparison, and at significantly lower speeds.

Is there any reason we can't just rely on autonomous cars in areas where these edge cases don't happen? Then as the companies behind this technology make more money, they can start working out the edge cases, but before then, even getting it to work in a given city seems like a huge win.

>Is there any reason we can't just rely on autonomous cars in areas where these edge cases don't happen?

Because there are still ripple effects to doing this affecting areas with edge cases.

So let's say you've reduced the number of manual-driven cars. That has implications for:

* Economies of scale. Autonomous cars will likely be of a different design than manual ones so the price of a manual-drive will likely increase- where will the increased money to buy them come from?

* Insurance costs. Drivers in less-dense cities are currently safer than the national average: what does the risk and cost for a safer driver become now that the bad urban driver benchmark disappears?

* Government regulation and other politicking. What happens to everyone else in, say, California or New York State or a European country if an urban-focused legislature creates laws restricting ownership/use of a manual-driven car in some way (anything from a tax to an outright ban)?

Whether these implications are small or large (or whether they cancel each other) isn't adequately answered right now; but I suspect that at least one (or all 3) will turn out to be a major hindrance to autonomous car deployment. Or they don't, and some section of the population gets screwed anyway.

>what does the risk and cost for a safer driver become now that the bad urban driver benchmark disappears?

I am not sure if an autonomous car will be safer. I mean, can they even do panic breaking right now (For example, if someone or something jumps in front of the car?). You know, safety is often not about having the fastest reaction time, but also about having good anticipation that any decent human driver will develop in a short time....

I think any hopes that cars will be better than humans at it is quite naive...

I think you'd be surprised at how many people "test" Google's self driving cars.

Pretty sure they can handle it if someone jumps in front of them simply because people play chicken with them.

I searched, but couldn't find one video that demonstrates this. I mean, I was not expecting someone to throw themselves infront of a car and video it hoping it would stop in time. but I thought someone might have tested this using a dummy...Found nothing yet...

> I am not sure if an autonomous car will be safer. I mean, can they even do panic breaking right now (For example, if someone or something jumps in front of the car?). You know, safety is often not about having the fastest reaction time, but also about having good anticipation that any decent human driver will develop in a short time.... I think any hopes that cars will be better than humans at it is quite naive...

Even if we assume that cars won't be as good at stopping from people jumping in front of them, I would bet the number of accidents that happen from something like that are significantly smaller than the number of accidents from human error like texting while driving, being drunk, or just not paying attention due to fatigue or whatever else.

Not to mention, how can you really even tell the difference between someone who's walking towards your car but will stop and someone who isn't? Not even humans can do that since we can't read minds.

Luckily there is already seemingly an oversupply of human-drivable cars (in both automatic and standard, aka "manual" transmission) today. The price of a working car has dropped so low that it is barely above the scrap price, and people turn their nose up at such perfectly working cars. The human-driving enthusiast will have plenty of material available for decades, and this is before even factoring in the great reduction in competition said enthusiast will face in the marketplace as most people begin to demand self-driving cars and turn their nose up at a 2016 human-driven camry without self-driving the same way they turn their noses up today at a 1999 camry that lacks features they think so essential like automated lane maintenance, push button start, heated seats, or whatever half ass features car manufacturers keep coming up with in order to con $20,000 out of the gullible.

> * Economies of scale. Autonomous cars will likely be of a different design than manual ones so the price of a manual-drive will likely increase- where will the increased money to buy them come from?

I'm not sure how big of a threat that is. The design ought not need to be so different that the same production lines can't easily make both versions. Relative to many newer cars (anything drive-by-wire), autonomy probably only requires the addition of sensors and a computer. Even getting rid of the steering wheel and all driving controls could probably be done modularly on a car model that's offered in both self-driving and traditional options.

> Economies of scale. Autonomous cars will likely be of a different design than manual ones so the price of a manual-drive will likely increase- where will the increased money to buy them come from?

The people who continue to buy manually driven cars, I presume.

> Insurance costs. Drivers in less-dense cities are currently safer than the national average: what does the risk and cost for a safer driver become now that the bad urban driver benchmark disappears?

They will be benchmarked against more dangerous suburban drivers (e.g. Marylanders) instead.

If twenty percent of the population controls eighty percent of the wealth, why should you ever care about the remaining eighty percent of the population? They can't pay you, so your success or failure in their markets need not determine your success or failure in general.

Just like with electricity, the Internet, and air travel, the poor will be last. And a lot of people can do very well for themselves by putting the poor last.

I'm not really sure what your point is here, other than "global inequality sucks"?

Should we arbitrarily stop development of self-driving cars until poorer countries' infrastructure can support them? How would that help anyone?

There was no implied position on global inequality being good or bad - just a reality check. Peru will be irrelevant to the viability of the project in the short term

This is the same case in India, no way an AI is able to drive on Indian roads before they pull the plug on the project altogether. I feel that self driving cars will become fancy toys for the ultra rich. Our roads are not predictable, they are good in a few places, but mostly, they are exceedingly bad. For a human itself they are impossible to travel let aside have a robot driving it. There are crazy pedestrians who don't move out of the way, they walk in the middle of the f* road, crazy people try to squeeze through the gap between your vehicle and the next one even if there is no discernible gap.

In India, in majority of roads you don't just have to look sideways, you also have to see if some stupid idiot is driving at the speeds of higher than 60 while you cross an intersection. Plus in places of high traffic the self driving car would be deadlocked because here, nobody gives anyone a lead, you have to "earn" the lead by stopping others, no way a robot is going to do that.

Good point about India! I wonder how a self-driving car would handle someone coming towards you on the wrong side of a divided highway, or know how to get a cow out of the middle of a crowded street..

Like the AI in any driving game with similar conditions?

It's still driving, just with different degrees of freedom.

It might be various degrees of freedom but in India there are a few people who follow traffic rules, everyone acts as if they own the street, they drive on the wrong side of the highway, pedestrians cross the street even while being fully aware that there is a vehicle coming at 150+ kmph speeds etc. It is crazy and chaotic, plus forget the highways, in smaller alleys the road is so narrow that can barely accomodate a bus, if the AI gets stuck in such a deadlock it'll be stuck forever, because nobody gives you side, you have to force someone to stop

>if the AI gets stuck in such a deadlock it'll be stuck forever, because nobody gives you side, you have to force someone to stop

Well, a sufficiently advanced AI would be able to "force someone to stop".

Perhaps the tolerable accidents rate will be increased for India.

Or, even simpler, India wont get self-driving cars -- the same way countries in the developing world don't get lots of modern services and/or conveniences until they have the infrastructure for them.

>Or, even simpler, India wont get self-driving cars -- the same way countries in the developing world don't get lots of modern services and/or conveniences until they have the infrastructure for them.

Precisely my point. But one has to note that the TATA group is going to start testing self driving cars in india.

Not necessarily. Sometimes lack of infrastructure fast forwards the technology to next iteration instead of holding it back. For example in India, majority of people's first phone was a mobile phone & they never had landlines & first experience of internet was via mobile phones instead of dial-up or fixed line. Now with a company called Jio providing free 4G internet lot of people are directly going to experience 4G from 2G without transitioning to 3G since the cost is going down. Similar shift might happen in payments with models like UPI (unified-payment-interface) where many people will do digital/cashless payments without ever doing netbanking or owning credit cards.

I expect a truly intelligent AI to navigate around such issues in fact, such issues will provide playing field to create robust AI since it will learn to handle & respond to unpredictability.

I agree with the point of robust AI, also with the point of mobile phones without landlines, mcommerce before credit card/debit card etc. Let's see how things go. btw I live in India :-)

When Google maps came, we said the same. It won't work in India. There's no order in door numbering or Street names. They have to use landmarks and all that. But that's what Google did. In the driving directions in India, you can see lot of landmark references. We always underestimate what a tech can do. I am hoping we are wrong on this one too.

I was a kid when Google maps came, so I won't be able to comment on that. Self driving cars will work, nobody is doubting it, what I am doubting is the extent of self driving car, yes, in US like streets where everyone follows traffic rules it'll work without a person in driver seat, in India it'll be difficult for a truly self driving car like Uber is betting

Google Maps did not work in India for a long, long time.

Google Maps launched in 2005; if it works in India now, then it took at most 11 years more to get it working in India than in the US. That's not that long. Self-driving cars are more complex, but it'll happen eventually.

I remember when I first went to India and saw the traffic. My family would say that traffic in the US and Europe was more "streamlined."

You make some good points, as does this article. In the article though, I think at least some of these points will just involve more visual indicators on cars themselves that people will need to get use to (and yes, I do realize that won't solve many of the other class issues the author raises).

I have my own concerns. So much money, even tax money, is being poured into this tech that I don't see as viable for at least 7 ~ 10 more years. We could be building automated trains, which do exist, are viable and can operate even on late night runs without drivers or strikes. All of Singapore's lines are automated and several of London's are; both which have trains arriving on platforms at less than two minute intervals at peak hours.

Self driving cars also will not solve the capacity problem at all:

http://penguindreams.org/blog/self-driving-cars-will-not-sol...

You'd have to have self-driving car only motoways to even being to approach the theoretical maximums of car capacity, and it still pails in comparison to train capacity, or even bus capacity for that matter.

The other thing I don't think people realize is that we (most likely) won't be owning self driving vehicles. I feel like if States get their acts together, we're going to need FAA style regulations. Cars need to be inspected at regular intervals to ensure all the sensors are functioning. Software updates need to be audited, and safety algorithms will need to be shared. I mean, do you want to live in a world where Audi claims their driving algorithms are safer than Nissan's, or Honda charging extra for more comprehensive sensor packages?

I don't think we can have a self driving where people own vehicles, which goes well against the OSS creed of "Would you buy a car with the hood welded shut?"

"Would you buy a car with the hood welded shut?"

"The new, maintenance-free iCar! So reliable, we welded the hood shut!"

Add a better copywriter, and an actual product, and I think a lot of people would buy that. Same with car-as-service. A lot of people don't actually want to own a car, they just want the transport capabilities and convenience.

"Would you buy a car with the hood welded shut?"

Sure why not? When do most people open the hood on their car? To top up fluids and, occasionally, to jump start the battery. Both of those things could be done without opening the hood with a slight redesign. Modern cars are both so reliable and so complex that there is hardly anything under the hood that most people either can or have to do, and this is only getting more true for electric cars.

I think self-driving cars can make trains much more viable by providing last-mile support to extend the practical range of stations. With self-driving shuttle taxis, a single station could serve a vast sprawling suburb easily.

And a switch to a transport-as-a-service model can make the advantages of the train much more apparent. Open Google Maps and ask for directions: Do you want to take a single 1 hour drive, or a 30min train ride with 5min connections at each end that costs less?

And then when people in the office hear about how your commute is 20mins shorter thanks to the new train line, maybe they want one too...

>I think self-driving cars can make trains much more viable by providing last-mile support to extend the practical range of stations. With self-driving shuttle taxis, a single station could serve a vast sprawling suburb easily.

Why then, is there no similar service with human drivers right now?

I'd imagine primarily because it's too expensive with human drivers, sufficiently so that no one has considered this a viable market. Not to mention we've only recently started getting digitally connected taxi services at all, and they're still experimenting with their models. I've certainly heard people talking about taking Uber to their nearest station, so I can imagine a more organised timetable synced service being a available in future.

And I think Google Maps starting to integrate things like Uber is a sign of a more integrated market for transport. If you can say "directions to work" and get a one button option to book the whole 3 part journey, alongside the other options, it suddenly becomes a lot more practical.

> I'd imagine primarily because it's too expensive with human drivers, sufficiently so that no one has considered this a viable market.

Why do you think removing human drivers will make enough of a difference? From the numbers I have seen (I can look them up when I'm not on my phone) the driver is only about half of the non-tip cost of a hired car. Do you think cutting the cost in half is sufficient?

The other factor here is that traditionally taxis charge a minimum fare for short distances making them much less cost effective.

If you charge close to linear pricing over short distances, I think it could be very affordable.

And another advantage is that the service doesn't have to be immediately popular to be cost effective, since you don't need to pay drivers to sit around to provide availability.

The practice of minimum fares is orthogonal to the driver. Why would a driverless arrive not do the same?

The issue of profitability again is orthogonal to drivers. How do you pay for vehicles and maintenance of your service is not profitable?

I'm not sure the exact reasons for minimum fares, but I'm not sure the driver is entirely irrelevant here. There could be a certain amount of emotion driving that decision.

In any case, to answer your original question: Yes, I think halving the cost of taxis will make a significant difference to their viability in many use cases. I think costs may also fall further than that, as regulations have been limiting competition in taxi markets. (And just intuitively, one would expect that sharing a car should be cheaper than owning one.)

Moreover, I think taxi markets were stagnant for a long time, so the lack of such a service in the past doesn't necessarily mean it couldn't have been viable, and the new players may simply not have explored this kind of service yet.

>How do you pay for vehicles and maintenance of your service is not profitable?

The point is that without a driver, your ongoing costs (i.e. fuel and maintenance) are more proportional to usage. A driver sitting doing nothing is costing you money, but a car sitting doing nothing barely does.

>new players may simply not have explored this kind of service yet...

Think it is too obvious to miss...

>The point is that without a driver, your ongoing costs (i.e. fuel and maintenance) are more proportional to usage.

isnt it the same with a driver too?

In the US I have yet to see a train route that much faster than driving the same route.

It seems to me the autonomous car is a car for wealthy countries, where wages are high, personal time valuable and roads in good shape.

Why is the situation in Peru at all relevant? I too have been to lots of places where self-driving cars would have a ridiculously hard time.

Those places are also very far from the cutting edge of technology.

Even if self-driving cars are confined to US highways, cities, and suburbs they have terrific potential. The future doesn't have to be evenly distributed to arrive on time.

Seriously, I don't even understand you argument. Why would the existence of challenging backwaters in Peru prevent the development and deployment of self-driving cars in the US?

Not the OP but I think he gave Peru just as an example. I would personally say that "normal" and non-hazardous traffic like that from the US and some Western European countries is the exception rather than the norm if we look at the whole planet.

Sure, but population is far from the most important metric when it comes to developing new technologies. The iPhone was out of reach of most of the world when it launched (from a financial perspective), and it was only available in the US. That didn't make it an impossible product to develop.

Even if self-driving cars were confined to only the USA that would still be a massive market and provide a huge ROI for the developers.

Thinking self-driving cars have to work everywhere or even most places to be useful is myopic. It's like saying cell phones will never work because you can't get reception in Antarctica.

I think the biggest market for cars has now become China. India will probably reach the top-5 in the next 10-20 years. Looking forward to the next 30-50 years the biggest demographic increases (meaning future car buyers) will come from places like Africa and South and SE-Asia. It would be myopic to disregard those places as having people who would never afford cars. To continue with your example, almost all of Sub-Saharan Africa passed from having no phone at all to having cell phones (because it was too expensive to install land-lines).

I think you missed my point.

I'm not saying that China and India will never have self-driving cars. There's obviously tremendous profits to be made in bringing cars to those markets.

My point is that a self-driving car doesn't have to be ready for those markets to be launchable as an excellent and profitable product. The iPhone very successfully launched in only select markets—and then made even more money after expanding to China. It'll probably be similar with self-driving cars.

Why exactly do you think a self-driving car has to be globally deployable before launching in the US?

> Why exactly do you think a self-driving car has to be globally deployable before launching in the US?

I think it can launch in the US all-right, but I don't think it will easily export to the other markets because of the above reasons. In iPhone's case that was easily fixable by people in said markets reaching middle-class status, but in case of self-driving vehicles I don't see that happening because one would have to change very ingrained cultural traits (basically you'd have to re-teach most than half to three-quarters of the world on how to drive sanely, good luck with that).

I'd say that for such a capital-intensive market like the car industry only relying on the US and some select Western European markets (+ Japan and South Korea) won't be profitable on the long term, hence my pessimism that self-driving cars will conquer the world.

you'd have to re-teach most than half to three-quarters of the world on how to drive sanely, good luck with that).

This is already happening in the wealthier parts of the world.

I don't think current self-driving cars will work very well in India. But I think they'd work just fine in the rich parts of Pune, to name one area I know well. That's because driving norms in this region are already becoming quite sane. From what I've seen the same is true in the wealthier parts of Mumbai. Lane discipline isn't quite as good as in the US, but I'm perfectly comfortable driving a two wheeler on these roads.

(I'd be a lot more scared driving in Delhi.)

I question whether "sane" driving is actually cultural. Is it possible that "sane driving" is actually just the natural human reaction to good roads and homogeneous vehicles capable of driving at similar speeds (as opposed to a mix of cars, auto rickshaws, pedal rickshaws, 3 wheeled mini trucks, etc)?

> I question whether "sane" driving is actually cultural. Is it possible that "sane driving" is actually just the natural human reaction to good roads and homogeneous vehicles capable of driving at similar speeds (as opposed to a mix of cars, auto rickshaws, pedal rickshaws, 3 wheeled mini trucks, etc)?

The difference in driving styles from Switzerland to its neighbor Italy is quite big, at least from my own experience as a tourist of both countries as a driver and as a pedestrian. The roads in Switzerland are a little better, I can agree with that, but the vehicles present on the roads of both countries are from the same class (no rickshaws and I don't think I've seen 3 wheeled mini trucks in Italy, though AFAIK they used to be quite popular in there not that long ago).

So, self-driving cars will be limited to about a billion people in the short-term? In the US alone we have about 4.5 million accidents and 30,000 deaths yearly.

Perhaps the rest of the world will see the benefit and adapt to the cars.

A stagnating billion people market, both when it comes to incomes and demographics (even though the US is doing slightly better here). AI-run cars would also have to be price comparable to many of the hatchbacks now dominating Western Europe, we're talking about a maximum price of 20,000 euros (with an average purchase price of 15,000). A month or two ago I was told on HN that AI cars are not for people like me, for whom the $35,000 price of a Tesla 3 model was too much (and Tesla 3 is seen as the base, cheap model), so I'd cut a big percentage off of that 1 billion people market you're talking about.

With self-driving it is important to separate owning and using.

> In the US alone we have about 4.5 million accidents and 30,000 deaths yearly.

More like 40,000 deaths yearly, these days. In 2015, it was 38,300 [1], and it had increased 8% (or 2067 deaths) in the first nine months of 2016 [2].

[1] http://www.newsweek.com/2015-brought-biggest-us-traffic-deat...

[2] http://www.nydailynews.com/news/national/traffic-deaths-surg...

> Even if self-driving cars are confined to US highways...

Self-driving cars "confined" to anywhere are not self-driving cars, because you obviously need a human driver to get to and from that special place.

... unless your entire life fits inside that special place, in which case they're good enough.

Indeed. It's easy to imagine urban centres replacing internal traffic with a fleet of self-driving cars. If you're coming from somewhere else, park & ride.

Nice selective quoting.

Millions of cars spend their entire existence confined to US highways, cities, and suburbs. In what way is a car not self-driving if it can successfully drive you on you everywhere you go in your daily life?

You don't "obviously" need a human driver if the car is used entirely in that "special place" (ie. where most cars spend their entire existence).

Your argument makes about as much sense as claiming cell phones aren't truly mobile because you can't use them in Antarctica.

> if it can successfully drive you on you everywhere you go in your daily life?

Well, simply because I don't know everywhere "I would have to go in my daily life" when I am buying the damn car.

>Your argument makes about as much sense as claiming cell phones aren't truly mobile because you can't use them in Antarctica.

By that logic, your argument also makes only as much as sense as calling a walkie talkie with 100m range a mobile phone.

So many people get by without owning a car that if you have to go somewhere where your existing car doesn't work (for whatever reason) you'll be just fine

You can rent or take a cab or an uber. I don't own a car. I get around just fine at home or when traveling.

Which bring us back to the original question. Why the need for self driving cars?

If your time for commute is that much valuable, just hire a driver or call a taxi. Like most people who's time is actually valuable, have been doing for years...

Quite honestly, it is quite a joke when people on HN brag about how "valuable" their time is....

Why? Okay let's make a list

* Safety - they'll be safer. Period.

* It's a natural extension for cos like uber - how else can they bring costs down enough?

* It meshes well with other aspects of tech reducing friction - deliveries will be cheaper (think dinner, groceries, taskrabbit)

* They damn near already exist.

> Safety - they'll be safer. Period.

No. I don't think so. Not in the foreseeable future at least..

>* It's a natural extension for cos like uber - how else can they bring costs down enough?

I don't know or care.

>It meshes well with other aspects of tech reducing friction - deliveries will be cheaper (think dinner, groceries, taskrabbit)

Well, why do you think the providers will reduce the cost? They can always add to their profit, instead of making the service cheaper.

>* They damn near already exist.

So does cold fusion.

>> Safety - they'll be safer. Period.

> No. I don't think so. Not in the foreseeable future at least..

According to the NHTSA[1], they already are.

>>* It's a natural extension for cos like uber - how else can they bring costs down enough?

> I don't know or care.

The point is that they're dumping money into the tech for their own purposes, regardless of effectiveness as a consumer product

>> It meshes well with other aspects of tech reducing friction - deliveries will be cheaper (think dinner, groceries, taskrabbit)

> Well, why do you think the providers will reduce the cost? They can always add to their profit, instead of making the service cheaper.

I sure hope they do; I'd doubt any of those comapanies make any profit as is. And again, they'll be willing to spend lots of money on the cars, making them a viable product without the consumer market.

[1]: https://techcrunch.com/2017/01/19/nhtsas-full-final-investig...

There's a lot of places a normal car can't go as well.

You would think a level 5 autonomous car would not try this; unlike some human drivers.

https://www.youtube.com/watch?v=GgzGaGuNDq0

How would an autonomous car go about judging water depth?

Sensors on the bottom of the car and in the tires? Image mapping. There will also be hazard protection as other cars can talk to your car too.

Correct me if I'm wrong, but isn't the advice to humans to never cross a water-covered road? I thought the idea is that it is dangerous to try to judge the depth.

Yes - depth and water speed. There's also the issue that the road may have subsided or completely washed away.

Because there are 1st world countries where the situation is not that different? Poland for example, actual road: http://i1.ytimg.com/vi/FESuhya2Ig0/hqdefault.jpg

Signage is absolutely atrocious, there's too many signs, some of them wrong, some of the rules are implied(speed limit ends at any intersection, but if a road is going to a private property, then it's not an intersection, so good luck knowing where a speed limit ends), people drive like crazy, there's loads of unpaved roads.....and yet a person from a very rich western country(say UK or Germany) can drive over there in few hours driving time.

Every time someone makes an argument for self driving cars, I keep thinking if they would work back home - the answer is, they wouldn't.

>and yet a person from a very rich western country(say UK or Germany) can drive over there in few hours driving time.

And experience the same issues that a self driving car could. But at least on the journey there and back they would be considerably safer.

Not to mention do something constructive like their taxes or read a book. Heck, I would just look out the window.

I would buy a self-driving car (even if it only works on normal roads and refuses to drive anywhere else) in a heartbeat. Can't drive some road in Poland? I'll manage. But if it takes the driving out of my daily commute I would pay good money for it.

On a separate note, try replacing "self-driving cars / normal cars" with "cars / carriages and horses" in such discussions. It make many answers obvious. (Hint: I am sure there were people who pointed out that the cars couldn't drive over narrow winding paths where horses ruled)

How is the self driving car going to know it can't drive on that road? I imagine the argument is that the self driving car won't notice that the signs don't make sense (like a human would) and causes an accident.

You're making the mistake of assuming people designing self-driving cars are idiots. Being able to detect scenarios where overwhelming signage makes driving impossible would be a basic problem. Even I could do it today.

By writing code that makes sure it can identify and understand the road ahead? If not, then stop and give an error to the user.

Secondary to this, it can be handled by GPS and history imaging. I think GM is doing this to handle snow conditions if I am not mistaken.

That seems like one of the easiest problems, much easier than actually navigating a well-organized road system like in the USA. If the car can follow a lane then it can certainly know when it detects no discernible lane. If the car can avoid obstacles like a pedestrian or a car crossing its lane, then it can certainly know when there is a chaotic mess of obstacles crossing its lane.

Peru's a poor country but it doesn't mean you don't see your fair share of midrange Corollas, BMWs, and Mercedes. You're not going to get commercially successful self-driving cars if they can't target that market.

Yes, you can. It's called market segmentation.

You can't go after every single market at once with the same product. Building and selling for the high end to bring costs down and expanding the market is econ 101.

It's part of the reason the more-powerful-than-Apollo-computer I'm typing on was $1400.

Or to put it another way: "The future is here, it's just not evenly distributed."

You don't have to go to Peru. Just try driving in Boston.

Anyone that's driven in downtown San Francisco, or just seen it without even driving, knows those arguments are empty.

That's why, when the Uber self-driving car blatantly ran a red light, the joke was that it was learning from its environment.

Well they're not gonna wait the whole world to be ready to start introducing self-driving cars in areas where they're ready, so it really depends on where live those people who "jump up enthusiastically"

I think you missed the point of the post entirely.

You talk about technological problems, which, while formidable, are not insurmountable. Surely they are hard but at least we have a direction of attack (as opposed to say teleportation or time travel or AGI, where we don't even know where to start). Just as in the case of Go, the performance of artificial systems will at first be borderline acceptable, then it will be on par with strong amateurs, and then it will surpass performance levels of the best of humans.

The post in contrast talks about social problems and shows how they will cause much strife and friction and will delay the adoption of driverless vehicles even after all technological problems are solved.

> As the designers move away from the relatively trivial confines of well-maintained urban spaces and into different countries and cultures, I seriously question their ability to keep things reasonable for the rest of the human drivers

And why should we support and enable unacceptable infrastructure and a broken culture? Isn't it better to force everyone else to conform to a single well-thought-out and managed system?

You are right in saying that autonomous cars, as we envision them for the near future, require some form of "proper" infrastructure to be able to perform (safely). However you are wrong in seemingly thinking that this infrastructure will never come.

As with many technologies, it is the infrastructure that is important ( cell phones vs cellular network, ethernet-enabled computers vs the ISP broadband installations, iPhone vs iTunes store, etc...).

Here it is similar and it is not so far fetched to think that in some future the road system will be modified in some ways in order to accommodate level 5 autonomous cars and people.

For example, the city of Drachten in the Netherlands has no traffic lights, road signage (stop signs, etc...) at all. They modified their infrastructure and got better pedestrian security.

From Wikipedia:

"Drachten received international attention for a traffic experiment known as shared space, a concept pioneered by Hans Monderman. Almost all traffic lights and signs have been removed in the town's centre in an effort to improve traffic safety, based on the theory that drivers pay more attention to their surroundings when they cannot rely on strict traffic rules. Previously the town's centre had an average of 8 accidents per year. In the first two years after the system was introduced, yearly accidents were reduced to 1."

It would be very interesting to see how level-5 autonomous cars would fare in Drachten...

After a skim of this sub-thread I'm surprised no one mentions that it might be a benefit that as autonomous-car driving tech spreads around world (presumably with SF/SV being the epicentre), countries which are poorly equipped for autonomous cars start to find the momentum to overhaul their road network and driving education system.

Having been to several countries in Asia, it would be no bad thing to have a stepwise reform in how driving is done in these countries - think of the number of accidents that would be reduced.

Or take advantage of the car's systems to steal from drivers, scam them, or take precedence on the road.

Self driving cars are designed for well maintained first world situations where the rule of law is followed (or at least the driving regulations are). Sure in India or Thailand where it is a free for all and the infrastructure is poorly maintained they will not be suitable but in these countries the death toll on the roads is also ridiculous. In the USA, Canada and Eurioe, the largest markets for these vehicles (at present) the cars work fine and will ultimately reduce the number of deaths on the roads. Eventually other nations will catch up and self driving or at least assisted driving will be much better for everyone.

Self driving cars are designed to learn.

People are way overestimating how much time it's going to take to retrain them for cities in new locales.

Will they sell in India to start? No, of course not. Most tech in the US takes a few years to launch internationally. But I think it will transfer much faster (5 years for internationalization) than the doom and gloom prophets on here say.

Potholes actually seem like a situation that autonomous cars would handle far better than humans.

And yeah I expect that the jump from level 4 to level 5 (assuming that level 4 does not include the most chaotic of world streets) will take decades, but that's fine.

I'm more skeptical that a human can deal with this complexity better than a computer could.

Many people think that self driving cars will lead to less cars on the road, but I believe they will lead to more cars and therefore worse traffic. Since you won't be actively driving the car, the time the commute takes will become less of a chore and you will put up with it taking longer. Think self-driving Winnebagos that go off and get cleaned and re-fueled when you are at work. Additionally, why pay for parking? Just send the car about the block for 2 hours while you watch the movie. The cost in gas/wear-n-tear will be less than the $12 for parking in some areas. That crowded concert that already has parking issues? Those are going to be much worse with a self-driving car. Also, what about teenagers? Take a blow-up doll, tie a rock to the feet, and put it in the middle of the road that leads to school. Do that in 5 places. How long would it take for a person to look up from their phone and realize the car is no longer moving? Essentially, taking the 'common sense' out of driving for an algorithmic approach is not going to work very well either. Yes, we may kill less people than before, but I believe that you can't just algo your way to less traffic. You have to use both the self-driving and the person's brain (what little motorists seem to have) together. Taking people of the picture won't work either.

> Many people think that self driving cars will lead to less cars on the road, but I believe they will lead to more cars and therefore worse traffic. Since you won't be actively driving the car, the time the commute takes will become less of a chore and you will put up with it taking longer.

No, you just won't own your own car, you'll car share via a service like Uber. It would be way cheaper since you would no longer carry the maintenance and insurance costs for a vehicle. That's why it would lead to fwer cars on the road.

When in the history of economics has changing nothing but making a resource _cheaper_ lead to consumption of that resource _decreasing_?

Don't kid yourself, if car driving gets cheaper people will do more of it. Or will you not be more likely to take a cab ride when it's $1 than when it's $20?

Why will car driving get cheaper? I think the argument is that Uber will get cheaper, not that owning and driving your own car will get cheaper. Perhaps it will get slightly cheaper if self-driving is much safer and thus leads to lower insurance costs, but that's about the only effect I can think of.

You are ignoring a major cost of operating a privately owned car: the time you spend driving it.

In the greater NYC commuter area, people already willingly pay more to live near a train station so they can more easily commute without having to "waste" the time they'd spend driving. (I'm sure this is true elsewhere) People are obviously already putting a price on this. Why pretend the cost of driving yourself is zero? It's absolutely not, unless your time is worthless.

Thought experiment: What's the longest commute you'd be willing to bear, if you drive yourself? How about if the car is self driving? If the answer to the latter is larger, you just proved you'd save literal money with a self driving car, not just figurative -- because you could move further away to cheaper housing.

> When in the history of economics has changing nothing but making a resource _cheaper_ lead to consumption of that resource _decreasing_?

Autonomous car price will be higher, maintenance will be higher due to more costly electronics, insurance will be lower due to lower collision rate. Given these 3 unknown variables, how have you definitively concluded car price must be lower?

Finally, your economic argument is incomplete. When renting a resource is much cheaper than purchasing it outright, renting always wins.

Owning a car will be the domain of car enthusiasts, not the typical consumer.

I'm sorry for any confusion but I did not mean to artificially distinguish driving an owned car versus renting a ride. Same thing, a 10 mile trip on either will put the same amount of congestion on the road. People will want to ride in cars more if it's cheaper or easier so we'll have more cars on the road. The total number of cars required to service this traffic may be less, but that will only reduce the number of parked cars.

As for maintenance/cost of hardware, I've heard that the LIDAR kit is under $10,000. Given how expensive new cars are and how cheap the same car becomes 10 years later, and given that that's a starting point, I'm not too concerned about depreciation cost.

In the western world there are also already so many suckers willing to pay the initial depreciation by buying new cars and then immediately selling them 5 years later for much less so they can pointlessly upgrade again, so the used car market will be filled with affordable self driving cars a few years after they come out. Like, you can already buy a car for less than the combined value of all its parts, because people put a negative premium on old cars. Why would it be any different in the future?

And given that LIDAR doesn't have moving parts, I'm not too concerned about the cost of maintenance. When you buy a 15 year old car, it's rarely the ABS or ECU or radio you need to replace, it's always moving parts like parts of the valvetrain, transmission, suspension.

And that's if you even use LIDAR. Tesla's cars apparently just use good old cameras. And we know how cheap and reliable those are

> I also remember run-down suburbs where massive potholes and ditches make navigations of certain streets a puzzle by itself, compounded by garbage trucks ahead just doing their thing, improperly placed signage ahead, and those badly-designed intersections where you can totally count on drivers misjudging distances.

Even assuming your anecdote is at all relevant, why would any of these factors impact the ability of an autonomous vehicle to navigate these streets? The car knows where it has to go, and it senses its environment more frequently, with more precision, and with faster reaction times than a human ever could. Either the objects surrounding it are on vectors that will impact its chosen path or they won't. If they don't, then it proceeds. If they do, it must slow, stop, or find an alternate path that avoids those vectors. There's nothing earth shattering here.

Can an autonomous system judge the difference between a harmless piece of trash on the road and something spiky that's going to blow out a tire?

If it has seen millions more of both things than any human, then I don't find it hard to believe that it could distinguish them better than most humans.

We don't have to go that far; anyone who thinks an autonomous car could compete in normal weekday midtown Manhattan traffic hasn't had the pleasure. You have to change lanes constantly to avoid double- and triple-parked trucks and taxis, and street construction – and every lane change is a little game of low-speed Chicken, in which each driver pretends they don't care if they get hit, and presumes that the other will yield. Autonomous cars can't pretend; every human driver will know that they have to avoid accidents whenever possible. And paid human drivers will be particularly eager to block their progress, to penalize their passengers.

I thought it was a very good article, but disliked the comparison to the sad state of US trains.

The author knew he was being a little disingenuous with his comparisons to the sad state of US infrastructure, which is why he was careful to preface every reference to poorly automated train systems with the US. There are a number of automated systems worldwide that work pretty well. The Yurikamome was the first line that came to mind. Its not an extremely packed line, but having safety systems similar, like door gates, works very well. Automating a more used line would of course be more difficult, but not impossible. Aggressive door closing seems to work pretty well in taming crowds. It might take an education period in the US, but once people realize holding a door won't allow you on the train and that it will only allow you to remove a trapped limb, people will be less inclined to do it.

Haha, yeah. Entirely agreed.

There is, within 15 miles of the continental United States, a Level 4 automated train, that has been running for over 30 years, with 50 miles of track, and 300,000 passengers per day. And it works great. It's off-the-shelf technology backed by a moderately sizable 70-year-old company, so you can just go out and order a system like this yourself next year, if you want to (and you have a few billion dollars).

That doesn't solve any of the self-driving car problems in the article, so it's totally a side point, but Level 4 trains are a solved problem, for over 30 years.

I have to disagree with Social outcasts; A study this year concluded that driver pedestrian interactions are at an all time low; likely because of both parties using cell phones.

As the article states, I often enter cross walks to test if a car is going to slow down; and they often don't. Similarly, I recently had a near miss when a pedestrian ran across the street when I had a green light (late at night too).

(in Atlanta) Currently, cars are 1st class , Pedestrians are 2nd class, and when I bicycle, I'm often 3rd class.

Self Driving cars, I hope, will make this more equal because I'm tired of almost getting hit when I walk across the street in my residential part of town and a "showboat" enjoys flooring it (going 2x the speed limit) down my street for the 1 block between traffic lights.

Twice now in the last year I've waited at intersections for the walk signal and when it came on I almost got run over and got beeped and screamed profanities at by drivers taking a right. I guess now drivers believe that they have the right to the road 100% of the time and nobody should ever be allowed to cross, even at crosswalks during the prescribed time for crossing.

Drivers have become increasingly hostile in the last several years.

The author is wrong, the manufacturers are already considering the problem of communicating with pedestrians.

Mercedes: http://www.electronics-eetimes.com/news/mercedes-benz-self-d...

Google: http://time.com/4129247/google-self-driving-cars-patent/

And this article in popular science: http://www.popsci.com/people-want-to-interact-even-with-an-a...

Do pedestrians now have to learn the UIs of different brands of cars in order to not be overrun?

Many good points in that article. Probably self driving cars are finally going to force us to review and to fix in many occasions the way we are using cars today. Especially the ways which are already broken today. A good example is the competition for the parking in front of a Starbucks or other similar shops. The common usage pattern is still crafted by how traffic was decades ago, and you could still just drive to a location and leave your car. But that depended on there being fewer cars than parking lots. In most locations, this is a thing long past - so I would consider the current situation already as broken. Self driving cars would make it worse and such could force us to finally find solutions for these situations.

There are two obvious ways of fixing that situation. The first, recognize that trying to reach a location by individual cars cannot work out and ban private cars. Many European Cities have large pedestrian zones where private cars are locked out of whole streets in city centers.

The other option would be, and fortunately self driving cars would be part of the solution, to create enough parking spaces. Parking in the streets would still be disallowed, creating large pickup areas instead. So the road sides would be free of parking cars, enabling drivers to stop at any desired spot. As soon as the passengers left the car, the car would leave immediately for a nearby garage. This would of course depend on the creation of the necessary amount of parking garages.

Banning private cards, I believe, will be unnecessary. I know free market is often folly but in this case, I believe, the advantages of just tapping in the app to call a driverless cab will be huge. In almost all cases it must be cheaper than owning a car which doesn't work most of the time just sits there.

I was not speaking of a total ban of private cars, just banning cars from certain streets. Converting streets to pedestrian zones is quite successful in the center of large cities, where the distances are short and the traffic density higher than individual transport could deliver.

Self driving cars could solve the problem of congestion and the problem of parking quite easily. If all cars were autonomous, then the flow of traffic could be optimised and things like traffic lights and roundabouts could be eradicated, this would drastically improve travel times. In an ideal world, no-one would own their own self-driving car, there would be a fleet (ideally state owned and taxpayer funded) that you simply order on-demand, like uber. This way there is no need to park. You simply arrange your pickup times either on-demand or ahead of time, so a vehicle is always there to scoop you up wherever you are.

"(ideally state owned and taxpayer funded)"

Why "ideally"? That would be a recipe for broken, dirty cars. People won't take care of government property (or property that doesn't belong to them in general). You don't usually see privately-owned cars driving around with ugly graffiti tags, nor do such cars typically smell of urine, the way that public buses and train cars often do.

With respect to cars specifically, ask anyone who's worked for a rental car company what kind of condition the cars often come back in. People abuse the crap out of rentals to the fullest extent they think they can get away with. This isn't specifically a modern phenomenon, either -- consider the old-time phrase "I'm going to beat him like a rented mule."

Nonetheless, public transit is an immensely popular way of getting around in cities all over the world, and it comes with the advantage of providing greater access at lower cost than privately owned transportation. There are many people willing to make this trade-off.

Public transit does not need to be owned by the government.

But governments can work too, so I don't think there's a universal best way to do it. (Here at Brazil we'll certainly be better with no government interference at all, but other places have other trade-offs.)

> Why "ideally"? That would be a recipe for broken, dirty cars.

Not necessarily. Granted, very few countries run a well-functioning socialist state, but that doesn't mean it isn't possible, more that many countries do a poor job when it comes to implementation.

The solution would be quite simple. These smart autonomous cars would contain security cameras that ensure passengers don't vandalise the vehicle. If they do, then it's a simple case of locking the vehicle and driving the passengers to the police station, where they would then get booked for vandalism of public property, and banned from using the service for a period of time. I should imagine that this would quickly curtail the issue.

The only problem with this is I can't imagine the logistics of out-of-town travel, say I want to go 200 miles away for a week. Have the car travel 200 miles and drive back? Have the car stay in the destination city? Then you have to have thousands of cities with some sort of ownership/maintaining agreement.

Don't forget that once we get to level 4, every time there's rain or snow, there might even be more loss of life as everyone gets to learn how to drive again in bad weather.

This is a great article that covers not only edge cases, but almost subliminal behavior that people perform almost every day.

To me, it is clear that our behavior will change. As an American in Amsterdam a few years ago, I was horrified to learn that at some places pedestrians can just walk out in front of you, and you have to stop. After a few days of observing this (as a passenger, thankfully), I noticed that you could tell the tourists - the natives would walk into the street without looking, knowing the drivers would stop. Tourists would hesitate, looking to the driver for feedback. Oddly (to me, as an American), I found that the natives had it right, and that blind faith made things smoother. The tourists, in checking for driver feedback before waling into a street, would actually slow things down more than necessary.

This is but one anecdote of thousands, but it shows how the expected behavior of a person from one place does not necessarily translate to another place. I would extend that same thinking to the future, in that or future will be a behavior change from what we do today.

I'm sure it will come slowly (although occasionally startlingly), and there will be new signs, new laws, and new behaviors. But for me, the best comparison is to when we introduced cars into the streets that had previously been owned by horses and carriages. a bit of chaos, but fairly rapid adoption (within a decade). It wasn't uniform, and it was messy, and some places lagged the modern world by decades. But it came, we adjusted, and now we can't even remember what that previous world was like.

What's that quote about the future not being evenly distributed? I think it applies here perfectly.

Weird, but somewhat relevant question (given that self driving capabilities combined with buses can potentially disrupt public transportation in urban areas): why do we even allow anything that's not a bus (say, 10+ passenger) or industrial vehicle on the road?

The downsides of such a thing seem small.

1. There would be less traffic

2. The speed limit could be raised

3. Due to higher occupancy, more people could go to places, faster.

4. Cheaper per person (maybe) than most other forms of public transportation

5. High density areas can have fleets of buses go there, express. This would have only a little overhead compared to driving your own vehicle, yet would be substantially cheaper.

Depending on destination, you might need to change buses several times, which is of course inconvenient.

OTOH, booking a busride far in advance might mitigate this. Also different price-levels could mitigate this to some extent. I'm not sure if it would be sufficient, though.

If and when all cars are computer-controlled another consequence may be that children will be free to roam about. This assumes that those same cars are built to be incapable of hitting pedestrians. Is this possible?

Children will be able to manipulate the cars and therefore create a bottleneck for others, this is a major issue outlined in the article.

Well, this sort of behavior has (presumably) already happened [1], but this also gets to the crux of "unintended consequences". Will we see kids deliberately walking in the street, knowing full well they aren't going to get hit and extorting drivers to let them pass?

1 - https://electrek.co/2016/07/21/tesla-autopilot-saved-life-pr...

Yes I'm sure this sort of thing will happen however I imagine the overall social consequences would be beneficial. Children are highly confined at present and the ability to roam and talk to different people, walk to school, etc, has been lost. Largely for fear of their being run over by traffic.

Largely for fear of their being run over by traffic.

Are you sure? I get the feeling that good old Stranger Danger is still alive and well, and automated cars won't in any way change that.

I expect as far as the social aspect of driverless cars making their way through our neighbourhoods goes, eventually most humans will just come to see them as an odd type of animal. Like a horse. It probably wouldn't hurt to give them some active animal-like cosmetic features that help nearby humans anticipate what they are "thinking".

Delightful symmetry there, no? The first great task of mechanized transport was getting rid of the horse aspects, the second was building them back in.

Nice thought, but as the article states, car manufacturers do not want their cars to be recognizable as autonomous, because others might take advantage.

If the advent of self driving cars corresponds with reduced private ownership then cars will not need to wait and our children just need to hop into the next free car and their phone or smart school bag will contain their home address. For adults, cars will not need to hand around near venues because we will just get into the next available car.

I agree that the transition will be messy. Perhaps we will need to redesign our local communities a bit to help. More off street parking for example.

> Since there are no current ways that driverless cars can give social signals to people, beyond inching forward to indicate that they want to go, how will they indicate to a person that they have seen them and it safe to cross in front of the car at a stop sign?

There is no reason that autonomous vehicles can't give other rudimentary social signals in ways similar to drivers that can't see each other by using their headlights, turn signals, emitting short 'beeps' from their horn, and so on.

For that matter, autonomous vehicles can be given greater range of expression by adding a 100^2 pixel RGB LED display or similar behind the windshield or on the hood that can display emojis, or by actually giving them the ability to speak (KITT-style): https://www.youtube.com/watch?v=i-UqF5ElduY

Eventually perhaps. But until then, there is a very large uncanny valley to bridge.

The comparison to level 4 or 5 trains doesn't make sense to me. For a train you've got millions of dollars worth of hardware and track so the driver is only a tiny component in the overall cost. For trucks and taxis the driver is the biggest costs. The incentives for autonomous cars are much higher than autonomous trains.

What about that part where all of our computers are riddled with zero-days, and wide open to shadowy government organizations with massive technical budgets, and the part where we weren't 100% cool with that.

And like, what about the part where we were worried about nefarious anonymous super hackers cyber-attacking the electric grid, and the gas grid, and other critical infrastructure, and we were worried about that sort of thing spiraling out of control into a federal emergency, because heat and potable water were taken down in the middle of winter, when a utility worker's laptop was spearphish hacked, because internet of things? I guess that can't happen anymore?

So now, right in the middle of this, the cars get to drive wherever they want because computers, and nothing can go wrong. One has nothing to do with the other, and we'll all be safe and totally okay, right?

One thing about self-driving cars that I'm not looking forward to is the proliferation of zero-occupant vehicles on the roads.

Remotely connected AI with ~1MJ of KE? No thanks.

Having the description of abbreviations all the way at the bottom of the article so that one has to scroll a mile to read what they mean is a bit annoying. Instead of:

>When the first IMPs^1 for the fledgling ARPANET were being built starting in 1969 at BBN^2 in Cambridge, MA, I think it safe to say that no one foresaw the devastating impact that the networking technology being developed would have on journalism thirty to fifty years later.

>[A million words]

>^1 Interface Message Processors. Today they would be referred to as Internet protocol routers.

>^2 Bolt, Beranek and Newman in Cambridge, MA, a company that was always known as BBN. As distinct from BBN, the Buckingham Browne and Nichols school in Cambridge, MA — no doubt many employees of BBN sent their kids to school at BBN.

I would have preferred for at least the first of these two to be baked into the text itself.

While I agree to some extent, I find it funny that I've somehow been around long enough to not realize people needed your example acronyms and names to be defined. Including the reference in the BBN definition to the other bbn :) I guess that's what happens when you live it first hand.

What about privacy?

There could be some serious privacy issues introduced by self-driving cars if they take off.

Cameras and microphones patrolling neighborhoods of the world, constantly feeding data into "the cloud" with absolutely no accountability?

Just no!

I had to stop reading because the blog poster's level of thinking is really superficial.

If you're going to think completely autonomous self-driving cars, why the hell would you bother with things like parking? Why would I even bother "owning" a car? What the more reasonable approach is time-sharing of multiple vehicles, like Uber.

So instead of worrying about parking, you order a car, and it picks you up, and then you get dropped off. It drives off and services other customers. Then, when it comes time to get driven home, you order another car that comes and picks you up. You only pay for the car while it's driving you, not while it's sitting there doing nothing.

Or, if you did own the car, when you are not using it, it could go off and service other people and earn you money, and then return to pick you up.

And why would you go to Starbucks to pick up your own coffee? Why not send a self-driving car through the special self-driving car drivethru to pick up the coffee and return it to you? It could wireless transmit the order and payment details for you, and have a special slot for the coffee or food order.

Seriously, the article is really superficial thinking when it comes to self-driving cars.

That's certainly true for some drivers. But there are still plenty of good reasons to own your own car without sharing it. For one, sharing it won't earn you much anyway, because everyone tends use their cars at about the same time, travelling the same direction. Then, it is very convenient to keep some stuff in your car that you might need but don't want to carry around all the time, especially if you have kids. So I don't think exclusive car-ownership will disappear.

I don't think you can own a self driving car (or at least regulators shouldn't allow it). Safety updates and inspections need to be on the level of FAA plane inspections. I mean really, these cars should drive themselves to a service centre at least once a month. When you couple all those costs together, and the liability from car modification compromising self-driving hardware, I just don't see it as a viability safe option.

I think a lot of jobs would become flexible with self driving cars. Imagine 8 hour commutes being completely viable.

8 hour drive even as a passanger on a bus is a hassle. Self driving car will:

- still shake on bumps,

- make reading uncomfortable for many,

- not be quiet,

- be too small to stand up and stretch

- not have a toilet so you still have to stop somewhere

- not make eating on a run any easier.

It is certainly more comfortable than driving for few hours so it can extend commutes, but not by that much.

Nobody is going to buy a self driving car unless it has a bed in it ;)

Nobody can sleep in moving self-driving car. Sleeping in stationary self-driving car can work (somewhat) if it is connected to the grid, because of necessary climate control. Then it still will be worse experience than sleeping in a camping trailer.

You can't expect to be able to sleep comfortably in a moving car. It will be like trying to sleep while someone tries to wake you up - noise and constant erratic movement. Even if you can get over it, to maintain safety you will have to be strapped to the bed and make sure that nothing heavy (even your phone) would struck you in case of an accident.

I once had a great idea - a car that would drive backwards with complicated display system to show a road in the direction you are actually driving. The idea was that it would be safer than seat-belts. But there is a reason why seats (even passenger seats) in a car are front facing. In case of an accident everything will try to maintain it's momentum. If you would have a coffee or a phone in your hand it will struck you in your face.

That's a problem that makes sleeping in comfortable bed inside a car dangerous. You can't get around it for foreseeable future.

Cars won't be dangerous once they are all automated. Once cars are automated, my guess it'll take years, rather than decades, for regular driving to be outlawed.

I keep hearing this rhetoric and one could say that this superficial thinking because it's very out of line with how people all over the world use their cars today.

I don't feel comfortable renting out my car to strangers while it's not being used. I'm afraid of it getting dirty, them not operating it properly and causing mechanical damage that won't be seen for weeks.

On the other hand, having an "on demand" car service as my only option would increase my transportation costs. My car is paid off, insurance is almost free and gas is dirt cheap in the US it might as well be free. If I had to Uber around with self driving cars my costs of getting around would skyrocket.

Are you suggesting having your own car as well? That's the only way I could see this working.

> I'm afraid of it getting dirty, them not operating it properly and causing mechanical damage that won't be seen for weeks.

With level 5; this doesn't happen.

Level 5 doesn't mean the driver can't take over - just that they should never have to. The driver will always have the ability to take manual control; something that won't change within our lifetimes. Even if a company manages to legally release a "level 5" car in the next 50 years, it will really only be "level 4.5" - 100% automation under all conditions isn't in the cards for the very near future.

It astounds me how naive and over-optimistic we are seeing enthusiasts of self-driving cars behave. The industry is in its infancy, but people talk about it like our roads have been exclusively filled with them for decades, and all challenges and bugs have been conquered. Surely every person who talks about these cars like it's a solved domain can't all be prospecting, bandwagon investors.

Can you speak more to why you're so confident that this won't change within our lifetimes? I don't know how old you are, but my remaining lifetime is a pretty long time. It would have been impossible to predict the technological and legal conditions of today an equivalent number of years in the past. And I doubt we've gotten much better at predicting future advances.

> and bugs have been conquered

I can't get a bug free desktop OS after 40 years of development, what chance a fully automated vehicle?

So level 5 is supposed to magically stop people from barfing in the back seat, getting things accidentally wedged in door and trunk hinges, smearing dog poo on the carpets or scratching up seats with paws, opening car doors into traffic, or lighting up a massive doobie while riding?

Level 5 is a fantasy world as it is, let's not pretend it can do even more than is described.

Yes. It means that people who do those things often get their account suspended. If an Automated Uber arrives in an unsatisfactory condition, you could just reject it and send it back for maintenance/cleaning, and a new one will be on its way.

You can do all of that today in non-magical normal Ubers, and yet it still happens.

Or they get charged significantly extra for it.

I was focusing on "operating it properly"

> On the other hand, having an "on demand" car service as my only option would increase my transportation costs.

I've done some of the math on this; here are some very conservative figures.

- $50/mo for reasonable high deductible coverage.

- $100/mo gas

- $100/mo parking at the office

- $x /mo parking at home (not realized in my case)

- $80mo misc expenses (tires, wiper, diy synthetic oil change, taxes, car wash)

$330/mo pays for a lot of $5 uber pool trips.

And while our cars are paid of; The average American does not and is increasingly spending more. This adds an additional $300 to the amount I can spend on trips.

This says nothing of the current liability of driving a car. Where an accident with a > $50k car with over $100k of medical costs creates a gap in coverage for the average American.

I don't commute to work by car so for my occasional use of getting groceries, going out to eat, getting a haircut or doing the occasional shopping, I don't spend anywhere near that much.

- $40 for insurance

- $30 gas

- $5 (Maintenance)

$75/month on my end. If I were to UberPOOL everywhere, besides the fact that it would be inconvenient because it would take longer and I would have to share a ride, it would cost more.

You'll have to calculate how much it costs to buy the car and the loss of value when you estimate to sell into that as well to get a fair comparision.

If you don't drive much the fixed costs of buying the car factor in much more negatively compared to the variable costs. Every trip you make costs much more than someone who drives a lot.

This is no easy function as it depends on your fixed costs vs. variable costs and trip lengh.

Especially for people going shot distances on rare occasions a rented car might be the better option.

$5 in maintenance sounds low. $5 a month in maintenance is $600 over 10 years, a new set of tires would cost around $600 and you should replace your tires before 10 years no matter the tread depth(1). I believe you are probably doing more maintenance to your car than new tires every 10years.

You also can't just totally discount the initial cost of the vehicle like that as "free."

(1) https://www.edmunds.com/car-care/how-old-and-dangerous-are-y...

But at $40/month for insurance, you likely have liability insurance only and a gap in coverage if you hit a car > $50k or significant medical costs.

Most states have tax depreciation/deductions per mile that are considered by most to be much lower than accurate. In GA, it's $0.55/mile (real costs were estimated $1/mile at $4 gas); so, at $30/gas, you're being very "efficient" because it should be costing you at least $150/mo for that kind of mileage.

While autonomous cars will definitely deflate prices downward and make it more practical to not own a car, comparing to Uber pool is ridiculous for most people's personal car usage. I live in a relatively dense city for the US with a large Uber presence, and I've completely stopped taking Uber pools because they take twice as long as a direct Uber at a minimum. And then account for the fact that you now have special considerations for grocery shopping or any other type of hauling, children with car seats, out of town trips, etc.

Uber and public transportation has made it extremely easy for my family to get rid of one car, but not both. Maybe if I were a light-packing 20-something single person in SF I could get away with Uber pool all the time, but for most people it's a non-starter.

Even if the car is paid off, there is still the opportunity cost for the value of the car being invested.

$100/mo gas is ~1600 miles/month. That's 20,000 miles/year. Where are you commuting from? The moon?

I don't know what kind of car you drive, but my wipers cost $40, and don't get replaced every year, my oil change is ~$50, my tires cost me ~$400, but last for 60,000 miles.. At some point in my car's life, I will probably throw a few thousand dollars into repairs - but its cost of maintenance is not going to be ~1,000/year.

There's also various licensing and registration fees, which are, essentially, roadway and driving taxes. You're going to pay your share of them, regardless of who owns the vehicle you will be riding in.

These are not conservative figures, unless you're driving an M1 Abrams, or, like my partner, a 20-year-old junker...

And, of course, if you're driving 100 miles/work day, I doubt those uber trips are going to cost you $5.

> some very conservative figures.

> $100/mo gas is ~1600 miles/month. That's 20,000 miles/year. Where are you commuting from? The moon?

UK example here. During the course of my work career I have driven 700,000 miles, which is to the moon, back to the Earth, and most of the way back to the moon again. Or about 28 laps of the equator.

I live in a semi-rural village 65 miles from my place of work (a town about an hour away from central London) and visit clients scattered around the south west of England. Several of my colleagues live 30 to 40 miles from our main site, but far enough from me that we can't car-pool. This pattern of commuting is quite common in the UK for people living outside major cities with approx 3.7m workers having a commute of two hours [1] (including trains as well as cars).

By the way, I am really bored with the M3 motorway.

[1] https://www.theguardian.com/money/2015/nov/09/million-people...

My current commute is about $90/month. I eat about two galleons round trip, every work day so around 20 days, gas is ~2.30 from the right place, but it's been rising. Excludes commute for social activities though so it ends up higher.

I guess it beats $25 Uber rides though

- Gas is $2.50 A gallon (in Atlanta, the average commute is 33 miles in traffic, so a weekly fill up in relatively ordinary, this is not a uber compatible number)

- Synthetic Oil parts cost $26 for 6 quarts + $10 for a premium filter.

- A Car Wash Once a month is $20

(I didn't list any surprises, alternator, cooling system, ac, blower, emissions, brakes, suspension, oil leaks, wheels, etc - and I can assure you that plenty of people spend over $1000 on many of those when they go to designer car shops or dealers.

I definitely forgot that I was talking within the context of trips that are your typical uber trips.

>On the other hand, having an "on demand" car service as my only option would increase my transportation costs.

I don't think anyone is suggesting it would be your only option, simply that most people would prefer it because it will be cheaper (since you're amortising the overheads of car ownership over many users).

And while owning a car might feel prudent, if taxis become cheap enough, it might not be the best investment for your money anymore.

The fact that your car is paid off does not mean it will last forever.

The on demand service will have well maintained cars, where the biggest costs (depreciation and insurance) are shared between a large number of people. There's just no way¹ private cars will be cheaper.

1 - Ok, there's government intervention and monopolies. Given a competitive market, there's no way.

depends on where you want to go with the idea. Of course you can go all the way and say noone will own cars anymore and personal transportation will be ubiquitous everywhere, but it's much more likely that it will start slowly.

People owning cars like today that drive and park themselves, after that maybe people are willing to rent out those cars when they don't need it. Many people probably would not do this all the time because they don't want to risk the car not being present when they actually need it in some cases.

Only after that and only if enough of self driving cars are available at all times, would people really stop buying their own cars.

Your idea of the drive-thru for self driving cars will also take some time to actually happen, self driving will not happen from 0 to 100 with all possibilities a reality from the beginning.

Keep in mind: at no point has any manufacturer stated that a self-driving car - even at level 5 - would be allowed to drive itself around without a person in the driver's seat. The idea that people will avoid parking by having their car drive around empty is absurd, and completely out of touch with how self-driving cars would operate within the next 20-30 years.

Audi have already shown off something like that: http://www.zdnet.com/article/audis-self-parking-car-whats-st...

Elon Musk has stated that you'll be able summon your car from the other side of the country. It will even charge itself itself on the way.

I think not having to deal with parking will be one of the major benefits of self driving cars. It gets even better if you just rent a car for a couple of minutes instead of owning it.

Hopefully it'll be like a taxi at a much lower cost.

Considering that as of now, not having to deal with parking already is one of the greatest benefits of not having a car I think this is spot on.

Creeping to the next free parking spot while scanning online availability data could easily be the one task where robotic cars would be consistently safer than drivers.

Me too, why would you even drive to UPS if self-driving cars are the norm? UPS would drive to you, where ever you are and pick up the package...

I think self-driving cars and the ever cheaper cost of electricity will fundamentally change everything.

While there were some interesting insights in the article, they are not the unintended consequences alluded to at the start with the reference to IMPs and the early Internet leading to the disruption of journalism.

An example of an unintended consequence of fully autonomous vehicles would be what might happen to where people want to live and hence the structure of communities and property values.

Based on the journalism example in the introduction I was expecting the unintended consequence to be for the professional drivers to lose their jobs. Which will be more impactful on the labor market than any changes in journalism so far and can completely eliminate "driver" (truck, taxi, etc) as job title. Moreover, this is a part of a broader trend of automation that is real but still left out of mainstream political discussion. If Trump hopes to bring back employment with traditional protectionism, he'll just incentivise faster replacement of human workers with robots and automatic systems. I believe this is a more important topic for discussion than the social and antisocial aspects of driverless cars.

Aside: not to be that guy, but the article says "unexpected"; the word "unintended" (or anything related to intention) isn't used in the article at all. At least not explicitly.

Just mentioning it as a bit of "expectation management" for what the article is about, since there's overlap between consequences one doesn't anticipate and consequences that are unintentional, but it's not quite the same thing.

With the development of high speed internet, I would have expect a radical change of nature of work where teleworking would be the norm. Maybe this change is happening, but at a very slow pace. IMHO, the changes caused by autonomous cars will also be progressive.

The internet has been fast enough for 10 years. Something else is preventing teleworking from becoming the norm.

I might want to own my self-driving car if it is my private space where I sleep, eat, work.

Owning a self-driving car may be cheaper and therfore move viable if they are safe (a few sensors instead of expensive construction and safety technology).

Self-driving cars could be cheaper if they turn out to be domesticated nodes in a network instead of status symbols.

I wouldn't completely rule out car ownership just yet...

Are you suggesting car manufacturers should forego airbags and seat belts because self-driving cars obviously would not crash? You do realise that someone else could still crash into you? Like getting T-boned in a busy junction.

Assuming it's all self-driving and networked, why not? The next smartphone (or implant) could warn cyclists and pedestrians as well.

That's a little futuristic, but sensors and software could potentially be good and cheap to get things safe.

That's assuming it has somewhere to go, i.e. escape.

Two reasons to own a car, even if level 5 autonomous is available.

1. Storage. As a family with little kids, the car is a mobile storage for diapers, toys, and other stuff. This was the main reason, why we got our own car. We used car sharing before.

2. Customization. Cars are still a status symbol and for heavy commuters also living space. Car manufacturers already provide more and more options to customize.

The second reason is not that strong. I can imagine that cars lose the status symbol appeal, because economic reasons always win. Eventually, only rich people own their cars.

The first reason is difficult. My prediction would be that many people will switch to car sharing. It is already happening even without autonomous cars. Families with little kids still want their own cars. Effectively, it becomes more expensive to have kids due to autonomous cars, relatively.

I think you could have a transferable storage unit for people who need it, especially if the cars were specially built. So once you get out, the car goes somewhere (probably nearby) and drops off your stuff, and then the next car picks it up again. It would probably be more expensive, but potentially still worth it compared to storing the whole car for people who need that capability.

That said, I don't think self-driving taxis will completely replace car ownership, at least in the medium term.

That requires the stuff to be in the storage unit rather than randomly distributed across seat cushions, door cubbyholes, cupholders, and so on.

(Cheap robotic guaranteed theft-free car valeting might change this, I grant you)

Some people will still want to own a car, some won't. However, parking is going to be a lot less of a problem.

1. Some people will used shared ride services, so the shared car rarely needs to park.

2. Cars will be able to much more densely park. Autonomous cars that can communicate with each other can pack themselves in on the side of the street bumper to bumper. In parking lots with rows, spaces can be much narrower since people don't need to open the doors, the car can just pull forward itself. Even the rows between cars can be filled up. You see this at the facebook campus now because they don't have enough parking, but it takes an army of valets to move all the cars around.

3. People will own and drive smaller cars on average. Even for people who do own a car, the ability to have a larger vehicle available on demand will tend to make them buy smaller cars.

4. Even when it's hard to find parking in an area, the car can drive itself to areas where parking is less scarce.

Overall my guess is that all of these are going to effectively increase the amount of parking available by 4-5x. Parking simply won't be a problem anymore, unless of course this induces demand, we build higher, and without parking garages...

I agree. But a lot of people want to own the car because they want it to be clean and not used but others in a way they wouldn't.

> Why would I even bother "owning" a car?

Talk about superficial thinking.

Half the time I get in the car it's for a spur-of-the-moment run to the store.

And while the bus stops 300 ft from my house I've never ridden it.

I don't want my car off earning money when I need it for something important like running up to Home Depot so I can finish whatever project I'm working on.

"The ideal state of a resource is idle and ready for work."

> I had to stop reading because the blog poster's level of thinking is really superficial.

And why would that bother you to the point of having to stop reading? Because if it causes rejection on an emotional level, I'm more inclined to think the discomfort is a consequence of it challenging your worldview. Which is likely one of driverless cars being the magical technological cure for all transportation problems. Especially since your suggestions, at least when formulated as briefly as this, are demonstrating quite a bit of superficial thinking in themselves:

> So instead of worrying about parking, you order a car, and it picks you up, and then you get dropped off. It drives off and services other customers. Then, when it comes time to get driven home, you order another car that comes and picks you up.

A the car is being shared without supervision, and somehow this will all magically go right, because vandalism doesn't exist in public spaces right now while we still have human supervision either, right? As Nathan Pyle once wrote about NYC public transport: "Beware the empty train car, it is empty for a reason."

http://also.kottke.org/misc/images/nyc-tips-etiquette.jpg

I mean, come on, the tragedy of the commons is ancient and well-established economic theory by now:

https://en.wikipedia.org/wiki/Tragedy_of_the_commons

> You only pay for the car while it's driving you, not while it's sitting there doing nothing.

Except that in the scenario you suggested cars are never "doing nothing," they are constantly driving around. Even if they have no customers. You suggested we wouldn't bother with parking, remember? So we'll get into a scenario where swarms of empty driverless cars are clogging up the road waiting for customers.

And this can realistically happen if we don't pre-emptively regulate this, because we'd have competing different companies with swarms of cars who don't want to share their parking infrastructure for "resting" cars, so they'll use the shared free common infrastructure (the road). Tragedy of the commons again.

> And why would you go to Starbucks to pick up your own coffee? Why not send a self-driving car through the special self-driving car drivethru to pick up the coffee and return it to you? It could wireless transmit the order and payment details for you, and have a special slot for the coffee or food order.

People don't bother with food delivery most of the time right now, why would the presence of driverless cars change this? More importantly, why would sending an entire car for a coffee be a good thing? What you describe is a form of Jevon's Paradox, where increases efficiency (or in this case, convenience abstracting away (externalising) the real cost from the user) actually creates more wasteful use of a resource

https://en.wikipedia.org/wiki/Jevons_paradox

But don't take my word for it, how about Mark Hallenbeck, who is the Director of the Washington State Transportation Center (TRAC)?

http://humantransit.org/2015/11/self-driving-cars-a-coming-c...

Or this Elseview publication on how automation likely increases reliance on cars, creating more traffic and energy consumption:

http://www.sciencedirect.com/science/article/pii/S0965856415...

> Seriously, the article is really superficial thinking when it comes to self-driving cars.

This isn't an argument. You're just asserting that we shouldn't even bother reading it. You are basically shutting down opposing voices before they are being heard, because they go against your own preferred line of thinking. And I have serious issues with this attitude, since we have more than enough of that in [current year] as is.

> Why would I even bother "owning" a car? What the more reasonable approach is time-sharing of multiple vehicles, like Uber.

A huge problem remains though : how do you avoid having dirty/smelly/broken cars all the time when everybody is sharing cars ? People are usually way more cautious with there own stuff than with others'.

The city bike-sharing systems around the world usually chose «indestructible» bikes to deal with this problem, but as a result you get a really heavy and uncomfortable bike and I don't think you could do this for cars anyway: no matter how robust is your car, spilled drinks or food on the seats are still a pain … This kind of system definitely needs a way to avoid the tragedy of the commons.

Elinor Ostrom[1] showed that there are mechanisms to regulate this efficiently, but you need to actively deploy them and it's better not be an afterthought.

If you do nothing and let people be toxic human beings, you'll end up with shared cars in worst shape than public transport (because in public transport you have a kind of social pressure that limits the bad behaviors)

[1]: https://en.wikipedia.org/wiki/Elinor_Ostrom

They could put a button on the car to report dirty/smelly/broken car. When pressed, the button would let the car know that it should go to some place for cleaning/maintainance soon and also, depending on how comfortable we are with the required surveillance, fine the previous user of the car either for spilling the drink of for not reporting it.

This is very interesting to read from a UK perspective. We don't have the concept of jaywalking, except for on motorways (I was actually shocked when an american friend explained it at uni, how the hell does anyone get anywhere, then I realised they mostly drive). So these social interactions don't just occur in certain areas, they're widespread across most roads and pedestrians technically have right of way over cars. Could autonomous vehicles cause the end of walking places?

At some point, we will debate against manual driver cars. As a petrol-head, I love driving and riding but that is the truth. the sooner we get rid of manual-drived cars, the faster we switch to self driving cars because the existence of manual cars are a big problem for self driving cars and they limit the potential the self-driving cars can achieve. We should at least do that in some cities. Imagine a town with no traffic lights and cars move in a harmony and scary precision in traffic.

Imagine a power outage. Imagine solar flares. Imagine a town with no backup plan for complex systems suddenly encountering events with 1 in a million odds.

One aspect that is not covered is armed conflict (or terrorism). Imagine what would happen if enemy take over a fleet of autonomous cars, be it through hacking or by physical attack on a control center.

Another thing: imagine robbery with autonomous robots and cars.

I am waiting for Hollywood to catch up on these ideas - this would be very cinematic.

Does the car need to be completely autonomous for that scenario to be a huge issue. Isn't completely controllable through on board computer bad enough? Someone could remote control and thus weaponize such a car already, could they not?

Yeah - indeed, remote control would be enough. There is a problem with latencies though - so you need at least some autonomy in the car. And then there is also the thing that if it is remotely controlled - then it is easy to stop it - just jam the communication link.

Eliminating the passenger lets you make the tool more expendable. The attacker can control more than one and afford to lose a few.

Most of #2 sounds fascinating and not anti-social to me at all.

Except, very few parents are gonna send a car and forgo helicoptering over their kids. Isn't having kid unattended in car illegal most places?

> taking away much of the rest of advertising revenue from print, radio, and TV

TV advertising revenue continues to grow, and is still larger than online ad revenue.

Bait title for blog hits. These aren't unexpected consequences, they are some of the specific benefits. It adds nothing to the discussion as it's neither comprehensive nor insightful. Why does the author think autonomous cars are a net good? Thanks, he went on and on about a few things he wants to pretend are "unexpected". SMH

I think that driving in places such as India with little traffic rules might be easier for an AI because the traffic flow is self-organized. There might not be any rules, but that is not relevant as we are not talking about expert rule based systems. If the AI car sees a cow it'll just slow down and move to the side to overtake.

I live in this area, completely agree, author is spot on.

Social mobility is a major problem in democracy http://www.bbc.com/news/uk-politics-24936416

Reply on news.ycombinator.com