Here’s what it would take for self-driving cars to catch on
washingtonpost.com
washingtonpost.com
Bad press. Being killed by a machine is soo much worse than being killed by a fellow human. This won't be a problem. Yes, it will make the news, but if self driving cars turn out to be more reliable than human drivers, the numbers will show, and the insurance companies will follow. (Funny how talks about not controlling your car will die out in the face of monetary savings.)
Liability. Self driving cars are not the only thing that kills people when it goes wrong. Think airliner software. I think we'll get over it. Also, current cars already face similar challenges. I recall a sticky pedal problem that caused some Japanese cars to speed up until death ensues. As far as I know, unlike some unlucky drivers, the company is still alive.
Law. Not a big problem, as long as self driving cars are legal somewhere. Once they have demonstrated their worth, other states and countries will follow very quickly.
Privacy. It kinda sucks, but it won't matter one bit. If people really cared about not giving away their trip data to some for-profit corporation, they wouldn't be using Facebook in the first place. Or Gmail.
Price. It will inevitably go way down. Self driving is a software problem. A human driver could manage with only a couple web cams. So will programs, as they get better. Not to mention economies of scale.
Unemployment. Here be Unions. But they won't stop the advent of self driving cars. Driver-less companies will simply drive the Driver-full companies out of business. Unlike rail-road companies, they don't own the tracks. Now this will raise unemployment unless we do something about it. (Enforcing a 4 day work-week would be a good start.)
Overall, I expect all those problems to set us back about a decade, 15 years tops. But they won't even arise until we actually start selling those cars.
Also enforcing a 4 day work-week? How does this lower unemployment, except on the paper? And keep in mind this solution would come with it own bag of problems, like administrativ overhead for more employees etc.
I only have so much space here, and I only have so much time to actually think these things through. So I'm just going to ignore this fully general counter argument.
> And saying that the problems won't even arise until we start to sell these cars is just nonsense.
Let me put it this way: if no self-driving car is ever sold, these won't be a problem. These problems are conditional on a working self-driving car actually existing in the first place. The only problem right now is getting the car to drive.
Now of course, foreseeable future problems do influence current decisions. But let's not get into Timeless Decision Theory.
---
> Also enforcing a 4 day work-week? How does this lower unemployment, except on the paper?
Why wouldn't it?
Speaking for about 400 actual companies of various size in France, the slight increase in hourly productivity does not match the rather sharp drop in hours per week per employee. To keep their total production from going down, they had to (and did) hire more people (about 15%, I guess, it depends). It meant more employment. In the real world. Now.
Also, the salaries were maintained. No pay cut. And the costs did not increase either. This is because they have a deal which allows them to pay less taxes (specifically, those who go to our national unemployment insurance). So, so increased costs and no pay cut. In the real world. Now.
The only thing that's only on paper right now is the macro-economic calculation: what if everyone did the same? So far, it checks out: it doesn't cost nor bring any money. You just have more people working a bit less.
---
There is this saying that employment is not a cake, that more work spurs more employment or something. But at any given point in time, it is a cake. Basically, work equals demand divided by productivity. In a consumption society, to have more work, either you increase demand, or you lower productivity. There's limits to the first, and the second one is Ludism. So there's a point where you have to divide work among the workers somehow. You can let the market deal with it, and that gives you unemployment, or you can enforce shorter work weeks.
Unemployment is bad, because it drives down demand, which ultimately drives down work, which under the current system increases unemployment… One of the causes of World War II by the way.
Shorter work weeks on the other hand, provided there's no pay cut, drives demand up, through the sudden decrease in unemployment. Also, less unemployment means better salary negotiations, which means more money to the workers, and less to the "stakeholders". (Mind you, it can go too far and stifle investment. Currently however, workers get 10% less of what they did 30 years ago. And that money don't tend to go to investment. It mostly goes to the Global Casino.)
---
Source: http://www.roosevelt2012.fr/ (In French, sorry. I hope automatic translations are good enough.)
The labor surplus is caused by decreased scarcity. Prosperity has decoupled from labor. This is not a problem! Even if it was, creating an artificial shortage of labor would not be a sensible.
Not exactly. The number one factor is hourly productivity, which more than doubled since the 70s. At the same time, demand also doubled, just not as fast (there are a few missing %). That difference is where labour surplus mainly came from.
> This is not a problem!
If your an employer, sure: this is an employer's market. But ultimately, this drives salaries down, and consumption, then employment. There's a reason why Ford advocated the 5 days work-week when everyone else went for 6, and why he paid $5 a day when the norm was $2.5. He wanted his employees to buy his cars.
> artificial shortage of labor
We're already under such "artificial shortage". Minus the shortage, of course, but we have vacations, 5 day work weeks… If you let the market choose, you would soon have many people working 50-60 hours per week for less than the minimum wage. (I mean, we would have more of them.)
I just want to push the cursor further on the "artificial" side.
Currently everyone has to buy cars that will be suitable for any journey they might have to make. This means that people commute to work in cars that can take a whole family on holiday. Self-driving would mean you could always get the most suitable car for the journey you're about to make. Perhaps you're feeling tired and want to sleep on the way to work? Maybe you want to do some work so you'd like a desk? Put your feet up in an entertainment centre?
Self-driving will become popular because it will mean the end of wasting hours of your life watching the car in front of you. Just think of the ads, what could you do with an extra 3 hours a day?
I agree with everything you've said but the economics will be very important. If the cars are more expensive than ordinary cars getting back 3 hours a day could be enough to justify the cost. But how long you drive per day obviously depends on where you live. Most people I know (and I live in the suburbs) don't commute more than 1 hour total per day. They might not be able to justify the extra cost of a self-driving car.
One weird effect of effectively using cheap robo-taxis for all our transport is that it will put the cost of each journey right in your face. I imagine this will have all sorts of weird effects on people's habits. Right now you spend the price of a car upgrade every few years, taxes & insurance once a year and petrol every week or two. You don't really feel the cost of a supermarket trip. If you drive your niece to the airport, you don't think about the cost at all. A drive to the woods with your family counts as free even if you use €20 worth of petrol and €20 in depreciation.
Governments might be more inclined to get the full cost of roads from transport companies one step removed from the public than directly from them.
The effects of a new mode of transport are hard to predict. Few people would have guessed that family cars would lead to supermarkets & disneyland.
Not sure if this is what you meant, but I read it as: You can bang a lot of 'product information' into someone you hold captive for three hours. So this could be a 'interesting' way to finance self driving cars. You play ads inside the car, and the customer has to pay to switch them off.
Addendum: I don't like this future, they promised me rocket ships, and all I get was a cell phone full of spam.
1) Some roads can have sensors embedded in them. People would have to get an account to drive on them (eg EZpass or SelfDrivingCarID). The sensors would recognize a self-driving car and allow it on the road. Those roads could have 200mph speed limits. Cars attempting to enter the road which arent self-driving, or too old of a version, would not be able to pass the tollbooth, and will be fined for attempting to and wasting everyone's time at one of the booths.
2) Parking tickets and metermaids should go the way of the dodo. Cars can easily be detected at the side of a curb, and the money deducted from the EZPass or whatever account. Instead of parking, though, you'd just timeshare your cars in a demand based marketplace (type of car, time you need it for, what the demand is for those hours etc) and then clean them up when you get out. They'll go on to pick up the next passengers.
The middle class will have their own cars and use high-efficiency parking. It will be no hardship to use a short, tight slot six blocks away.
The upper class will have their own cars and good parking. Some will even have "drivers", because babysitting a fancy car is one of the less unpleasant ways for the "driver" to earn his negative income tax credits.
I am extremely curious how far off the technology is to drive on less than predictable roads, such as narrow streets of London during rush hour or in the small villages of Yorkshire in the UK.
I love how the article refers to lives saved, whereas I always said, "we will allow computer driven cars when the number of accidents/lives lost will be less than those caused by humans". It goes to show that the nature of driving is now out of anyone's hands and we accept there are fatalities, akin to disease. With this acceptance is the race to save lives, rather than have preventable loss.
>> And roughly 40 percent of fatal accidents are caused by alcohol, distraction, drugs or fatigue. Letting robots take the wheel would save lives.
The part that I find interesting is that if you are in control over whether to drive with alcohol, distraction, drugs or fatigue (at least for the car you travel in); you can affect your chances of being in a fatal accident. With computers at the wheel, you don't have control and accidents become the lottery of driving. Unless of course, control over your fate is now down to whether you install the latest firmware update on your car.
I do believe that mainstream driverless cars will happen and future generations will look back at human driven cars like riding horses - mostly done for recreational purposes and not for long distances.
I don't know which aspects of roads you are considering, but there are few unpredictable roads nowadays, thanks to street view
I'm referring to really tight corners which you can't do it in a single turn and have to reverse back, all while there is potential for oncoming traffic.
I'm referring to single lane country roads with passing points every few hundred yards and having to reverse back to accommodate a vehicle or tractor coming the other way.
I'm referring to blind corners where the only way to know if it's clear is to watch for no cars between the visible gaps in a hedge or between cars.
I'm referring to the guy in the middle of the road with the road works sign saying waving you to go or to stop.
I really mean the less than predictable conditions on the roads.
Stuff like this:
Cars on left: https://www.google.co.uk/maps/preview#!q=holmfirth&data=!1m8...
Sharp turn: https://www.google.co.uk/maps/preview#!q=laithe+avenue&data=...
Pulling out onto this road: https://www.google.co.uk/maps/preview#!data=!1m8!1m3!1d3!2d-...
I'm not saying it's impossible, but I know it's not easy.
I'd expect them to do better, for the most part. They're able to take in significantly more information than humans so rushed, cramped situations is where they should shine. One thing I think they'll do 'worse' than the drivers in London is driving straight at pedestrians who cross when they shouldn't. If it stops for anyone, everyone will cross. They need to angrily drive straight at people and honk their horn.
The only two places self driving cars have a future are:
1. Planned communities which ban the use of personal cars inside that community. 2. An interstate highway system where there is a special lane dedicated to self driving cars.
If google is serious about self driving cars, they need to buy or build a town... or use this one.
http://theweek.com/article/index/219043/the-ghost-town-being...
[edit]
The best hope for self driving vehicles is as an anti-social public(or public-ish) transportation system.
Retrofitting all roads with sensors is just not an option. You can't fit all the small dirt road in the country with sensors, and neither can you have a car that refuses to go on because it can't talk to the road.
What ever humans do - reading the map, studying the surroundings, estimating the pavement/conditions, building on experience - a machine (with the right sensors, carefully designed algorithms and testing) can do much, much better.
Actually, how do you face any particular situation X? Why do you think the machine can't handle situation X?
Can't handle every situation, you say? Then just put the sensor's on manually driven cars, and build a corpus. It won't be foolproof, but I bet my arm it can be much more reliable than a human being.
---
Your objections only work with current technology. One day, you will see that self driving cars have less accident than human driven cars. And soon you will be faced with a tough choice: let the car drive for you, or pay more money.
And why would the roads need sensors?
I do believe it would benefit self driving cars to have areas where only they are allowed, and in the future I foresee that being the case as the cars can drive faster and closer together if not driven by a human. But this shouldn't be a necessity when cars gets more advanced.
Progress is fast and if history has showed us anything it's that history is impossible to predict. About 100 years ago flight was thought as an impossibility until the Wright brothers proved us wrong and look where we are today.
In 100 years what role will self driving cars have? I'm certainly leaning toward it being mainstream, even considered normal, to see them everywhere.
Then again, some would say we would have flying cars by then.
A B C
Car A is driven by a human and steers into car B. If car B could communicate with car C it could ask it to move over (or slow down) so that it (car B) can move out of the way of car A.
If A B C are all driven by humans, C could move over or slow down if it sees that B is steering into it. But self driving cars could do that as well.
That self driving cars could communicate is an argument for the self driving car, not against it as the parent made it out be.
Furthermore, special infrastructure is unacceptably vulnerable to deliberate disruption. A system that depended on radio communication could be gridlocked with a modified $50 microwave oven.
They know the road and where they are, they can see what's around them and how fast they're travelling.
For humans to beat machines at driving, we need to consider situations that require complex reasoning that humans are good at. Essentially, for any "But what would they do if X" you need to say what humans need to do, and why it's a complex task.
Whatever you can use to find out where the road is supposed to be, the self-driving car can use as well.
There will also probably be databases of landmarks like buildings and light poles that can be used to make a centimeter-accurate visual positioning system.
The vehicles' outlandish appearance doesn't bother me because I'm a long-time advocate of self-driving cars, but it may be an issue for other consumers.
Other car manufacturers such as Cadillac are working toward fully autonomous vehicles that look like stylish, regular cars. The computers and sensors are hidden in a spare tire compartment.
I wonder if Google is underestimating consumers' preference for traditionally good-looking vehicles. I think about how the company was criticized for underestimating the aesthetic roadblocks involved in the Google Glass rollout.