Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
If You Think the Cabbies Are Mad Now…
Buckle your seat belts, everybody. We’ve reached peak disruption: a story of the gig economy intersecting with the rise of the robots. From Thomas Lee in the San Francisco Chronicle:
From taxicab unions and package couriers to politicians and regulators, a growing crowd of people would like to destroy Uber. Add one more name to the list: Uber founder and CEO Travis Kalanick.
Somewhere lost in the scrum over whether Uber drivers are employees or contractors, or whether the company conducts proper background checks, is the simple fact that Kalanick wants to eventually replace all Uber drivers with software and computers. Like Google and Tesla, Uber is trying to develop a car that can drive without a human operator.
The piece, of course, dedicates a fair amount of time to what Lee perceives to be Uber’s hypocrisy: the company’s current PR strategy partially revolves around arguing that it offers drivers a better deal than cab companies, while its long-term strategy would remove drivers from the equation altogether.
I’m inclined to go a little easier on the company then that. If they think this is the way the market is inevitably going to move, I don’t know that they have much of a choice. How are you going to compete if your price has to include labor costs that your competitors aren’t shouldering?
What’s more interesting to me is the fact that the vision here is bigger than Uber. Indeed, what Kalanick and company have in mind is nothing more than the destruction of the idea of personal automobiles. From the Guardian piece by Alex Hern linked in the quote above:
While a full move to driverless cars would be Uber’s dream scenario, letting it cut the cost of a ride to little more than fuel plus wear and tear, it could very well be a nightmare for car manufacturers. Self-driving cars could prove the death-knell for private car ownership, with services like Uber offering a cheap substitution while avoiding the wastefulness of leaving an asset worth thousands of pounds sit unused on the side of the road. A self-driving car can carry someone from home to work, head off to a different office and pick up someone going to the airport, even take a package in the boot to be delivered to a client – all while a conventional car would be sitting in its owner’s car park.
Perhaps because of that, the focus from the conventional auto industry has been less on driverless cars, and more on using self-driving technology as a safety feature to augment traditional driving. For instance, a number of cars already on the market are able to maintain a steady cruising speed, stay in lane, stay a safe distance away from cars in front, and even park themselves, all without human intervention. In a patchwork fashion, those cars could eventually build up to almost full automation – but the signals coming from the industry indicate that the final step might be something they are loathe to take.
I own one of these “AI vehicles,” and I have to admit that this technology is very cool (though not yet reliable enough that you would rush out to get the fully automated version). Still, I’m conflicted — and I wonder if you are too.
I’m largely receptive to the intellectual case for driverless cars. But I’m emotionally resistant.
Now, granted, I’m an outlier when it comes to romanticizing driving. In just the past year or so, I’ve piloted my SUV through 42 states. For tax purposes, my official residence is the interstate highway system. But I don’t think you have to be as much of a road warrior as I am to be a little uneasy about this. (Please note before you jump into the comments that this in no way means I think we should attempt to arrest technological progress on this front.)
Here are my questions: Are those of us who feel like we’d be losing some ineffable freedom by having the wheel taken away from us just sentimental luddites? Should we just chill out and await the brave new world of driverless cars and hyperloops? Will personal automobiles become indulgences rather than necessities? For that matter, to what extent can they coexist with driverless vehicles?
I realize I’m getting perilously close to “Get off my lawn” territory. I’m just hoping I’m not the only one.
Published in Economics, Science & Technology
I’ll bet you that driverless cars will be widespread in 20 years, Hanson.
Pepper knows what you want. Pepper always knows.
Get off my lawn, Whiskey!
I was wondering when you were going to show up! Ever since I first read your writing on this topic, confirming my instinctual skepticism, I have been telling people “self-driving cars are the new flying cars”. It’s a pithy way to express a whole host of complex technical and social barriers and helps many people understand the problem quickly, even if they don’t really understand the technology.
WITNESS! Winner pays the car.
This is where it will be fantastic to free ride the system. All of those people taking driverless cars are finally out of my way in the left lane. If I decide to cut you off, your car is going to be programmed to brake to avoid the collision giving me total control of when I want to change lanes or merge for construction. Passing on the shoulder will be the new thing because those driverless cars won’t get mad and speed up to block me.
How will people know how to take over if they never drive anywhere themselves? Are we going to require a special off-roading license now for those who never drive themselves otherwise?
Except they aren’t. Segways, and Google Glass, and Flying Cars are solutions to problems nobody has. Driverless cars are simply the next step in the evolution of transportation.
The reason there will eventually be no manual override is because that will negate all the safety gains driverless cars will achieve. If humans can drive the car anytime they want accident rates will still be too high because human drivers are the cause of 99% of accidents.
That’s why they’ll make you take the driveless bus, WS
To give you an example of one big problem we simply have no idea how to solve, consider ‘moral judgement’.
Here is the scenario: You are driving along with a passenger, and a child steps into the road. You don’t have time to stop. But you could spare the kid by swerving into a parked car, perhaps killing everyone in the car. So what do you do? You could just hit the kid and kill him. Humans are socially ‘allowed’ to make this choice if no other option was available. Or, you could swerve into the parked car and kill your passenger. People would understand that choice as well. No one will blame you whichever choice you make, so long as there was no option for a good outcome and you did your best.
But how about a computer? If it is faced with a no-win scenario and makes a choice, are the people affected negatively by that choice going to say, “Oh, I understand why the computer chose to run down little Bobby. Its algorithm said that sacrificing Bobby led to the maximum utility.” I don’t think so. I think it’s much more likely that they will claim that a soulless machine was programmed to kill Bobby, and seek restitution. And lawmakers would be jumping up and down calling for driverless cars to be regulated so they couldn’t do that again.
These types of decisions require a lot of attributes that we simply have no idea how to give to computers. Emotion, empathy, a sense of duty, a sense of right and wrong in a world of ambiguity and fuzzy choices… These decisions are hard enough for humans, and when humans get them wrong we tend to understand. We won’t be nearly so trusting or understanding of machine failures. And they WILL fail. And sometimes there will be fatal consequences. We have no idea how that will play out in terms of public acceptance of the technology. My suspicion is that it won’t turn out well.
As a practical matter, even if we solve these problems truly automated cars will appear very slowly. The average auto lasts more than a decade now, so even if an automated car showed up tomorrow it would take several decades to convert the auto fleet. We aren’t just going to throw away our perfectly good cars.
The most complex and expensive part of the autodrive system is not the hardware, it’s the software. The examples you gave of miniaturization are the result of advances in hardware mass production and size. The autodrive software, if it is to have any serious capabilities (other than simple tasks like adaptive cruise, lane control, and autopark), will only grow in complexity as more and more failure scenarios are realized either through testing or real world accidents. Its development and testing is an enormous engineering investment with all kinds of insurance, liability, and human safety overhead costs added on top. I don’t think the the analogy holds.
I’ll see your 20 and raise you 15. Seriously Dan as much as I respect you my parents have a book from the 70’s that predicted a computer would never beat the world’s best chess player. The only exception to the rule of “that’ll never happen” happening seems to be nuclear fusion which remains 50 years away.
And so you agree that it’s a terrible thing for individual freedom and autonomy, right? You’ll never be able to go exploring down mountain back roads, or maybe not even be able to drive around town when a blanket of fresh snow covers the lines in the road.
Right but a lot of people may just leave their car at home when they do the math and find out it costs them half has much to commute by self driving car.
OK, but in a driverless car scenario the car would see the child sooner and communicate to the other cars what it intended to do. Potentially leading to no damage at all.
Mark why do you insist on a binary scenario? There will probably always be cars that can be driven by humans, but unlike you a lot of people don’t want to drive. For those people there will be cars that will do everything for them, for people like us who like to explore every now and then…..there will be options.
But why can’t I do those things? Why can’t I just say “Hey, car. Do these things.”?
I’m not averse to a bet, but we’d have to define the terms. What do you consider ‘widespread’? Say, half the cars on the road being totally autonomous? Or are we just talking about taxi fleets? Or what?
JUST SWALLOW THE GOLDFISH!
I could agree with you about Google Glass, but the Segway and the Flying Car are potentially VERY useful.
Let’s say these technologies were as cheap as their alternatives and had no other drawbacks. Are you telling me that it wouldn’t be useful to lift off your driveway, fly to work, and land in your parkade? It wouldn’t be useful to have a mode of personal transportation that can go 300 mph yet still land at your house and directly at your destination? Or that it wouldn’t be useful to have a car that could sprout wings on the highway and suddenly lift off and fly you directly to your destination at 3 or 4 times the speed of a car?
The Segway was targeted at a very real gap in our transportation – the first and last mile. People who use public transportation have to walk to their bus stop, and walk from where they are dropped off to their destination. A look at any city core in the daytime will show you thousands of people trying to get around at the same speed they did a thousand years ago. It’s not a stretch to imagine that a device that would let them travel two or three times as fast using the same infrastructure might be very useful indeed.
Furthermore, 80% of all driving is done within a a few miles of your home. A device like a Segway would allow you to make that jaunt to the corner store and back much more quickly and efficiently than walking or driving the car. So there was definitely a need for something like a Segway.
The problem with flying cars is logistical and technical. They’ve never made sense, and never will, even if would be great to have them. That doesn’t stop the media from predicting their imminent rise.
The problem with the Segway is more one of social response. The device works exactly as it was said to work. It’s just that it turned out to be socially unacceptable. The ‘dork’ factor is too high. That’s what sunk Google Glass as well. They’re not cool.
But being ‘uncool’ is not an arbitrary assignment. We decided Google Glass was uncool because of the privacy implications and the in-your-face pretentiousness. We decided that a Segway was uncool because it was dangerous on the sidewalk and people riding them used more than their share of the public space. These were not technological issues – they were issues having to do with whether the technology fit within the boundaries of social acceptability.
Whether an automated car will survive social acceptability tests is an open question.
Who buys the cars? Someone has to own.
Useful in that they do stuff but not in the entrepreneurial sense.
There just aren’t that many people who say “This car is very useful but I hate that it touches the road!” or “I really like the idea of going over to that spot over there but the moving of the feet! It’s too much!”
Dan Hanson
Flying cars ain’t going to happen. I agree that George Jetson’s car is very cool and what we all want, but not going to happen in our life time.
The car owns itself. When it is about to die it trades itself in for a new self.
Who pays to build it? Does it collect micro pennies from those who use it?
You have a point, in some sense we are arguing past each other. My main target is the idea of ubiquitous driverless cars with no human controls, which is the fantasy scenario of many who envision a future without traffic jams, or even without private ownership of cars, which is part of Troy’s original question.
On the other hand, your comment to which I replied (“there will be no manual override”) seems to fit that description exactly, predicting no human driving whatsoever.
I love it when you make my point for me. You develop a piece of software. You sell ten copies of the software. You have to split the development cost over ten copies. You sell ten million copies. You have to split development costs over ten million copies.
How much does it cost to produce a piece of software? (Development cost / total sales) + $5, for the storage media.
The cost will initially be sky-high, then tumble to almost nothing once development costs are amortized. Why not? You get a huge return on the cost of the media holding the software once you pay development.
Seawriter
See my comment 33 on mountain driving.
My point with the snow was that the cars rely on certain fixed indicators of road direction which would not be visible. Freshly fallen or plowed snow looks different every time and would be impossible to accurately map and preload into the car’s memory — which is a requirement for Google’s cars to function properly at the moment.
My point was only that your analogy to transmissions and personal computers was flawed. I also think the nonrecurring development costs for the software will be more than enough to make up for the cheap recurring costs. It’s going to be the most complex software ever developed for a consumer product — more complex than even the Space Shuttle flight software, I would guess — with the added burden of being safety critical.