Finding driverless-friendly cities that allow car makers to test their technology has been one of the challenges facing the driverless industry.
Today, Portland stepped up to the plate. “The technology is coming,” says Mayor Ted Wheeler. “Either the technology will happen to us, or we are going to shape the playing field.”
Bloomberg reports Portland Opens Its Streets to Self-Driving Cars.
Autonomous vehicles need to drive and drive and drive, vacuuming up hours of real-life encounters on the road to make their algorithms smarter and safer.
But there’s one thing in relatively short supply: cities willing to have test cars on their streets. Portland, Oregon, is trying to change that and be what it says would be the first to issue permits for driverless vehicles, with the goal of getting them on its roads this year.
“The technology is coming,” says Mayor Ted Wheeler. “Either the technology will happen to us, or we are going to shape the playing field.”
The city will solicit proposals from companies working on driverless cars to gauge how they can help Portland reach its goals of reducing carbon emissions and providing equitable service. The city would also consider providing financial support for businesses to test autonomous transit vehicles, such as shuttles or buses, that could potentially connect passengers to its existing transit infrastructure. Wheeler says two years of pilot testing would inform final rules: “If we wait five years, my concern is we are not going to have a say in the matter at all.”
General Motors, Lyft Inc., and Daimler AG were among those that wanted to partner with Portland on autonomous transportation as part of Portland’s submission for the U.S. Department of Transportation Smart City Challenge last year.
If the city’s plan works out, driverless cars can learn more about maneuvering in the rain while dodging Portland’s many modes of transit, from light rail trains and buses, to pedestrians and unicycles.
Most Disruptive Force Since the Internet
Some of my readers think I am obsessed with this topic, and most of those who do, still have their heads buried in the sand believing driverless technology is decades away from happening.
By now it should be crystal clear the nay-sayers are all wet. Driverless is going to happen, in a major way, by 2022 at the latest.
I write about the topic because driverless technology is the most disruptive force since the internet. It will bring about major changes in the car ownership, the way people commute, and the need for insurance.
Accident rates will plunge. The gas station model and the car dealer model will have major changes as well. Instead of buying two cars at $20,000 to $40,000 each, families will share a single car. A growing percentage those who live in major cities will choose to not have a car at all.
Instead of buying two cars at $20,000 to $40,000 each, families will share a single car. A growing percentage those who live in major cities will choose to not have a car at all.
Mass Adoption When?
Driverless long haul trucks will be the norm within two years of interstate approval. I expect approval by 2021 at the latest.
In contrast to trucks, mass adoption by the general public will not be immediate, but it will progress rapidly within a decade.
Portland now leads the charge in testing. Competition between cities is sure to heat up. Expect another dozen announcements this year.
A Study Says by 2030 1/4th of Miles Driven will be Driverless. I expect 85 percent of miles driven will be driverless by 2030.
For further discussion, please consider Second-Order Consequences of Self-Driving Vehicles.
Mike “Mish” Shedlock
I hereby call to order the Never Going To Portland Society. Have at it!
Portland is barely a city, more like a large suburb. I have spent a lot of time there, and like it. But driving conditions will in no way be typical of cities…
Other cities will quickly scramble to adopt a plan.
Mish,
think it’s important for your readers to become familiar with the differences between level 3, level 4, and level 5 SD vehicles
as I understand
level 3 is very widespread today millions of cars
Level 3 1/2 is what Tesla installs today
level 4 what everyone is beta testing these days, likely what the Portland project will be. Also
what the Pittsburgh taxis are, the LA harbor trucks
level 5 currently in alpha testing
level 4 will be useful for
1) hub to hub trucking
2) taxis and for hire vehicles
3) delivery routes
Tesla is likely going to transition from level
3.5 to full fledged level 4 by end of 2017.
level 4 is real and will happen quickly
I was in Portland a few years ago on one of the trolley-buses, and the driver had to stop the bus on the narrow downtown street and get out to physically move a truck rear-view mirror out of the way so that the bus could pass without hitting it.
I’ve been in Portland 25 years, and after several years of mass immigration to “Portlandia” and a few decades of the environmentalist blockade of any meaningful additions to highway infrastructure, traffic is now as terrible as any other city. Try a 7am commute here some time.
We recently drove through LA on I-5 round trip Anahiem to Universal Studios during rush hour. This was proof that no matter how many lanes are added, traffic will not move like it does at 2am. We need to spend money fixing the existing lanes so a set of shocks lasts more than 5000 miles. This is especially true of the freeways we encountered around any large city on our recent cross country trip.
Urban, particularly LA, gridlock, is mainly just another predictable side effect of the same zoning and regulatory and “planning” idiocies that is the cause of much of today’s miseries.
Always some tax feeder on lobbyists’ payroll “planning” where others shall drive. Or, more relevantly, where they shall not drive. Making sure that noone drives where the well connected lives. Instead, the yahoos are busy installing speed bumps, road blocks, gates and permit parking to make driving as difficult as possible. Then complaining that it’s too difficult…..
I’m not sure what part of Portland you visit, but it certainly is not a giant suburb. As a driver the cities two biggest problems are pot holes and streets designed for a different era. (Way too narrow.) Many of those “suburb” streets don’t allow for two cars pass each other.
I follow this issue from afar. It would seem that electric cars – the current trend – could conceivably stay on the road for decades beyond the current internal combustion engines. For one, there are fewer parts in electric cars. Also, the powertrain will not wear out; only the batteries and stuff like brake pads and tires will need to be replaced semi-regularly. So what happens to the auto manufacturers? Car companies remain a big part of the economy, They depend on a business model driven by internal combustion engine powered vehicles being replaced every 10-15 years and a steady stream of advanced features fueling demand.
I think the solution is that they morph into software companies. With the steady state of impending autopilot deaths and accidents from robot cars, there will be plenty of work upgrading the software. Imagine Dearborn, MI becoming Silicon Plains.
“Mass Adoption When?”
Probably when the chickens come home to roost. My guess is not for decades and after most of us have turned to room temperature.
None of the unintended consequences have been considered. We have lots of denial going on. It’s part and parcel of the human condition.
This old timer agrees with you, LF. Just take a look at the number of sensors on the ‘self-driving’ car in the photo, just to try to do what Joe Public can do without much effort or cost. Then imagine the software that analyses all that real-time data, then the compute power to run the software, then the controls that allows the car to respond to the command signals from all that technology. I gave worked in technology for over 50 years and lately on the integration of ‘automated’ mining trucks into the vast amount of IT systems that the typical large mining company uses to run its operations, and they have hardly started to come to grips with basic issues like how the vehicle should react when it develops a mechanical fault. The next generation (or two) of techies will have to figure this out, it certainly won’t be mine.
I can see how this will eventually work in large metropolitan areas. But I can’t imagine how it would work in the kind of suburbia where I live. I have two kids and a job, and on a normal day I make 4-6 trips easy. And many of the trips I make are at rush hours, when presumably everyone else would want a car too. Yet everything I do is extremely time sensitive, so it’s not like I could wait for a car at peak times. And it’s not like my trips are easy straight a-b kind of things… I have to drop one kid off at school, then the other at another school, then I have to stop at the store for office supplies, then I head to work- and that’s just one trip. And there are 10,000 people in my small suburban town doing the same kind of crap at rush hour. What happens to all the extra cars needed for rush hour, when it’s not rush hour? How is that any more cost effective than just owning the car?
why don’t your kids walk to school ?
Because the schools are miles from where we live, with no sidewalks, and they are four and nine years old?
Do you notice that it’s the liberal that are the first to roll out the autonomous vehicle programs? (ie, San Francisco, Portland, etc…)
That should be your first clue that something is awry.
More denial in play.
You noticed thee very big clue to the whole scheme. There was no original consumer demand for the autonomous feature and there is a reason for that. That’s because the original demand was not set forth by the private consumer, it was set forth by tne government and they are funding the hell out of it because the auto makers would never go out on a limb to find such an emdeavor. You can trace it all back to the UN and politicians being coherced into supporting the scheme.
Thanks for joining us Mr. Alex Jones, happy to have you here at MishTalk.
You say there is no original consumer demand, but I beg to differ:
I would much rather ride to work in a self-driving car, and be able to sleep, read, or do work.
It would be great if my elderly relatives could ride to see me more often without sitting on a a slow and uncomfortable plane or bus.
I would be more than happy to participate in a movement that would end the business of personal injury lawyers, car insurance, auto body repair, and all of the economically worthless industries built on the broken window fallacy.
If I could ride to work in a car service I would much prefer to live farther out, where I could afford more land and a larger house, and not be under the thumb of my current city’s politicians.
I would very much enjoy NOT running errands to the grocery store, getting my oil changed, getting vehicle tags at the DMV, and all of the other time wasting things modern Americans have to do on a regular basis.
I will never miss having to protect the lives of my family members from idiotic drivers who drive like they have something to prove and no fear of death. Ever.
I would much prefer that 1 million people wouldn’t die every year needlessly from auto accidents.
I could go on, but I think most people can pick out at least a couple of the above that would be worth the price of admission. If you’re so worried about the government, take another look at the point above regarding exurban growth. Rural counties of every middle-sized city will get a population boost. This will create a natural check against urban cores that tend to lean left. If anything, the new exurban growth spurred by self driving cars will cause the country to become more conservative politically.
Thanks RF
Those who say there is no need are of the same mentality as those who said there was no need for an internet
It’s embarrassing how ridiculous their position is
+infinity RF
RF,
Don’t assume no demand for SD vehicles based on your own ambivalence.
There has been plenty.
Your world may not may not include everyone, or even a majority.
There’s plenty of consumer demand for any kind of genuine automation. Cars no different than washing machines and roosters. It’s not like you have to be some sort of communist to mind a robovacuum vacuuming your floor for you, instead of you having to do so yourself.
Salt Lake to Reno along i80 is pretty darned boring by the time you’ve done it a few hundred times.
The most concerning thing about the fad’s concentration in the most statist parts of the country, is that those are the places where; when the tech disappoints the fad followers who have sunk money and credibility into it, those same starry eyeds are most likely to get a bailout. In the form of draconian “road zoning,” negatively affecting those who can’t afford the new toys, in favor of the well connecteds who can.
After all, it is extraordinarily rare that anyone stupid enough to still, after all this, believe that government is som sort of useful institution; will be able to accomplish anything useful themselves. Yet, in order to keep the progressive illusion alive, it is imperative for the government to ensure that money, resources and influence is, to the greatest possible extent, transferred to that specific cohort of self identified halfwits dumb enough to stomp for government. It’s the same reason banksters, lawyers, realtors and the like, as well as those on asset pump welfare, are being treated so well by the redistributionists.
So, just like they are currently being zoned away to live ten to a shoebox in ghettos under freeways, “autonomous vehicles” will likely be the excuse du jour to make sure their driving is zoned away similarly. Unless they pay rent to keep their overlords’ latest “car sharing app” from appearing to be as underwhelming as it really is, of course…..
Liberal, by definition, means less respect for tradition and an openness to unproven concepts. Both liberalism and conservatism have their respective places.
Name on liberal program that actually worked.
From LBJ’s Great Society, to free lunches in public schools to increase academic achievement, to the War on Drugs, to ObamaCare.
LOL.
Gimme a break!
What stupid examples. Maybe we should go back to letting old people work until they drop. Unbelievable level of cognitive disconnect among American right wingers.
“We” should most certainly “let” old people, or any people, do whatever the heck they want to. And mind “our” own life and business. Not anyone elses.
The War on Drugs? You mean Nixon’s program?
I am not at all liberal. I strongly support SD.
Is Mish a liberal?
If you ascribe to “the other side” all things you’re against, eventually “your side” becomes a team of one.
Cities are where people live, and thus where this tech will be most useful soonest. That’s why Uber and Left started in big cities too.
All major US cities are politically liberal.
You’ll get self-driving cars soon after they’re mainstream in Portland and New York.
I fall in the camp believing that driverless cars are decades away. No-one has yet gotten a robot to do anything useful in the messy real world, only in highly controlled environments such as factories. Show me a grocery store self-checkout that doesn’t need human intervention every few minutes before you even think about about making a driverless car…
I think there will be a serious realty check soon for these companies chasing the driverless dream. The most likely result will be that “driverless” technology will be used to provide safety assistance to a human driver. This is where the sweet spot will be for a long time – potential collision detection and avoidance, predictive advice to the driver “you can safely turn left now”, etc.
UN Agenda 2030 will become UN Agenda 2039 in a few years as public acceptance fails to materialize.
“No-one has yet gotten a robot to do anything useful in the messy real world”
I haven’t vacuumed my carpets, nor mopped my kitchen tile, in three years.
Robots do that now.
The same will be true of driving, and sooner than you think, thanks to cognitive computing and the incredibly cheap high performance computing hardware of today.
A $50 throwaway smartphone today has way more computing power than the most advanced and powerful desktop computers of 10 years ago. 10 years from now, the same will be true but on an even greater scale.
You’ll feel as silly with this prediction at that time, as people who were skeptical of smartphones and e-commerce feel today.
People want this technology so it will come. This liberal city idea/arguement is like saying people did not want iPhones because they weren’t available. Once available everyone wanted one.
Exactly, 10 years ago Nokia was the leader in smartphones. Most people today would barely even recognise a “smartphone” from less than 10 years ago. If there is one thing people are astoundingly good at, its not understanding exponential growth
I neither live in or near a giant city like Chicago. So I have no idea why I would use a driverless car. I have to commute about 40 minutes each way to work. My wife will still need a car while I am gone.
When I want to go somewhere, I don’t want to wait 20 – 30 minutes for a driverless car (or Uber dude) to show up at my house to pick me up. I want to go right then. And so does my wife.
Perhaps for people in highly urban environments they see this as a big deal. It has no impact on my suburban lifestyle. So I don’t think it is that revolutionary, at least for me and the folks like me.
Now show me an EV that can go 250 miles at 80 mph with the A/C blasting in summer, recharges in an hour, and is affordable. No more worrying about gas prices and mechanic rip-offs. I’m in like Flynn.
My situation is like yours and my skepticism is too.
you unable to plan 20 minutes ahead ?
Sometimes yes, sometimes no. The sometimes no thing will create tons of resistance to the idea for many, many people. Americans value the freedom of mobility that car ownership provides.
I’ve got a 2012 car with parking sensors. Five years later the sensors are worthless. Likewise with the electric parking brake. Both of these systems would be needed to replace human control. Both of these systems are $1000+ each to fix. I’ve also got a 2008 & 2003 with none of these complex systems and have not had a single repair hit the $500 mark, much less four digits.
Given the increased complexity and cost to repair such complex systems, how much demand will there be for a $40,000 driverless car that will require thousands of dollars in repair within five years? Maybe this is fine for the affluent, but what about people who can’t or don’t want to flush $40,000 down the toilet every few years?
And if the owner chooses not to fix the car, will the car brick itself? If the self diagnostics fail to detect a failure, who will be responsible when the car is 20 years old and kills someone due to the failure? Will repairs require certification similar to an aircraft mechanic? Will people no longer be legally allowed to repair their own cars? Will the cars even be considered their own at that point?
I think the idea is that although driverless cars will super expensive, they will be affordable because they will be shared across many users. You just pay for the time you use it. Some problems with this though:
– it doesn’t really alter the value equation, because the cost of ownership for a vehicle is more related to number of miles used rather than age, and a shared vehicle has no advantage here.
– shared vehicles will actually need to do MORE miles than individually owned vehicles due to deadheadings to go and pick someone up. This will increase vehicle wear and maintenance costs, plus put extra traffic on the roads.
– in a shared vehicle you have to sit on the previous user’s mayonnaise drips / baby vomit / breadcrumbs etc
– if you ride share as well as vehicle share, you will spend longer in the vehicle (to drop off / pick up others), plus have to endure their music / morning breath / banal conversation / occasional murderous attack, etc
What percentage of all driving is people showing up for work at 8:00 and leaving at 5:00? If 80% of the driving population is doing that M-F, then you still need nearly a 1 – 1 relationship between cars and people.
Bingo.
Desktop computers were once a major purchase.
Now they’re less expensive than a telephone.
Your math doesn’t make any sense. The equipment required to make a car autonomous will be a few thousand dollars to replace, let alone fix/refurbish. So no, your $40K car doesn’t become worthless when the auto features go bad or are damaged. Also, you’re simultaneously going to have a lot of out of work auto body repair shops given that crashes will effectively end. Smart owners of these businesses will be learning about autonomous driving systems now so that they can be prepared to service them.
“The average age of light vehicles in the U.S. has risen once again as consumers continue to hold onto vehicles. IHS Markit said today that the average age of light vehicles is now 11.6 years, based on a snapshot of vehicles in operation on Jan. 1, 2016.”
As you imply, useful high capital cost do not get scrapped.
See also:
While the average age of a U.S. domestic commercial airliner is 11 years old, it is not uncommon for aircraft to still be in service at 24, 25, even 30 years old.
This is just speculation, but I’d wager that when autos are a subscription based good, the average life will increase substantially. Your lowest tier subscribers will be picked up by the proverbial jalopy and have to potentially share the car with other passengers. Your highest tier will be picked up in a new Benz and have a private ride.
Michael, your last paragraph seems like very good questions to me. Likely none of us know how they will be answered exactly. Here are some wild guesses:
Brick? … No. Not if the car has human controls. If the driver breaks, the human must drive.
Insurance at 20 years? … Same as a new car. I think it’ll be the manufacturers but would not bet on it.
Aircraft-like cert? … Impractical. For this reason and others, the driver will be pretty self-contained and its interface to the vehicle will necessarily be as fixed and simple as possible. Nice engineering challenge.
Owner repair? … Again, argues for the driver being self-contained. As is the case now, owners won’t be able to futz with the driver. (Currently, if my car had a computer controlled steering wheel, it would be fun to build a driver for it, but I’d be putting my legal neck out when my driver took the wheel on ordinary roads.)
Is “your” car your own? … Great philosophical question! Is your car your own right now, given that you’ve off-loaded liability to millions of other people through insurance? Ownership involves both control and responsibility, as balanced as is practicable. If you shed some responsibility, it won’t be long before you shed some control. Welcome to city living. 🙂
As to your experiences with expensive repairs to flaky, opaque items in newer cars: That’s how we older folks think about cars in general when we’re not running up 3 hundred thou on the odo.
Anyway, don’t assume robots won’t affect car repair. Currently, the hold-up for automated car repair is the same as for many things, including self-driving vehicles: Nobody knows how to build a robot for the job. But such knowledge is being gathered and spread at an astonishing rate as we speak. And no one knows the limits to the current knowledge-exploding areas. Only that the main push-back is lack of people working in the field. With 7 billion prospects, 2 billion of whom are still mostly wasting their time, this is a pretty pleasant limitation to work with.
https://www.wired.com/2015/04/dmca-ownership-john-deere/
Driverless will eventually get here but …
“But don’t expect to toss out your driver’s license in 2021. Five years isn’t long enough to create vehicles good enough at driving to roam extensively without human input, say researchers working on autonomous cars. They predict that Ford and others will meet their targets by creating small fleets of vehicles limited to small, controlled areas.”
“Probably what Ford would do to meet their 2021 milestone is have something that provides low-speed taxi service limited to certain roads—and don’t expect it to come in the rain,” says Steven Shladover of the University of California, Berkeley, who has worked on automated driving for more than 20 years.
Shladover says many media outlets and members of the public are overinterpreting statements from Ford and other companies that are less specific than they appear. The dream of being able to have a car drive you wherever you want to go in the city, country, or continent remains distant, he says. “It ain’t going to be five years,” says Shladover. “The hype has gotten totally out of sync with reality.”
https://www.technologyreview.com/s/602210/prepare-to-be-underwhelmed-by-2021s-autonomous-cars/
There’s more at the link. I consider it to be a more realistic take on the subject.
Well isn’t that special? I’m typing this from Portland, Ore, the place that won’t even do driverless light rail. So we have all these toy trains running all over town with “drivers” in them on dedicated rail track, at a, what, 120K/year? public union job, while they are now authorizing driverless cars. It will be awesome when the driverless car plows into the drivered light rail.
Yes Portland …….. where the state requires an attendant to gas your car. You as a driver are not allowed to which adds to the cost of fuel.
I’m imagining a shared vehicle service where you request a ride on your phone much like you do on Uber today, and then a system dynamically re-routes a small bus to come pick you up and take you where you want to go (together with a bunch of other people on their trips). This bus is driven by a human, and he/she is constantly updated with instructions from the central planning system about where to go next to pick up or drop someone off. With a decent sized fleet this could be very efficient and user-friendly. What’s more – it could be done with today’s technology.
But of course people don’t seem to want practical solutions, preferring pie-in-the-sky autonomous, shared fleet, techno-fantasies.
RJS, mustn’t the bus be big enough to support the driver’s cost? Or why small buses coming often are more normal in areas of the world with cheap drivers available for work. And why gigantic buses coming twice a day are the American way.
Also, the passengers must be cheap to be routed circuitously.
All of a sudden travel times are randomly doubled and you have no control over when or why. No thanks.
They should start by making the light rail and the street cars autonomous. Seems like that would be easy.
Autonomous vehicles are not about liberal or conservative. They are not about any government or the UN. They are not about some plan to compromise your privacy. I am amazed at the ridiculous statements that are made when this topic is discussed.
They are about human ingenuity developing a technology that will be beneficial to those who adopt it. No one says that you will have to adopt it. I still know people who have never touched a personal computer and they have been around for over 40 years. You don’t have to use a smartphone, or even a cell phone if you don’t want to. But most people do use computers and smartphones because of the benefits they provide to the user.
That is why autonomous vehicles will be adopted. Because their benefit will be too big to pass up.
Personally, I look forward to both the personal benefits, and the benefits to society that will result from the adoption of this technology:
Millions of lives saved worldwide.
Millions of accidents and injuries avoided.
Dramatically lower insurance costs.
No more traffic jams.
Driving safely in bad weather and bad road conditions.
Reducing the number of vehicles I own from 2 to 1, and maybe zero.
Being able to have my car drive me while I focus on other things.
Less fuel consumed, less pollution, lower costs to drive.
Etc. Etc. Etc.
The only disagreement I have with Mish, is that I believe the adoption of this technology will be a little bit slower than his predictions. He is more optimistic than I am on the time line. We are both very optimistic on the benefits, and on how this technology will be incredibly disruptive to society.
The naysayers are the types who a hundred years ago would say that man will never fly. But today, everyone accepts as normal that millions of people fly every day because it is in their best interest do do so. We don’t use air travel because it’s some kind of conspiracy. We use it because human ingenuity developed the technology and it is in our benefit to use it.
Not only is technology evolving at an exponential rate, the rate of adoption is also exponential. This implies that the next wave of technology will be developed and adopted roughly twice as fast as the wave before it. Doesn’t mean that current tech will be completely replaced – people still use feature phones and dial up internet today, but the gap between early adoption and mass adoption will be surprisingly small this time around.
Bingo.
It took 20 years for the mass market personal computer to be generally adopted across society (the first mass market units appeared in 1978, and household penetration exceeded 80% by 1998).
It took 10 years for the consumer internet to become ubiquitous. The first web browsers appeared in 1993, the mass adoption of internet use exceeded 80% by 2003.
It took 5 years for smartphones to become 80% of the mobile devices used in the USA, if you count the iPhone launch in 2007 as the “first consumer smartphone.”
Uber and Lyft went national in 2013… by 2016, just 2.5 years later, it had destroyed the taxi cartels in every major city in the USA.
When self-driving cars arrive, the uptake in urban areas will be dizzyingly fast, and parking garages and other urban car infrastructure will collapse within 12 months.
Disagree to an extent. I’d say the naysayers are the types in the 1960’s who said “flying cars are not just around the corner”. Sometimes you have to spend a lot of time trying to create something to find out why it is almost impossible to do.
I’m sorry, Mish, it’s still pie in the sky. So Portland wants to be in on the cutting edge of law suits. Good for them. Yes, driverless trucks will ply the Interstates but in a very limited way. If you have never driven truck then you have never encountered all the exceptions that automated systems cannot handle. Road conditions and weather are two of them. Other vehicles on the road driven by drivers of varying aptitude and awareness are two more. The list is very long. And when these problems occur then the automated rigs will be shut down rather than risk accident and injury. And I am just talking the Interstate, not the US or state highways. Most of the automated truck traffic will be shut down during significant portions of the winter.
As for Portland and other cities, those metropolitan areas that are small and compact will see regular vehicles prohibited from the downtown and a few residential areas in order to reduce the chance of accident. And unless a city is not located in a snow zone then we will see periods of automated vehicle shutdowns. City snow has a bad habit of making the commute hazardous, what with snow plows, very narrow lanes, ice formation, etc. So Mish, it’s time to get off the bandwagon for automated vehicles, you may never see it fully implemented in your lifetime. Oh, you’ll see pieces of it and a lot of failures and a lot of wrongful death suits, but it will never be the sunny day picture you keep projecting, so stop trying to blow sunshine up our skirts.
40,000 Americans killed last year by human drivers
But the liability belongs to the car manufacturer when it’s an autopilot, and there’s a lot of money in suing huge corporations vs. individuals.
I suppose then that if we let computers drive our vehicles then the vehicular death rate might only be 4,000 a year, an acceptable level, right? People still die every year in elevator accidents and most of them are controlled by computers. Okay, so we reduce the death rate. If that were the only measure then you would have a plausible argument. But there are many more factors to consider. Do we command that every one use driverless vehicles and no one may operate any motor vehicle? What about bicycles? What about pedestrians? Skateboards and roller skates, aren’t these problems of human actions that are not always based of the safe operation of such vehicles or bipedal movement? What’s your criteria?
I agree. Plus how many of those 40,000 fatalities were caused by alcohol-related factors? We are overestimating peoples’ driving (and decision-making) abilities while underestimating the ability of compute systems evolve and react.
WB2014, you seem to bring significant experience with the road to the table, but do you know any engineers who you trust to give you an overview of their job and what they know?
Here’s a flavor:
You know there are jillions of driving situations a driver must cope with. How many? Have you counted them in any way? You’d want to do so if you were given the job of making a driver. Going in to any engineering job, you want to find the scope of the job, right?
So, how do you count these situations? Well, an engineer might simply list them like this on the back of his envelope:
Humidity at 4 main values, including “rain” and “desert”.
Sun in 8 main directions.
Sun at 4 main azimuths, including “night”.
Temperature at 10 main values including several around freezing.
Traffic at 4 main densities.
Live animal in road ahead 5 main ways at 4 main sizes.
We’re already at 4*8*4*10*4*5*4 = 1,020,000 situations, increasing exponentially and being silly-conservative in “main” thing counts. And the list goes on and on.
But the list ends. Sure, to complete the list may take a while. Sure, you may miss some very, very rare list items. No biggy. You’ll easily get everything any large group of old, human drivers will ever have experienced. If it’s critical to do so, you can even mostly automate the job of making the list! Put sensors in, say, 1000 vehicles and let ’em feed back the road from around the world for a while. (Ever wonder what data Google has from street view cars having driven most of the world’s roads? Already. Today.)
When the list is complete, the situation count will be a number. A big number, but just a number.
Now, the thinking and calculation above is something the people who build self-driving cars all know as well as you know which sleeve to put your arm through.
So with, say, a trillion trillion situations, what are these people thinking? Many, if not most, are very good at detecting hopeless jobs. None of them want to waste their time on a hopeless job. None of them would have any trouble finding another job. Why have these people not given up?
Well, here’s something to consider: Winter driving.
How good are humans in the snow?
Lousy, right? There are good, experienced humans, but you can’t get much experience in a short, human lifetime.
But a single computer? With 1000s of trucks all feeding experience back to it from a world where it’s always winter someplace? With the ability to “drive” billions of miles under any simulated conditions, including millions of types of snow?
Yeah. Who’s going to be the better driver?
Your assumption is that somehow human experience can be quantifiable. The second assumption is that the computer will be able to recognize the pattern as fast if not faster than a human and act accordingly. What we are talking about is really AI in a nutshell. Then the third assumption is that the various sensors will be as good and the human senses of an experienced driver. Now there are other assumptions that many individuals/engineers may make but the point is, are the assumptions really valid? If not in whole, that at least in part. Do you start to see the problem? I have observed through my long life that Murphy’s law has not been repealed. I’ve got a half million miles of driving truck under a variety of conditions. Not every and all conditions, but most. the one “iron Law” of life that includes driving vehicles is “Shit Happens”. It literally does. The truck in front of you kicks up a large rock and throws it through your windshield striking you in the face. True, it is a very rare occurrence but it does happen. I’ve lost headlights that very way twice. Other drivers, never happened. What happens when you blow a steering (front) tire? Nails and glass and pieces of old steel belted radials cause blow outs. It’s hard enough during the day to avoid them but at night it becomes worse. Your scanners may not pick up the pieces of tires but a driver usually does. Now we have an evasion process that means you have only a couple of seconds to react. If you swerve you better check to make sure no one is beside you. High wind gusts can knock over a van, particularly one that is empty. You have a trailer that whose side area is approximately 10 by 53 feet. Empty, that van weighs in about 6000 pounds, loaded close to 60,000 pounds. You see the difference? Some parts of the country are subject to tremendous downpours of rain that come abruptly and may last as long as an hour. the human eye can’t see through these downpours. What are the chances a lens will? Same thing happens in snow country when a faster truck passes you while spraying flying snow causing a “whiteout” for several minutes. Your engineers will never solve these types of problems, never. But the last problem is the most deadly. When your computer data base is wrong and sends your 80,000 pound semi over a bridge that is rated for only half that weight. A driver will see the sign and stop. an automated truck will continue on. Want to know how many of these little “mistakes” there are? At least a thousand. God knows that I’ve been trapped that way and had to back a 75 foot articulated vehicle seven miles back to the Interstate and onto the off ramp. These little mistakes are very common. The best that interstate automated trucking can do is to ferry loads from one terminal on the interstate to another. it will be strictly van and reefer loads. Ain’t no way in hell any state will let an automated truck haul a fourteen foot combine down the interstate. No way, no how. Usually a driver needs about two years experience behind the wheel before they let him haul one of those loads. Oversize loads and heavy haul pay more, a lot more. That’s why that CDL means you are a professional driver instead of a four wheel seat cover.
So to sum up the argument. Driverless vehicles will work with a reasonable certainty when the conditions are fairly uniform. They will start to have problems when conditions change unfavorably, even slightly. By the way, go read some literature on how they drive the Mars surface rover and many of the problems they face. Makes interesting reading. Maybe your engineers might want to do the same.
I agree with your three assumptions. They involve some work to be true, depending on the situation and whether in a given situation they are already true or not. No. I have no problem with them.
The first seems true on the face of it.
The second is old news. (Example: Think of the joke rock-paper-scissors machine a while back. It cheated and won all the time by being fast.)
The third: Last I checked humans don’t see in infrared very well. The human brain does a really great job of processing what the “eye” sees, but if you want to really “see” something you use that human brain to process a picture from machine sensors. Say, do you wonder whether machine sensors can make rain invisible? Sensors, themselves, that is. I can imagine a fairly easy way to process rain out of video. Drive without wipers some time and notice how you can see better if you constantly shift your head back and forth. But that’s not the trick I’m thinking of.
To argue when self-driving vehicles will be around and in what circumstances, don’t bother listing issues you see with self-driving vehicles. You may be a professional driver, but you’re not a professional issue lister.
That tiny list of categories I made would, if filled out, certainly include things like wind speed and direction, vehicle weight, breakdown of parts, recognizing and reading signs, road condition (as, for instance, what’s the condition of this bridge?), state of other vehicles, local customs and laws, and a generalization of the silly “live animal” category I listed (which includes tires and other objects, of course). Those categories would be listed in the first few minutes of informally scoping the project. There are teams who have been doing this process formally for 10 years now. They know those categories and many more.
The weather condition items you mention (e.g. whiteouts) come as a consequence of more general categories already listed (temperature, humidity, etc.).
I didn’t even mention time. All parameters change over time, short and long.
So, again, we agree there are a lot of things a self-driving vehicle must handle.
Let’s take an example of a particular item you mention:
“A driver will see the sign and stop. an automated truck will continue on.”
If the automated truck were as dumb as a fence post.
But if such a truck were reasonable, we’d have had self-driving trucks long ago. I would expect a self-driving truck to:
1) Read the sign. Quicker, from further away, in less favorable circumstances, and more accurately than a human.
2) Maybe even recognize the carrying capacity of bridges by sight, if possible. Certainly more accurately than any human driver.
I’d assume some groups have already built #1. It’s so similar, if simpler and easier, than other things already done at super human levels.
#2 is probably a big, straightforward data processing job, nowadays. Low priority? Good as a PR stunt?
As a electrical engineer with 22 years experience, I’ve seen many seemingly insoluble problems become trivial.
e.g.
Home computers will never be useful or simple enough for the general population.
There will never be cell coverage in rural areas.
Full size cars will never exceed 30 mpg.
If you are an engineer and you believe a software problem is too complicated, you likely retired before 2000. I would not hire an engineer afraid of this problem.
You are getting complex again Kev. You and Steve are working on some theory, aren’t you?
If DWI accidents decline, then so will the number of DWI and personal injury lawyers.
I have a question. As a longtime resident of Oregon I want to know what happens to the sensitivity of all of those lenses on the cars roof when it rains, snows or sleets?
Over half of Americans according to AAA don’t feel comfortable with driverless cars on the road. I think adoption rates may lag. Count me as a non-adopter.
http://www.businessinsider.com/aaa-survey-americans-fear-driverless-cars-2017-3
Also, I drove home from Newark airport last week, about 100 miles from home. The amount of potholes I avoided on the roads/highways, especially in NY/NJ was shocking, and in April! Will driverless cars see potholes and avoid them?
so what?
What about the huge percentage of people who struggle with night vision, little vision?
How many of those Neanderthals are computer illiterate?
“Will driverless cars see potholes and avoid them?”
Why not? Sure seems like a plan.
Har, har. I lived for a year in that area 40 years ago and could not believe the jeep trails these guys called roads. Still the same, eh? Are the cops still inept drivers?
It doesn’t matter if you’re a non-adopter – you will be adopted involuntarily sooner than you think.
First, early adopters get to pay less auto insurance. Some estimate premiums slashed by 90%. Build that savings into the extra cost of the new gadgets and the initial purchase starts making more sense. The auto insurance carriers have fewer and fewer customers paying the human driver rate, meaning the only people having accidents and paying for full insurance are human drivers. Rates will increase on an S curve.
Maybe you’re wealthy enough to withstand that premium. Many will not be able to. I hope I’m able to because I would enjoy the ability to drive once in a while.
I don’t get this fascination with driving. I remember the euphoria of driving unsupervised for the first time when I got my first DL at age 17. Aside from that fleeting moment of joy, driving has been over 30 years of exhausting drudgery. I can’t wait for the extra relaxation of being driven everywhere — why do wealthy people have chauffeurs? If I want the sensation of self-directed movement I will buy a nice bicycle.
Couldn’t agree more. Not only that, but many of us were dangerous behind the wheel in our teen years due to impulsive hormonal stuff that is par for the course at that age.
Side point on biking: In high school I went on multi-week cycling trips with a group of other kids – it was a blast. Whenever I returned home and got behind the wheel of a car I was bored out of my mind.
The autonomous systems will improve (learn) through driving millions of miles. It will take some time but eventually they will reduce the number of auto accidents by close to 99%. It is impossible to reach 100%.This alone will save millions of lives and countless injuries. Some accidents, though, are simply unavoidable. Of course, there will be some high profile lawsuits along the way, but those will be insignificant compared to what we have today.
One area that is still under development and rarely mentioned, is the programming of ethics into the systems. For example, when an accident is unavoidable, but the vehicle has a choice about what action to take; hit the animal, hit the cyclist, run into the crowd on the sidewalk, or drive the car into that wall and injure or kill the driver. Which should the vehicle choose? I can tell you that animals do not fare well, and drivers are not always at the top of the list to protect.
I suspect this will give the naysayers fodder for discussion.
Regardless, these systems are already on the road, and they are getting better and better. Eventually they will be commonplace. And society will reap the benefits.
This feels about right:
Animals
Drivers
Cyclists
Adult pedestrians
Young pedestrians
Some might say such a list is ridiculous. I’d say it is ridiculous for people to think that tired/stressed/medicated/drunk/agitated/distracted humans are capable of making the right ethical split second decision.
It seems to me that it would be technically and economically more favorable to upgrade and automate long haul rail transport than long haul trucking. Even now a 100 car train with a 2 person crew can move an amount of cargo that would take about 400 long haul trucks on a weight basis using much less fuel. Rail can also be easily electrified.
Rails are too static, and goods end up having to be moved by truck anyway. I’d argue it would be more efficient to convert some rail into self-driving semi-truck only highways.
Remember when cell phones were acquired because saving one trip a month or handling one emergency kinda paid for the phone? Or at least gave an excuse for getting a phone.
One less fender bender pays for self driving equipment by the time such equipment is available in stores near you, as seen on TV.
One year of auto insurance.
Five years from now I will enjoy reading all this nay-saying poppycock while being chauffered in the back seat of my self-driving car. I’ll wave to y’all as I pass by.
It’s going to be glorious.
It’s decades away, forget about it.
https://www.recode.net/2017/3/16/14938116/uber-travis-kalanick-self-driving-internal-metrics-slow-progress
Approval is predicated on very low accident rates. Once the approval is there, it will become old-fashioned or illegal to drive manually in very short order, because of the preventable accidents that such choice brings.
Interesting data, but there is another data point that would be informative – the number of trips completed without safety driver intervention vs total trips. The trend of this ratio would be a good indicator of when self autonomy will become a reality.
Considering the fact that safety driver had to intervene every 0.8 miles, that number is close to zero I guess.