Autonomous Cars

Editor, Brutalhorse.com
May 11, 2014

Like any group of friends, we here at Brutal Horse aren’t strangers to the occasional heated debate. But unlike most friends, we’ve decided to put our arguments to the test through the Brutal Horse Debate Club. The rules are simple: Each side presents his or her argument in a 1,000 word essay. After that, each side reads the other’s essay and pens a 500 word rebuttal. Fairly standard stuff, but we are gong to take it a step further and put up real cash money to settle our dispute. We’re taking our debate to Amazon’s Mechanical Turk service (http://www.mturk.com/, which allows people to pay a set of “Human Intelligence Workers” money to perform some mundane task) and paying four people $5 to choose a winner and write a paragraph in support of their decision. We’ve decided to allow ties, because sometimes that’s the way life is. Oh and, loser pays of course.

The question

Will autonomous cars be ubiquitous in 15-20 years?

This question was distilled down from the original argument Ray would make, that if any of us has kids (none of us do at this point), will they actually drive a car? This was a little vague for the general public, so we came up with the question above.

Ray’s argument

Currently, technology and car companies are racing to produce a self-driving, fully autonomous car (http://en.wikipedia.org/wiki/Autonomous_car) that can perform all functions of a human driver automatically. The self-driving car has a long history, essentially dating back to the 1920ís, though they entered public consciousness with the 2004 DARPA Grand Challenge, which pitted self-driving cars against a tough desert track. There are many reasons companies are pushing autonomous vehicles to the public, safety and convenience being at the top. In fact, in 2012, traffic accidents accounted for 34,080 deaths in the United States alone. I believe that autonomous cars will be ubiquitous within the next 15 to 20 years (between 2029 and 2034). Even though the necessary technology seems futuristic, driving a car causes injuries and death, wastes time, and increases stress, and so the auto industry and government have a huge incentive to bring them to the public. I will break down my argument into two main points: First that self-driving cars are inevitable and second that the timeframe of 15-20 years is correct.

Cars are becoming more autonomous with every new model year. The public can currently buy cars that can park themselves (http://en.wikipedia.org/wiki/Self-parking_system), adapt cruise control speed to traffic speed (http://en.wikipedia.org/wiki/Autonomous_cruise_control_system), and warn a driver when they veer out of their lane (http://en.wikipedia.org/wiki/Lane_departure_warning_system). Other necessary components of an autonomous car have been familiar to drivers for years: GPS, anti-lock brakes, electronic stability control, and electronic traction control. Further easing the transition, mechanical linkages of driver controls to the car (steering, throttle, brakes, etc.) are being converted to electrical systems (http://en.wikipedia.org/wiki/Drive_by_wire). These are all examples of subsystems that will eventually be combined into a fully autonomous car. Other technologies needed by an autonomous vehicle but not currently in modern cars (radar, lidar, cameras, computers, etc.) are well understood and widely available. In fact, the auto industry and technology companies (Google, for example) are working to produce autonomous cars by the year 2020 (http://en.wikipedia.org/wiki/Autonomous_car#Official_predictions).

A roadblock to autonomous vehicles becoming widely accepted is liability/insurance. It is common to hear an opponent of autonomous cars claim that they will never be accepted because if they are ever involved in an accident then it is impossible to assign blame for liability or insurance purposes. While this is understandable, I believe that it is a non-issue. First, many states in the US use a ďno-faultĒ auto insurance system, in which drivers involved in an accident are only liable for damage to themselves or their own car (http://en.wikipedia.org/wiki/No-fault_insurance). (Typically, if a driver is proven to be negligent, they are liable for paying the expenses of the other party involved in the accident.) Insurance for autonomous cars can exist under a no-fault system, as operators of self-driving cars involved in an accident would use their insurance to pay for damage to their car only. A related issue involves accidents due to failure of the self-driving car. In these cases a car company may have to offer warranties, recalls or, alternatively, car insurance may pay for it. Car companies may in fact be liable for any accidents or injuries that result from the faulty operation of their cars, as has already been the case with the Toyota recall from 2011 of cars with faulty accelerators. A company is still liable for the product it sells, whether itís autonomous or not, though it remains to be seen whether these cases will be dealt with through insurance or the car companies directly.

If we assume that an autonomous car will be produced by 2020 (according to predictions by the auto industry), we are left with about 10 years for these cars to become ubiquitous. For a product to become ubiquitous there must be demand, which is made up of desire for the product and ability to afford it. I believe that the public will want autonomous cars. As a personal example, I drive 90 miles roundtrip to work everyday, which is essentially two hours of wasted time. Public transportation options are nonexistent for me, so a self-driving car would be entirely welcome. The lost time in my commute would not be so bad if I was able to read a book, watch a movie, or even get some work done. Fortunately, I donít encounter much traffic, but many believe that autonomous vehicles will reduce traffic by reducing accidents and poor driving habits (http://www.vtpi.org/avip.pdf). Most people that experience stressful, congested commutes will surely welcome the reduced traffic that comes with a society using self-driving cars. Finally, with reduced accidents comes reduced injuries and death. Traffic injuries should not be accepted as normal when we can use technology to alleviate this problem.

Of course, the technology needed to produce an autonomous car will add additional costs to a car. Many components are necessary: radar, lidar, cameras, computers, and possibly networking equipment. The first autonomous cars will surely be expensive and only available to high-income individuals; however, if the first self-driving cars are produced in 2020 as predicted, there is an additional 9-14 years for economies of scale to reduce their price. As I believe that autonomous cars are simply cars with additional features, the history of car features can be used as a guide to predict the future of self-driving technology. As a first example, Ford first placed airbags in cars in the early 70ís, and about 20 years later in 1994 Ford made them standard equipment (http://en.wikipedia.org/wiki/Airbag). Next, electronic stability control, which is used by cars to maintain traction in slippery conditions, was first developed in 1987. Twenty years later, in 2007, GM made electronic stability control standard equipment in its SUVs (http://en.wikipedia.org/wiki/Electronic_stability_control#History). Finally, while GPS has a fairly long history, starting with initial testing in 1972, Magellan Navigation Inc. produced the first hand-held navigation system in 1989, about 24 years ago. Today, GPS is in every smart phone and used in applications as mundane as tracking your walking route.

Autonomous cars will improve society by reducing traffic, death, and injuries, lead to greater convenience, and reduce stress. The auto industry is predicting the first autonomous cars will be commercially available in 2020, leaving over ten years for them to become ubiquitous. As history shows, many technologies take around 20 years to become universally available, and as autonomous car research is in full swing, I foresee our roads as nearly driver-free in the year 2034.

Neal’s argument

I was supposed to have penned this argument months ago. And I did. But the dog ate it. Kinda. I had laid out what I thought to be a pretty sophisticated and air-tight case for why self driving cars will not be ubiquitous in 15-20 years. I made economic, political, legal, and social points. I never questioned the technical feasibility of this endeavor but rather said that society as a whole was not ready for this. I had about 900 words on paper and then I stupidly overwrote the file with an older version that had about 100 words. So I gave up, and more or less conceded victory to Ray. But then this morning happened.

I was doing my usual Tuesday commute up to Philly, in the rain, on I95. In other words, it was slow going. As I looked around me I tried to play a guessing game of the age of cars, brought on by a slow moving jalopy in the left lane. We had a 1999 CR-V, a 2005 RSX (mine), a 2002 Honda Accord, 2005 Prius, 2013 Ford Taurus, 1986 Olds Cutlass, and so on. These are more than shots in the dark at the years Ė I fancy myself somewhat of a car enthusiast Ė therefore these years are probably correct plus or minus one.

My point? The average model year of cars on the road is probably early to mid 2000s. In fact, recent data suggest the average age of cars on the road is 11.4 years old and five years from now will continue to be at least 10 years old (1-6). So what does this mean for this debate? Letís assume that self-driving cars will be widely available for sale around 2025, a not unreasonable estimate by experts (7,8). If the public continues to hold on to cars for an average of 10 years Ė a conservative estimate Ė this means that self-driving cars will not EVEN HAVE A CHANCE of being widely used on the roads until 2035. This fundamental assumption implies an immediate and 100% conversion of the public to buying ONLY self driving cars. Of course, we all know this WILL NOT OCCUR. Therefore, the idea that self-driving cars will be ubiquitous and make up the majority of the cars on the road 15 years from now (or letís round up and say say 2030) is ridiculous.

In fact, even if I concede that the myriad obstacles in the way to self-driving cars being widely available (vis-a-vis the economic, political, legal, social) are resolved (ha!), just from the rate of public consumption of cars, ubiquity by 2030 is an IMPOSSIBILITY

Ray, care to comment?

Neal’s rebuttal to Ray’s argument

As expected, Ray laid out a thoughtful insightful argument, which I concede. Wait, what did I just say? Yes, Ray I concede that this is not a technological issue! In fact, in my opening remarks I even make the point that this is not a technological hurdle. This was never a technology debate. The technology exists to a large extent today, and in 15-20 years self-driving cars will be readily available for sale. The point I wish to return to is that this will be a SOCIETAL decision for the ubiquity of autonomous cars.

Take, for example, the mid 90s push for Segways. You remember the advertisements, right? ďITĒ is coming! It was likened to the most significant device since the computer (9). And in fact would revolutionize transportation. People will no longer walk anywhere; everyone will use the Segway! Roughly two decades removed from its introduction, the devices are hardly ubiquitous. In fact, the only people who have seemed to have adopted Segways are the pigs that work at the airport and college campuses. So what happened? Society deemed this is unnecessary, a luxury rather than a necessity. I believe the same is true for self driving cars. These are a luxury for the few well-to-do elite that can afford them (kind of like hybrid cars at their introduction).

Now to return to a point that Ray makes in his argument, the public will demand this, I have to respectfully disagree. While it is true that his commute his long (2 hours round trip) and he may benefit from this, the average one way commute is well under half of an hour (10). And in fact, Rayís sample of n=1 for this opinion is hardly a sample size sufficient of demonstrating anything except that he agrees with himself.

So I know that I framed my entire opening tirade in terms of a practical argument (is there even enough time for the auto industry to turn itself over in 15-20 years), and I would like to conclude by saying that regardless of whether these numbers are correct or not is NOT THE POINT. The point is simply this, society will ultimately determine if and when these cars become ubiquitous in the time frame we agreed upong. Based on history, anytime there is a new technology, there are early adopters, the greater public follows (i.e. the sheep), and then the luddites remain steadfastly behind (including yours truly). So while a hundred years from now we may no longer need a driverís licesnse, 15 years from now we will merely have prevalence of autonomous cars among the early adopters. And that is not ubiquity.

As one final example, Iíll leave you with the mobile phone. No one would argue this device is NOT ubiquitous. So letís trace how long it took for this to occur and compare that to autonomous cars. Mobile phones largely became available for sale in the mid 1980s (11). Letís say they were ubiquitous by mid 2000s among the US population, which implies a 20 year interval. With self-driving cars first becoming widely available for sale in 2025 (a point I made earlier) this still means 2045 for ubiquity assuming the same level of interest in self-driving cars as mobile phones. A very lofty assumption indeed. Does this date fall within the original argued point (between 2029 and 2034)? Iíll let you be the judge.

  1. http://money.cnn.com/2013/08/06/autos/age-of-cars/index.html
  2. http://www.edmunds.com/car-news/average-age-of-cars-in-us-jumps-to-record-high-of-114-years.html
  3. http://blogs.wsj.com/corporate-intelligence/2013/12/26/cars-on-american-roads-are-older-than-ever/
  4. http://blogs.cars.com/kickingtires/2013/08/study-cars-on-the-road-continue-to-age.html
  5. http://wot.motortrend.com/average-vehicle-age-in-u-s-hits-11-years-ford-models-most-abundant-on-roads-224491.html
  6. http://abcnews.go.com/Business/average-car-american-roads-time-record-10-years/story?id=15406801
  7. http://viewer.zmags.com/publication/f5bffeaf#/f5bffeaf/19
  8. http://autos.aol.com/article/self-driving-cars-coming-soon-dealers-near-you/
  9. http://content.time.com/time/business/article/0,8599,186660-1,00.html
  10. http://project.wnyc.org/commute-times-us/embed.html#5.00/42.000/-89.500
  11. http://en.wikipedia.org/wiki/Mobile_phone#History

Ray’s rebuttal to Neal’s argument:

While Neal makes an astute point, he is pulling a Malcolm Gladwell worthy hack-job by hastily attributing predictive power to a fun anecdote. Nealís statement is correct, the current average age of cars in America is 11.4 years; however, average age is a bad measure in this situation, as it ignores the actual car age distribution. In other words, Nealís focus on a single statistical measure fails to capture a more complicated reality. (Further, the reason the current average age is remarkable is that it is at an all-time high due to the recession and is probably anomalous, so weíll use numbers from before the recession in what follows.)

First, leaving aside the average age of cars, letís look at the actual breakdown by model year. In 2001, 38.3% of cars were older than 10 years (not 50%), 22.3% were between 7 and 10 years, 25.8% were between 3 and 6 years, and 13.5% were less than 2 years1. (The reason that only 38.3% of cars are older than 10 years but the average is around 10 years is that itís a skewed distribution. Many of the cars over 10 years are probably much older than 10 years, thus skewing the average age.) If we assume that autonomous car sales start in 2020 (not 2025 as Neal states), then by 2030 if all new cars sold were autonomous weíd have 61.7% of cars being autonomous. Of course this is being generous, but it defeats Nealís assertion that it is impossible.

More realistically, letís create a model of autonomous car sales growth. Weíll start at zero autonomous cars on the road in 2020, when it is estimated that they will first go on sale. Of course, the initial adoption rate will be low—-letís assume a mere 1% of new car sales—-but will increase steadily if the technology works as advertised. Say the sales rate increases modestly by 5% per year. Currently there are about 250 million cars on the road, and before the economic recession 20 million new cars were sold and leased per year (http://en.wikipedia.org/wiki/Passenger_vehicles_in_the_United_States#Total_number_of_vehicles), giving an 8% replacement rate per year. Adding all this up, and assuming that a negligible amount of the autonomous cars sold are replaced, we have by year:

2020: 1% autonomous cars of 8% new car sales = 0.08% of 250 million or 200,000 new autonomous cars
2021: 6% autonomous of 8% new cars = 1.2 million new autonomous cars (1.4 million total)
2022: 11% autonomous of 8% new cars = 2.2 million new autonomous cars (3.6 million total)
Ö
2029: 46% autonomous of 8% new cars = 9.2 million new autonomous cars (47 million total)

Adding up all of the above gives 47 million autonomous cars on the road before 2030, or about 20% of cars on the road. Per the rules, we have another 5 years, which, due to increasing sales rates, leads to 123.2 million autonomous cars by 2035. 123.2 million autonomous cars in 2035 are nearly half of the 250 million cars on the road now. While this is cutting it close, I believe that my estimates are modest and confirm that it is very possible to achieve autonomous car ubiquity by 2035.

The results

And the winner is . . .

It’s a tie!

I know, somewhat anticlimactic, but I’m ok with this result. I think both sides were well-argued and the comments with the votes confirmed that. Here’s what our HIW’s thought:

First pro-Neal response

Summary of Ray’s argument: Ray’s stance is that by 2034 all cars on the road will be autonomous. The advantages that this type of vehicle offers speak for themselves, so it is just a matter of getting the price point and availability in line. He believes that even now, all the components to create such a vehicle are already in existence. The pieces simply need to be put together, tested, decreased in price over time and on to the next revelation. His argument is based largely on logical equations based on history.

Summary of Neal’s argument: Just because the car will be readily available for the right price, that doesn’t come close to meaning that it will become a priority for everyone. In fact, Neal believes that most will see it as just another luxury that is not feasible for the average American to own. He also uses logic and statistics to drive home his points. He states that there is no way that this vehicle will be the only option for future purchases.

Decision: I do not think that people will be able to afford the vehicle, the insurance or prevent them from being stolen. I don’t think that the economy is going to improve nearly enough to shift priorities to a self driving car. I think that perhaps a form of this vehicle as public transportation would be realistic in Ray’s time frame. Additionally, people are not just going to give up their current cars for the half hearted promises of being able to possess an autonomous vehicle. Not to mention, I think that there are ALOT of people that love driving and have no interest in losing control of the steering wheel.

First Pro-Ray response

Summary of Ray’s argument: Autonomous cars will become ubiquitous within 20 years, both because the technology necessary to make such cars feasible already exists, and because the demand, both from consumers in particular and society in general, will drive production by that point. The reason why this demand will exist comes from a range of factors, such as the increased safety of cars that don’t rely on faulty human judgement and reaction time, the ability to devote the time otherwise spent operating the vehicle to other activities, and the sheer novelty of such a futuristic construct. The estimate of 20 years reflects the average time it takes for new technology to mature and be widely-adopted in a general sense.

Summary of Neal’s argument: Ray’s forecasting makes a very optimistic assumption, being that society will both appreciate and buy into the benefits of a driverless car in such a short time frame. The cars we have used for a hundred years will still be sufficient, meaning that the driverless car will become much like the timeworn idea of the flying car: a technical possibility (working prototypes have existed for quite some time), but ultimately, an unnecessary frivolity. Driverless cars may become ubiquitous, but it will take far longer than 20 years for this to happen.

Decision: Safety features are big sellers on many products, not the least of which are automobiles. The idea of a driverless car, one which relies on the hair-trigger reaction time and decision-making of a computer, will certainly be the most effective new safety feature on an automobile since the invention of the internal combustion engine. Furthermore, Ray correctly points out that, after a certain level of adoption, autonomy will become a standard feature in new cars, which will necessarily make adoption skyrocket. Neal makes the counterpoint of how mobile phones took somewhat longer than 20 years to be adopted by the American population. While this is true, it ignores the fact that the pace of adoption of mobile phones was contingent on the creation, from scratch, of an entire infrastructure to support their use. Driverless cars are designed to operate on existing road infrastructure. Neal also made a comparison to Segways, but this was a case of a product made to meet a demand that never existed. Cars will always be in demand, and the driverless car is not a new and completely different sort of vehicle than the modern car. It is simply a version of the modern car with enhanced capabilities. Therefore, I believe Ray made the more convincing argument.

Second pro-Neal result

Summary of Ray’s argument: Ray gives examples of ongoing technological developments that are enabling increasing autonomous function in cars, predicts that this trend will continue, claims that auto manufacturers are actively pursuing the development of autonomous vehicles, and suggests that costs will go down quickly enough for autonomous cars to be economically feasible within the 15-20 year time frame given. He then gives a personal anecdote about his driving habits and also lists a variety of reasons, including reducing accidents, saving time, and cutting down on driver stress, that autonomous cars will be beneficial and drivers will choose to adopt them.

Summary of Neal’s argument: Neal claims that the average age of the cars on the road is over 10 years, and assumes that this tendency of drivers to hang on to older cars for will continue. Based on this, he concludes that even if the technology for autonomous cars is available in 10 years, it could take 30 years for them to become prevalent, if that happens at all.

Decision: Neal’s argument is simple but difficult to challenge, and his example of cell phone adoption supports it strongly. Cell phones were available in the early 80’s (I remember this) but it took about 20 years for the majority of people to use them (I remember this too). I’m skeptical whether the demand for autonomous cars is the same as cell phones, but that starts to get into a lot of opinion and conjecture, which are the things Ray’s argument revolves around: that the technology for autonomous cars is being aggressively developed by the industry, that it will be implemented within the given time frame, that autonomous cars will produce benefits to driver safety, that drivers actually want autonomous cars, and the like. Basically, Neal has at least one fact (or solid trend) on his side, which is parallel to what I’ve observed in adoption other technologies, and which even Ray at least partially concedes exists, whereas everything that Ray claims as support for his argument is subject to debate.

Second pro-Ray result

Summary of Ray’s argument: We will be technologically ready to produce autonomous cars by 2020. Liability and insurance problems can be easily solved. Even though the first ones will be expensive, they will become cheaper over time, just like other new technology. Given what will be the desire of people to avoid stress, accidents and other problems with driving, people are like to fully adopt the car by 2035.

Summary of Neal’s argument: Society will determine how long it takes before the autonomous cars will be ubiquitous. Given that automobiles are on the road for 10-11 years, it would not be likely, for the majority of them to be autonomous in 15 to 20 years, if the first ones are not produced until the year 2020. It is more likely the cars would be ubiquitous in 2045. Also, look at how long it takes the majority of people to adopt new technology. The mobile phone was available in the 1980’s but did not become ubiquitous until the mid 2000’s.

Decision: 1) My gut says that Neal is correct but his arguments were lackadaisical, anecdotal, and not supported by real evidence. Therefore, I have to support Ray. 2) The argument between the two men was not if the cars would be ready by 2020. There was no argument on liability and insurance problem, and no argument about the cars becoming cheaper over time. So Ray wins those arguments. 3) We are now down to whether the cars would be ubiquitous by 2035 or 2045. Although Neal made a good point about the phaseout of cars over 10-11 years, based on today’s driving habits, Ray demolished the argument with his chart showing the replacement rate from 2020 to 2030. Then he concluded we could get 123.2 million autonomous cars by 2035, about half. That is a good definition of ubiquitous so Ray wins the argument.

editor@brutalhorse.com
ink splash

Jacques Dangereux, app by WildTaters

Check out The Ringer by Camp Dracula,
available now.

The Ringer, album by Camp Dracula