There is no end to the list of questions—some of immediate concern, others exclusively the province of science fiction—we humans would like answered about the always-just-around-the-corner-but-never-quite-here technology of artificial intelligence. Will robots replace the working class? Will we wake up one day and find ourselves prisoners of The Matrix? Will realistic robots replace human relationships? Will AI supersede humans, rendering us obsolete altogether?
By comparison to these big, world-changing questions, the question I want to pose here deals with a more imminent and tangible matter: What, if any, paradigm shifts can we expect should we succeed in building genuine self-driving cars? The current discussion has addressed an array of legitimate concerns that can be summarized as a cost-benefit analysis:
Benefits: improved safety, reduced traffic congestion, better public transportation, improved fuel efficiency, car sharing, higher speed limits, elimination of drunk/distracted driving, reallocation of police resources, less parking infrastructure, increased freedom for the elderly, young, and disabled.
Costs: high price of vehicles and infrastructure, complexity of equipment, possible need to ban human drivers altogether (implying a wrenching transition period), security concerns (hacking), loss of transportation jobs (taxis, truckers), legal/ethical/regulatory issues, unreliability of vehicles, loss of driving skills, fear of relinquishing control, insufficient GPS/map accuracy/completeness, potential low acceptance rate by driving enthusiasts.
The question I want to ask will become relevant only after this debate has been largely resolved and self-driving cars have been accepted socially and legally. My focus here is on the decisions people will make in the marketplace about how to use this life-changing technology.
Point One: Autonomous cars will only be accepted by the public when they are reliable enough to completely relieve passengers of ALL driving responsibilities.
Technically, this point will be part of the debate and will arise during the acceptance phase. But is it critical to understand how it will be resolved in order to see what comes next. It is currently assumed, mostly by lawyers, that the autonomous driving function of a car—its autopilot feature—will be treated as merely a supplement or aid to the overall driving operation, and that the human driver will have to remain engaged to some as-yet undetermined degree, and prepared to take control should the autopilot run into problems. While this may seem like a prudent measure on the surface, it is a preposterous assumption, motivated entirely by liability considerations, and utterly oblivious to human neuropsychology. Consider the following future scenario.
It is three o’clock in the morning, raining, and you have been behind the wheel for six straight hours, drifting in and out of consciousness, while your autopilot dutifully takes you from DC to Chicago just as it has two dozen times before. You have owned this car now for more than two years and not once have you had a problem with it. Indeed you idly wonder now and then if you even remember how to drive. Suddenly, an angry red light and loud buzzer go off on your dashboard. Vague memories of your owner’s manual flash into your brain. You have exactly two seconds to: 1) wake up, 2) discern the meaning of a light and buzzer you have never experienced before this moment, 3) figure out the nature of the emergency (ice? a deer? computer malfunction? flat tire? jackknifed semi?) out there in the rain, 4) deftly execute a heroic, lifesaving maneuver despite having not touched the steering wheel or pedals in more than two years. And you must do all of this at 75 miles per hour on wet pavement in the dark.
If this seems like an extreme, unfair example, ask yourself, under what sorts of conditions is the autopilot likely to fail? Will it be on a sunny afternoon at low speeds in light traffic with perfect visibility? Or will it be in exactly those terrifying situations that would cause even Danica Patrick’s heart to skip a beat? I predict that this scenario will have to come up in a court of law exactly once, if at all, to definitively end the legal fantasy that drivers can be held liable to compensate for sudden failures of the autopilot. In a real, flesh-and-blood human brain, the only reliable way to remain vigilant (in any particular instance) and skilled enough (in general) to avoid catastrophe in such a case is to actually drive the car oneself, rendering the autopilot feature obsolete. Therefore, the only way anyone will—or legally could—accept the autopilot is if the human driver is never expected or required to step in and save the day. Ever.
Who in his right mind would ever use the autopilot feature if there were any chance at all that the blood-curdling scenario above might come to pass? Failure to act appropriately, besides potentially killing you and who knows how many others, would leave you legally responsible for something over which you had virtually no control. The autopilot is supposed to lessen the stress of driving, make it more predictable and safer. It is not supposed to be a terrifying Russian Roulette-style abdication of personal control, akin to boarding an Aeroflot flight during a bad year.
One consequence of this, perhaps counterintuitive but nonetheless common sense, observation is that the autopilot feature will have to be almost perfect before it will pass legal/regulatory muster. Automakers will not be allowed to roll out a series of iteratively improved but still shaky autopilot versions on the assumption that drivers will pick up the slack. Psychologically, there is a sharp distinction between the features we have today (e.g., lane assist, adaptive cruise control, automatic breaking), which leave the driver fully responsible for driving, and an authentic full-featured autopilot that takes over the driving function. The courts will get an earful from neuropsychologists, learn that the human brain simply cannot react to every possible situation in a split second without ongoing vigilance and regular practice, and rule that car companies cannot offload their liability onto their customers. It will soon thereafter be recognized that the autopilot must rival the safety of air and rail travel before the public, the regulators, or the car company lawyers can accept it.
For perspective on this matter, the table below shows the fatalities per billion journeys, hours, and kilometers for common modes of travel. It can reasonably be assumed that self-driving cars will have to fall squarely within the top three or four safest modes before they will be accepted. And they will have to be significantly safer than human-operated cars to be justifiable at all.
|Mode of Travel||Per Billion Journeys||Per Billion Hours||Per Billion Km|
|Car (human driver)||40||130||3.1|
In addition to the consequences discussed below, there is another one that is worth noting before moving on. Because they must be virtually perfect in order to solve the liability/safety issue, self-driving cars are likely still a very long way off. For air travel, the odds of a particular flight killing you are only 117 in a billion, or 0.0000117%. Put another way, even if you took a flight every single day of a 75-year life, your odds of dying of something other than a plane crash are 99.7%. These should be very sobering numbers for anyone working in the autonomous car business.
Point Two: Everyone will want an Autonomous Recreational Vehicle (ARV)
Assuming that the considerable challenges discussed above are eventually met, in the brave new world of self-driving cars every vehicle will do exactly the same thing. Once the algorithms take over, the car on the left (below) will behave exactly the same as the car on the right.
They will both obey all laws, including the speed limits, at all times. Most likely they will be programmed to maximize fuel economy by minimizing unnecessary acceleration. They will take turns at prudent speeds and brake judiciously, just in case the road conditions are not ideal. In effect, all cars will be driven by a risk-averse team of coders, engineers, actuaries, and lawyers. All of the fun, danger, excitement, etc. implied by the Corvette’s specification sheet and outward appearance will have nothing to do with what actually happens on the road. Buying such a car, in spite of that fact, would strike most of us as pointlessly wasteful at best and sadly pathetic at worst. When all the fun of driving is devoured by the National Autonomous Highway Infrastructure Grid, or some such bureaucratic contrivance, Corvettes and similar vehicles will quickly go extinct. But that does not mean the Yaris on the right will suddenly become the vehicle of choice either.
Remember point number one, above, self-driving cars will only be accepted after they have become entirely autonomous, with no driver obligations whatsoever, and only after they have proven to be comparable in safety to trains and airliners. Therefore, as with trains and airliners, there will be no need to pay any attention at all to either the vehicle itself or the driving conditions. The vehicle occupants will be passengers exclusively, never drivers. The autopilot will be, in effect, a personal 24×7 chauffeur. The pertinent question then becomes, What sort of vehicle will people choose to be driven around in?
This question instantly directs our attention away from the steering wheel and dashboard (the functional aspects of the car) and back into the vehicle itself (the amenities of the car). You have ten hours to kill and no driving responsibilities. What might you want back there? Looked at in this way, the question all but answers itself. Even today, when people (usually with money) are faced with this question, they choose a luxury tour bus or similar vehicle.
The game-changer here, the basis of the coming paradigm shift, is the ability to sleep (and do any number of other things) while riding. Imagine you live in DC and have a meeting tomorrow morning in Chicago. Currently, you have only one realistic option: fly, rent a car, and get a hotel. But with an autonomous RV you could easily drive there. It takes 12 hours to drive from DC to Chicago. No problem! Get in your ARV before 8pm the night before your meeting, set the destination on the autopilot, make popcorn, watch a movie, sleep 8 hours, wake up around 6 am, fry some eggs, ride your exercise bike, shower and shave, iron your shirt, get dressed, and look out the window just as your ARV pulls into the parking lot exactly on time. I assume we will have autonomous gas stations by then as well, so you can sleep right through any overnight refueling.
Here’s a quick, back of the envelope, calculation of the relative costs of driving versus flying from DC to Chicago and back for one and two travelers:
|Expense||1 Traveler||2 Travelers||1 Traveler||2 Travelers|
|Round trip airline ticket plus taxes, bag fees, etc.||$850||$1700||$0||$0|
|Hotel plus taxes and fees||$200||$200||$0||$0|
|Rental car or taxi||$75||$75||$0||$0|
1Assumes 22 mpg highway, $2.50 per gallon, and a 1,400-mile round trip.
2Assumes a $75,000 vehicle with a 200,000-mile useful life and a 1,400-mile round trip.
The small values for gas and depreciation under the flying option reflect the drive to and from the airport.
These numbers will certainly vary somewhat but, in general, the savings that could be realized through the use of an ARV, particularly as the number of travelers increases, are substantial. The option to sleep and multitask while riding—transforming a car into a mobile hotel room—will dramatically increase the acceptable driving radius for most people. The main obstacles to long-distance drives are the need to sleep, the expense of hotels and meals along the way, the discomfort of sitting motionless for long periods, and the tedium. Remove all of those things and driving (more accurately, riding) becomes an option in far more situations.
Point Three: The unintended (or at least unexpected) consequences of autonomous RVs will result in a major paradigm shift in our view of personal transportation.
Thus far this essay has emphasized only the practical, nuts and bolts reasons that people will rationally choose ARVs once the option is available. But very soon thereafter, a major social transformation will begin to set in.
Today, in 2016, long before autonomous cars actually exist, the ARV already functions as an extremely useful heuristic for AI prognostication. That is because an RV is at least as much a living space as it is a transportation device. In effect, it is any empty box that we can fill up with anything we can imagine. Using it as our model platform for the autonomous driving future focuses our attention on the million-and-one other things, beyond simply getting from A to B, that will suddenly become possible. Indeed, by the time the dust settles, the A to B part of the equation will seem like an afterthought by comparison.
Currently, there are two main ways to make a statement with a car: luxury and performance. The overwhelming majority of drivers who buy top-of-the-line performance cars never exploit even a tiny fraction of their cars’ capabilities, and the few who do routinely end up in prison or the hospital. Automotive technology long ago exceeded the driving abilities of all but the most skilled professionals. And even the most sedate and unassuming new cars are far faster than any sane person would ever drive them. This fact is equally true of off-road performance cars. Though the highways are choked with hulking four-by-fours and SUVs, the nation’s complement of mud trails, sand dunes, and back country roads remains pristinely untouched by such vehicles. As SUVs became mainstream, they quickly morphed into little more than beefy minivans or stylish station wagons. In general, performance vehicles make a bold statement that their human drivers are attracted to but rarely have the courage to repeat out loud. And, as mentioned already, once the algorithms take over, performance will cease to be a factor altogether.
Luxury is a different story. Whereas high performance is a largely unrealizable fantasy, it is possible, at least to some degree and in principle, to enjoy a vehicle’s creature comforts, style, and panache. But even luxury is tragically undermined by the need to actually drive the damned thing. The most luxurious cars (and tour buses) are designed to be chauffeur driven. They boast fully reclining rear seats, champagne fridges, TV monitors, curtains on the windows, and any number of other amenities that cannot be enjoyed by the driver. If you are forced to sit behind the wheel, even the most luxurious car is only marginally superior to a run-of-the-mill sedan.
So, with performance an already dubious pursuit and luxury an experience largely denied to the driver, the stage is set perfectly for a rapid acceptance of ARVs. Once performance has been completely removed from the equation and creature comforts are available to all of the passengers, ARVs will be fertile ground for all manner of creativity and excess.
The busy-bodies, bureaucrats, environmentalists, progressives, and other control enthusiasts will be horrified as this realization begins to dawn on them. The utopian dream of autonomous cars is currently arrayed by a fleet of hyper-efficient, soulless, Prius-like vehicles plying the nation’s roads and highways like a well-regimented ant colony. The hope is that car ownership will become an outdated concept inasmuch as all vehicles will behave in exactly the same way. It will be more efficient, it could be argued, to simply request a ride on your smartphone and treat the National Autonomous Highway Infrastructure Grid as a species of public transportation. If driving is no fun, why own a car?
But as soon as people begin to see the value and, quite frankly, the fun of an ARV, all of that utopianism and uniformity will go right out the window. Think about this: if car buyers were primarily motivated by practical considerations, almost everyone would drive a Prius. Instead, we spend exorbitant sums of money to nurture fantasies that are, as discussed, virtually impossible to realize out on the actual roads. Performance cars cannot be driven according to their specifications, and luxury is largely confined to the back seat. The fact that we continue to chase after these things, in spite of their futility, speaks to the overwhelming desire on the part of car buyers to use their cars as vehicles of fantasy as much as vehicles of transportation. With the advent of ARVs, performance will indeed disappear, but luxury will suddenly become a genuinely achievable goal, and car buyers will embrace it with relish.
The RV industry is already a mature institution, so we do not have to tax our imaginations much to envision the autonomous driving future. Type “luxury RV” into your search engine and you will begin to grasp the enormous breadth of options that are available today. Increase the size of this marketplace a hundredfold and the future of ARVs comes dimly into focus.
Soon after ARVs become available, they will begin forming the foundations of an entirely new facet of society, neither a car nor a house, but something in between and in addition to those familiar things. Today’s cars are not, for the most part, “social spaces”. ARVs will be. And social spaces operate according to very different sociological rules than mere tools and other utilitarian items. ARVs will be used to entertain friends, strangers, and business associates, and a whole raft of heretofore nonexistent expectations will magically materialize. I suspect it will be similar to the spontaneous emergence of netiquette, the constellation of behaviors that did not previously exist, but seem natural, somehow, given the unique dynamics of the internet. Exactly how the world of ARVs might evolve is anyone’s guess, but the sheer magnitude and reach of the phenomenon all but guarantees that it will transform our lives in many significant ways.
Scenario One: Flexible Daily Commute
In 2016, the daily grind for much of the working population begins with a dismal creep through bumper to bumper traffic from home to office. Few good options are available to mitigate this grim reality. Shortening the commute requires living closer to the city, and that entails either a much more expensive or much smaller residence. Such choices must be balanced against the primary disadvantages of a long commute: the emotional stress and wasted time. Yet ARVs will remove both of those unpleasant realities, radically changing the typical working stiff’s morning calculations.
Depending on the length of your commute, you could transfer as much of your morning routine as you like into your ARV. To maximize sack time, you could even roll directly out of bed and shuffle straight out to the garage, taking care of all of your personal primping and hygiene in your ARV, and even spending some time at your desk answering emails and getting a jump on the day. By removing the stress of gridlock as well as the lost productivity of sitting in it, the length of a commute will hardly matter at all.
Changing the commute calculus will have secondary effects as well. Suddenly, urban real estate will be less desirable and therefore less valuable. At the same time, the distant exurbs will no longer seem so far away. Together, these changes will have the effect of hugely increasing urban sprawl while equalizing real estate values across an entire region. This would be true of normal self-driving cars as well, though perhaps to a somewhat smaller degree. ARVs, unlike cars, will render the length of the commute largely irrelevant.
Scenario Two: Road Trip!
Nothing quite as dramatic as the epic road trip depicted below is needed to make this point, but it serves its purpose as well as any other example.
If you took it into your head to see the whole country one summer, you could use modern GPS algorithms to chart the most efficient possible course. But though that would minimize fuel use and driving time, it still leaves you on the hook for around 30 hotel rooms (some of the East Coast landmarks could be seen two to a day) and at least 60 restaurant bills (assuming you and your companion can live on only two meals a day). According to Randy Olson of MSU, the route in the above figure is 13,699 miles long which, assuming 20 mpg and $2.50/gal, would cost $1,712 to drive. We can liberally add $300/person for 30 days’ worth of food bought at the grocery store and brought along. If we assume that vehicle depreciation is comparable for an ARV and a regular car, then the savings for two ARV travelers are related to the hotels and restaurants [($120/night x 30 nights = $3,600) + ($15/meal-for-two x 60 meals = $900)] minus the $600 for the ARV-scenario grocery bill. Altogether, the ARV total is approximately $2,312 while the traditional car total is approximately $6,212. That nearly $4,000 difference is enough to tip the scales for most people who might consider such an adventure. Even if we assume that a car can get twice the mileage per gallon (40 mpg) as an ARV, the difference is still more than $3,000.
This example is admittedly extreme, but it is comparable to any road trip on which hotels and restaurants would be a significant part of the expense. If you plan your trip carefully, you could arrive at your destination in the morning, spend the day sightseeing or walking the beach, and then travel to the next destination during the night. Better still, you could program the autopilot to drive at a speed that will bring you to your next destination at the right time, not simply as fast as possible. For example, if the distance to your next destination is only 300 miles, and you do not wish to arrive at three o’clock in the morning, you could let the autopilot calculate your speed (300 miles/8 hours=37.5mph) letting you to get there at a reasonable time while saving gas in the process.
Ultimately, it won’t be the practical, economic considerations that make ARVs really take off. That dry calculus is merely what will make them possible. People will embrace them because they offer endless possibilities and will be loads of fun.