by senior futurist Richard Worzel, C.F.A.
➢ Within the next 5 years, self-driving cars will expand rapidly into those commercial sectors where the economic case is overwhelming, but will only replace privately-owned cars more gradually, say over 20 years. Some economic consequences will be profound.
➢ The benefits to cities are so remarkable that urban planners should immediately start preparing for smart cars and smart highways.
It’s always tempting, as a futurist, to make bold, sweeping prognostications, especially about an emerging technology. Some commentators have done just this, saying that self-driving cars will offer such compelling economic benefits that they will quickly reshape our cities, radically change our way of life, and eliminate many millions of jobs.[1]
Most other critics say the opposite, that self-driving cars will evolve gradually, and won’t make an appreciable dent in our driving habits or our economy for many years, possibly decades.
So, who’s right? As is often the way when you’re trying to puzzle out where the future is leading us, the answer is: some of each. Let’s look at the case for self-driving cars first.
Feature Creep Moves Us Towards AV Capability
The technologies behind self-driving cars (also called autonomous vehicles or “AVs”) are emerging very quickly, and may accelerate to the point where your next car may capable of driving itself if you so desire. Elon Musk, the head of Tesla Motors, as well as SpaceX, says that today’s Tesla cars could handle 90% of the driving conditions you face today[2], but not without supervision. He also says that he expects cars that can drive you to your destination while you sleep (i.e., operate fully autonomously) will appear within the next five years.[3]
These advances will be (pardon me) driven by a number of developments. Computers will continue to get more cost-effective at exponential speeds, allowing AVs to get smarter at least that quickly. Next, as companies like Google, Apple, and the major car companies race to claim the biggest possible piece of this new car market, the software will become steadily – and quickly – more sophisticated and effective.
Next, those cars whose operations are managed by software, as Tesla’s electric cars are, can be upgraded with a software patch, and don’t need to wait for the next model year. However, while software is easily upgradable, true AVs also need an array of expensive sensors, which will limit the number of fully AV-capable cars on the road.
Finally, AVs will probably be networked, which means that what happens with one AV can be quickly incorporated into software and mapping systems, and distributed to all AVs (or at least, all AVs of that manufacturer). This means that an unusual, potentially hazardous road, once it has been encountered by one AV, will become familiar to all AVs. As a result, the mapping of highways on the fine scale necessary for completely autonomous cars will happen at exponential speeds once the process begins in earnest.
Likewise, as computer smarts help AVs learn how to cope with more difficult kinds of driving conditions, such as snow-covered highways or freezing rain, these lessons will quickly spread throughout an AV fleet.
All of this would be like a teenage boy being able to learn from the experience of all other teenage boys, all over the world, in all kinds of climates, and on all kinds of roads through a kind of telepathic link-up. You can get smart awfully quickly that way.
Meanwhile, all cars will incorporate more and smarter features to assist drivers to drive more safely, always starting with luxury cars. This has been happening for decades, first with cruise control, then automatic braking systems, traction control, and, more recently, with things like lane-keeping, cross traffic warning, and blind spot warning systems. Such systems will continue to grow in sophistication and power, creeping towards AV capability.
Niche Markets Will Lead in Adopting AVs
One market ripe for AVs is Uber, which is currently challenging the survival of the taxi industry. Uber makes no bones about its desire to replace independently owned cars and drivers with Uber-owned self-driving cars. This would increase Uber’s effectiveness through computer optimization to serve demand. It would also decrease its costs, eliminate many of the current problems, such as worrying about drivers assaulting passengers, and allow Uber to keep virtually all of the money it collects rather than sharing it with freelancers.
So whether it’s Uber, or someone else interested in replacing the functions of cab drivers, it’s clear that in most urban markets, the idea of self-driving cars is compelling, and likely to be a niche where a number of players will become early adopters. Indeed, a Columbia University study indicated that you could replace today’s 13,000 NYC cabs with just 9,000 self-driving cars[4]. According to Columbia, such cars would offer faster response times (they estimate an average of 36 seconds) at significantly lower costs (50¢/mile), in large part because of the lower capital costs involved.
At a cost of approximately $25,000 per cab, reducing the number of vehicles required to supply NYC with cab service by 4,000 units produces a capital saving of $100 million. That’s a pretty enticing number for just one city.
This will also bring real benefits to cities that municipal planning groups should consider right now. Thirty percent fewer cabs means less gridlock, and a decrease of at least 30% in emissions from cabs, with lower health care costs. In fact, results would be better than that because a more efficient, responsive, and inexpensive cab system would also reduce the rate of personal car ownership, getting even more cars off the road. And if such cab-substitutes were also zero-emission cars[5], then emissions reductions would be even greater. All told, this could be a real win for cities, and at virtually no cost to them as it would be financed by companies like Uber.
The other related niche is for city dwellers who use car sharing schemes, such as Zipcar or AutoShare, rather than owning a car. Here again, the case for AVs is compelling: the car drives itself to your door, rather than forcing you to go get it. Then, when you’re done, it drops you at your door, and drives itself away. Indeed, it might be that cab and car-sharing companies would become one and the same.
Another niche where AVs will be welcomed is in long-distance trucking, where the skilled workforce is rapidly aging, and finding replacement drivers is increasingly difficult.
Moreover, as the pre-war and boomer generations age, more and more of them will reach the stage where they have difficulty driving, and some of them will lose their licenses. The ability of AV-capable cars to either compensate for aging reflexes, or take over the driving creates a large, and rapidly growing potential market for AVs and intelligent driver-assist cars, all of which will help finance the move towards AVs.
Smarter Cars Will Lead to Smarter Highways
As cars become autonomous, they will also become networked. As one example, a “car train”, being a set of individual cars, traveling in a bumper-to-bumper, coordinated pack, will be able to go much faster than they could under individual control, take less road space, and increase traffic flow and capacity.
And as this kind of smart coordination between cars increases, smart highways will begin to emerge to manage traffic flow, often using Fog computing techniques, which will, again, decrease congestion, emissions, and collisions.
It’s true that smart highways will require significant capital investments, but it’s money that would have to be spent on highway replacement and repair anyway, but it will just be done with sensors and better design in mind. Moreover, smart highways will actually save taxpayers money. Smart highways will cost more per mile, but they will cost significantly less per passenger-mile, reducing gridlock, highway repair and replacement costs, emissions, and the need for emergency services.
Best of all, smart highways can begin without having to replace existing roads at all by having a central computer interact with communication-enabled smart cars by radio – perhaps even by having cities buy digital messaging capacity on a wholesale basis from existing cellphone providers.
Then, as highway systems become more sophisticated, it will become easier to justify investments in ever-more sophisticated sensing and management systems that will make smart highways work even better with AVs, creating a virtuous circle, all pushing us towards smarter highways with smarter vehicles.
All of which means that city planners should be looking into the future, and planning for smarter, better, more cost-efficient infrastructure rather than just repairing the dumb infrastructure they inherited.
So, all told, the case for AVs seems like a slam-dunk. Or does it?
The Problems with AVs
The biggest barrier to AVs might be called Big Iron. That’s the stock of non-AV capable cars and trucks on the highways. Since most people aren’t going to junk their existing vehicles just to get somewhat more capable ones, it will take a gradual turnover as current cars wear out and are replaced by smarter ones. With the average age of cars and trucks in the U.S. being 11.4 years, it will take more than a decade, probably closer to 15-20 years, for truly smart, AV-capable cars to dominate the highways, especially as these features will appear first in luxury cars before trickling down to the rest of us.
Next, driving isn’t always easy, and there are still situations where developers acknowledge that AVs don’t do very well, such as in low visibility conditions. Heavy rain, snow, or fog, particularly when they obscure highway markings, can be particularly difficult for AVs. But then, such conditions are difficult for humans, too.
Likewise, driving on congested city streets, with bicycles, pedestrians, and other vehicles doing weird and often unpredictable things, present real challenges for AVs – just as they do for human drivers. Indeed, you could argue that the much faster computation and reaction speeds of AVs may do a better job of avoiding people and other vehicles in such situations.
However, while we accept that it’s OK for humans to take such risks, the creators of AVs can’t afford to do so just yet. Hence, even if AVs were better at driving in such conditions than humans, they might not be allowed to do so. More experience, improved results, and definitive statistics showing AV superiority will eventually allow AVs to function in such circumstances, but it will take time for humans to accept this.
The Non-Driving Problems of AVs
Which leads to the non-driving barriers to AVs, particularly laws, insurance, and liability.
Every state has a different attitude about self-driving cars, whether they are acceptable or safe, and whether to allow them to function on their own. This is creating a patchwork of laws that will take years or even decades to smooth out. After all, different jurisdictions have had almost a century of cars with human drivers to work out most legal differences.
Next are the interlocking questions of liability and insurance.
Google, the most prominent AV developer, says that its cars have been involved in a total of 17 collisions in more than 2.2 million miles of driving, and claims that none of these have been the car’s fault.[6] Typically what has happened is that an AV has been hit, either from the rear or the side, by human drivers who either don’t stop, expect the AV to keep moving, or to bend the rules of the road as a human would, rather than adhering to them strictly. The software will quickly overcome this problem, but collisions will still occur.
When they happen, they will create questions about liability, largely because people who sue in car collisions usually go after the party with the deepest pockets. I mean, who do you think would make a richer target: Google, or a little old lady in tennis sneakers?
And if an AV is in a collision, and is, for some reason, found to be at fault, who’s responsible? The human driver, whether he had his hands on the controls or not? The car company? The software provider? Or is there a chain of liability, where the driver is sued, then his insurance company in turn sues the software company? We don’t have answers to these questions yet, but you can be sure they will arise.
And whom does the insurer insure? The driver? The AV company? Or the software provider? Stay tuned to find out, because we don’t know at this point.
What Are the Broader Implications?
It’s clear that as AVs take the wheel from less-precise human drivers, there will be far fewer collisions, which means fewer deaths and lower insurance premiums, plus less need for police, fire, and emergency medical personnel.[7]
There will be fewer vehicles on the road, which means less gridlock, lower emissions, as noted earlier, plus a significant drop in demand for petroleum.
Fewer people will shell out the big bucks to own a car, but will instead buy car services on a piecemeal basis. This means more discretionary income will be freed up for other things, creating a de facto increase in standards of living – but at a cost.
Fewer cars also mean lower profits and the probable bankruptcy of many car companies, a decimation of jobs for auto workers, plus a lowering of demand for everything related to the world’s largest manufacturing industry.
And what kinds of jobs will cab drivers be able to find? Driving a cab is one of the grittiest, most dangerous jobs out there. When that’s not available, what do all those people do instead? With 171,000 cabs in the United States, that probably projects to between 300,000 to 400,000 people who will be out of work in the U.S. alone.
But the implications go far beyond this. Given the importance of automobile manufacturing, will the loss of jobs, and the decrease in economic output associated with making cars create a backlash against automation by consumers, voters, and governments? Surely such a backlash against automation is coming (which will be the subject of a future blog), but could this be the trigger that releases it?
Next, social patterns will change in ways large and small. The elderly will regain independence, whether or not they have drivers’ licenses. Children will be able to travel places unescorted – which means there will have to be a whole different kind of safeguards on AV travel. Sixteen-year-olds may not bother to get a driver’s license, and the parents of teenage boys may not see their insurance rates go through the roof when their kids start borrowing the car. Party hosts may once again start offering their guests “one for the road.”
Social status, and the car as a status symbol will change, although how is unclear. Will it be a real status symbol to own a car, or not to own a car? Will owning a pre-AV vehicle requiring manual control be a source of swank? Will we increasingly socialize in the flesh as mobility becomes easier, or will the trend towards virtual socializing continue?
We don’t know. The future of technology doesn’t depend on the capabilities of technology alone, but on what humanity chooses to do with new developments. After all, the Internet was originally created to be a robust communication system that could survive a thermonuclear war, not to allow people to shop for books, slinky underwear, and vacations at home.
The Bottom Line
Self-driving cars are going to emerge far faster than most people, especially most pundits, think, but they will not sweep all before them. Some sectors will race to be early adopters, like cabs, car-sharing systems, and long-distance trucking firms. Yet, most people will accept AVs more gradually, and AVs probably won’t constitute a majority of vehicles on the road until 2035 or beyond. But city planners would be well advised to drop everything and start work on smart highways right now, because the benefits are truly compelling.
So: baby, who will drive my car?
© Copyright, IF Research, November 2015.
[1] See, for instance, http://zackkanter.com/2015/01/23/how-ubers-autonomous-cars-will-destroy-10-million-jobs-by-2025/
[2] http://www.theverge.com/2014/10/2/6894875/elon-musk-says-next-years-tesla-cars-will-be-able-to-self-drive-90-percent-of-the-time
[3] http://www.cnet.com/news/elon-musk-sees-autonomous-cars-ready-sooner-than-previously-thought/
[4] http://www.slate.com/blogs/future_tense/2015/07/07/self_driving_cars_a_new_department_of_energy_sponsored_report_finds_self.html
[5] Ignoring, for the moment, that there is no such things as a zero emission car.
[6] http://www.nbcnews.com/tech/innovation/google-self-driving-cars-face-slow-going-california-dmv-n464331
[7] http://gizmodo.com/study-humans-driving-cars-are-more-likely-to-hurt-othe-1739440432
