Rise of the Robots: Technology and the Threat of a Jobless Future (28 page)

Three-dimensional printers can be used with virtually any type of material, and the technology is finding many important uses outside of manufacturing. Perhaps the most exotic application is in printing human organs. San Diego–based Organovo, a company that specializes in bio-printing, has already fabricated experimental human liver and bone tissue by 3D-printing material containing human cells. The company hopes to produce a complete printed liver by the end of 2014. These initial efforts would produce organs for research or drug testing. Organs suitable for transplant likely remain at least a decade in the future, but if the technology arrives, the implications would be staggering for the roughly 120,000 people awaiting organ transplants in the United States alone.
⁴
Aside from addressing the shortage, 3D printing would also allow organs to be fabricated from a patient’s own stem cells, essentially eliminating the danger of rejection after a transplant.

Food printing is another popular application. Hod Lipson suggests in his 2013 book
Fabricated: The New World of 3D Printing
that digital cuisine may turn out to be 3D printing’s “killer app”—in other words, the application that motivates huge numbers of people to go out and buy a home printer.
⁵
Food printers are currently used to produce designer cookies, pastries, and chocolates, but they also have the potential to combine ingredients in unique ways, synthesizing unprecedented tastes and textures. Perhaps someday 3D food printers will be ubiquitous in home and restaurant kitchens, and gourmet chefs will be subjected to the same type of winner-take-all digital market that professional musicians currently face.

The biggest disruption of all could come when 3D printers are scaled up to construction size. Behrokh Khoshnevis, an engineering professor at the University of Southern California, is building a massive 3D printer capable of fabricating a house in just twenty-four
hours. The machine runs on temporary rails alongside the construction site and has a huge printer nozzle that deposits layers of concrete under computer control. The process is entirely automated, and the resulting walls are substantially stronger than those built using traditional techniques.
⁶
The printer could be used to build homes, office buildings, and even multi-level towers. Currently the machine builds only the structure’s concrete walls, leaving workers to install doors, windows, and other fittings. However, it is easy to imagine future construction printers being upgraded to handle multiple materials.

The impact of 3D printing on manufacturing may be relatively muted simply because factories are already highly automated. The story could be very different in the construction industry. Building wood-frame homes is one of the most labor-intensive areas of the economy and offers one of the few remaining occupational opportunities for relatively unskilled workers. In the United States alone nearly 6 million people are employed in the construction sector, while the International Labour Organisation estimates that global construction employment is nearly 110 million.
⁷
Three-dimensional construction printers might someday result in better and cheaper homes, as well as radically new architectural possibilities—but the technology could also eliminate untold millions of jobs.

Autonomous Cars

The self-driving car entered the final stretch on the road that would take it from science fiction to everyday reality on March 13, 2004. That date marked the first DARPA Grand Challenge—a race that the Defense Advanced Research Projects Agency hoped would help jump-start progress in the development of autonomous military vehicles. Fifteen robotic vehicles set off on a course that began near the town of Barstow, California, and wound its way 150 miles across the Mojave Desert. At stake was a $1 million prize for the first contestant to cross the finish line. The results were underwhelming.
None of the vehicles managed to complete even 10 percent of the course. The best effort came from Carnegie Mellon University’s modified Humvee, which careened off the road after just seven and a half miles and plunged into an embankment. DARPA declared the race a bust and kept its money.

The agency saw promise, however; it scheduled a rematch and upgraded the prize to $2 million. The second race was held on October 8, 2005, and required the robotic vehicles to navigate more than one hundred sharp turns, pass through three tunnels, and trek across a mountain pass with sheer drop-offs on both sides of the winding dirt path. The progress was astonishing. After just eighteen months of continued development, five of these vehicles leapt literally from the ditch to the finish line. The winning entry, a modified Volkswagen Touareg designed by a team led by Stanford University’s Sebastian Thrun, completed the race in just under seven hours. Carnegie Mellon’s refined Humvee design crossed the finish line about ten minutes later. Two other vehicles followed within half an hour.

DARPA staged yet another challenge in November 2007. This time the agency created an urban setting in which robotic vehicles shared the road with a fleet of thirty Ford Tauruses manned by professional drivers. The self-driving cars had to obey traffic regulations, merge into traffic, park, and negotiate busy intersections. Six out of thirty-five robotic vehicles managed to complete the course. Stanford’s car was once again first over the finish line but was later demoted to second place after judges analyzed the data and subtracted points for infractions of California’s driving laws.
⁸

Google’s autonomous-car project got its start in 2008. Sebastian Thrun, who had come to the company a year earlier to work on the Street View project, was put in charge, and Google began to rapidly hoover up the best engineers who had worked on the vehicles entered in the DARPA races. Over the course of two years, the team developed a modified Toyota Prius packed with sophisticated equipment, including cameras, four separate radar systems, and an $80,000 laser
range finder capable of creating a complete three-dimensional model of the car’s environment. The cars can track vehicles, objects, and pedestrians; read traffic signs; and handle nearly any driving scenario. As of 2012, Google’s autonomous fleet had driven over 300,000 accident-free miles on roads ranging from freeways jammed with stop-and-go traffic to San Francisco’s famously convoluted Lombard Street. In October 2013, the company released data showing that its cars consistently outperformed the typical human driver in terms of smooth acceleration and braking, as well as general defensive driving practices.
⁹

Google’s project has had a galvanizing effect on the automotive industry. Virtually every major car manufacturer has since announced plans to implement at least a semi-autonomous driving system within the next decade or so. The current leader is Mercedes-Benz. The 2014 S-Class is already capable of driving autonomously in stop-and-go city traffic or on the Autobahn at up to 120 miles per hour. The system locks onto either lane markings or the car ahead and handles steering, acceleration, and braking. Mercedes has initially chosen to take a cautious approach, however, and the driver is required to keep his or her hands on the steering wheel at all times.

Indeed, the systems under development within the automotive industry are almost universally geared toward partial automation—the idea being that the human driver always maintains ultimate control. Liability in the event of an accident may be one of thorniest potential issues surrounding fully automated cars; some analysts have suggested that there might be ambiguity as to who would be responsible. Chris Urmson, one of the engineers who led Google’s car project, said at an industry conference in 2013 that such concerns are misplaced, and that current US law makes it clear that the car’s manufacturer would be responsible in the event of an accident. It’s hard to imagine anything the automotive industry would fear more. Deep-pocketed manufacturers would make irresistible targets for attorneys wielding product liability claims. Urmson went on to
argue, however, that because automated cars continuously collect and store operational data that would offer a comprehensive picture of the car’s environment up to the moment of the accident, it would be nearly impossible to succeed with a frivolous lawsuit.
¹⁰
Still, no technology is 100 percent reliable, and it’s therefore inevitable that an autonomous system will eventually cause an accident that confronts its manufacturer with a daunting liability judgment. One possible solution would be laws placing reasonable limits on such lawsuits.

The semi-autonomous approach creates problems of its own, however. None of the systems are yet capable of handling every situation. Google’s corporate blog noted in 2012 that, while progress on self-driving cars has been encouraging, “there’s still a long road ahead” and that its cars still “need to master snow-covered roadways, interpret temporary construction signals and handle other tricky situations that many drivers encounter.”
¹¹
The grey area where a car may need to detect that it is encountering an unmanageable situation and then successfully return control to the driver probably represents the technology’s greatest weakness. The engineers working on the systems have found that it takes about ten seconds to alert the driver and ensure that he or she regains control of the vehicle. In other words, the system has to anticipate a potential problem well before the car actually gets into trouble; accomplishing that with a high degree of reliability is a substantial technical challenge. This would be made worse if drivers were not required to keep their hands on the wheel during automated driving. One Audi official noted that when the system being developed by the company is engaged, the driver is “not allowed to sleep, read a newspaper, or a use a laptop.”
¹²
It’s unclear how the company plans to enforce that—or if using a smart phone, watching a movie, or engaging in any number of other distractions would be allowed.

Once such hurdles are overcome, autonomous cars offer enormous potential, especially in terms of improved safety. In 2009, there were about 11 million automobile accidents in the United States, and
about 34,000 people were killed in collisions. Globally, about one and a quarter million people are killed on roads each year.
¹³
The National Transportation Safety Board estimates that 90 percent of accidents occur primarily because of human error. In other words, an enormous number of lives might be saved by truly reliable self-driving technology. Preliminary data suggests that the collision avoidance systems now available in some cars are already having a positive impact. A study of insurance claim data by the Highway Loss Data Institute found that some Volvo models equipped with such systems experienced roughly 15 percent fewer accidents than comparable cars without the technology.
¹⁴

Aside from accident avoidance, self-driving car proponents point to many other potential upsides. Autonomous cars will be able to communicate and collaborate with each other. They might travel in convoys, riding in each other’s draft to save fuel. High-speed coordination on freeways would reduce, or perhaps even virtually eliminate, traffic jams. Here, I think the hype is running substantially ahead of any near-term reality. Benefits of this type rely heavily on a network effect: a substantial fraction of the cars on the road would need to be autonomous. The obvious reality is that a great many drivers are going to be, at best, ambivalent about self-driving technology. A lot of people simply like to drive. Enthusiast magazines like
Motor Trend
and
Car and Driver
have millions of subscribers. What, after all, is the point of owning “the ultimate driving machine” if you aren’t going to drive it? Even among drivers who embrace the technology, adoption is likely to be quite gradual. One consequence of soaring income inequality and decades of stagnant incomes is that new cars are becoming increasingly unaffordable to a large fraction of the population. Indeed, recent data suggests that American consumers are in no rush to trade in the vehicles they have. In 2012, the average car on the road in the United States was nearly eleven years old—an all-time record.

In some cases, a mixture of human and robotic drivers might actually lead to more problems. Think about the last aggressive
driver you encountered—the person who cut you off or perhaps weaved recklessly between lanes on the highway. Now imagine that person sharing the road with autonomous cars he or she knows are programmed to be flawlessly defensive in all situations. Such “wolf among sheep” scenarios might invite even more risky behavior.

The most optimistic boosters of self-driving car technology expect a major impact within five to ten years. I suspect that technical challenges, social acceptance, and obstacles related to liability and regulation may make such projections seem overly optimistic. Nonetheless, I think there’s little doubt that truly autonomous—or in other words “driverless”—vehicles will eventually arrive. When they do, they will have the potential to revolutionize not just the automotive industry but entire sectors of our economy and job market, as well as the fundamental relationship between people and automobiles.

Perhaps the most important thing to understand about a future in which your car is fully autonomous is that it probably
won’t be your car.
Most people who have given serious thought to the optimal role of self-driving cars seem to agree that, at least in densely populated areas, they are likely to be a shared resource. This has been Google’s intent from the start. As Google co-founder Sergey Brin explained to the
New Yorker
’s Burkhard Bilger, “[L]ook outside, and walk through parking lots and past multilane roads: the transportation infrastructure dominates. It’s a huge tax on the land.”
¹⁵