Washington Metro’s computers crashed twice this past weekend, forcing all trains to stop and stranding passengers for up to 30 minutes. This is just the latest example of how the aging transit system is slowly falling apart.
It is hard to imagine today what kind of computers Metro used in 1976, when it opened DC’s first new rail line. Programming probably used COBOL or some other now-archaic language. (The Antiplanner has heard rumors that the COBOL programmers who wrote the software that runs the San Francisco BART system refuse to ride the trains.) Anyone who has an older computer knows that things go wrong and those cumulative failures add up until eventually the system just does not reliably work.
In any case, the Metro system has roughly a $10 billion maintenance backlog. As a result, rails break; trains fall apart during operation; computers crash; and the agency’s bureaucracy can’t even keep up with the problems.
Rail advocates never mention the cost of maintenance when they propose new rail lines. Instead, they call highways “antiquated” and argue that a higher share of federal transportation dollars–most of which come from highway user fees–should go to transit. In fact, it is rail transit that is antiquated, while the real future of transportation will continue to use highways.
As Google chair Eric Schmidt recently predicted, self-driving cars will “become the predominant mode of transportation in our lifetime.” Ironically, George Mason University economist Tyler Cowan reports that DC police pulled over the Google driverless car for some apparent infraction.
Cowan, who has written favorably about driverless cars before, should have checked his facts. Slate staff write Will Oremus did, and learned that the car happened to be parked where Cowan spotted it and the police officers came over to learn more about them.
While the Antiplanner has previously stated that the main obstacles to self-driving cars are institutional, some technical obstacles remain. First, the laser-beam device that sits on top of most of the driverless cars that have been tested in the last decade costs $70,000. Significant improvements will be needed to reduce this cost to affordable levels. Second, the on-board software needs to carefully map every street, traffic signal, and traffic law in the country. This means people who travel all over the country will require reliable memory drives in the terabyte, if not petabyte, range. Both of these problems are solvable, and you are likely to see cars fully capable of driving themselves on the market by the end of this decade.
One data correction: Schmidt was quoted as saying that 35,000 people die each year in drunk-driving accidents in the United States, and this number has “remained constant over two or three decades.” Both these statements are wrong. In 2010 and 2011, fewer than 33,000 people per year were killed in all U.S. auto accidents, and less than a third of those accidents involved alcohol. Both auto fatalities and alcohol-related auto fatalities have declined dramatically over the last several decades. Still, self-driving cars can save a lot of lives.