The real problem with self-driving cars
Whatever happened to baby steps?
Last week a 49 year-old Arizona woman was hit and killed by an Uber self-driving car as she tried to walk her bicycle across a road. This first-ever fatal accident involving a self-driving vehicle and a pedestrian has caused both rethinking and finger-pointing in the emerging industry, with Uber temporarily halting tests while it figures out what went wrong and Google’s Waymo division claiming that its self-driving technology would have handled the same incident without injury. Maybe, but I think the more important question is whether these companies are even striving for the correct goal with their cars? I fear that they are over-reaching and simply trying to do too much too soon.
Let’s be clear: the obvious goal of Uber is to eliminate drivers from its fleet, saving money. It’s likely, too, that no Uber passenger will ever be harassed or raped by a self-driving car. Waymo’s goal is probably the same given that the company says it will shortly begin testing its own driverless car service, also in Arizona.
The self-driving goal for private cars owned by their passengers is likely to be somewhat different. I might want to read a book or watch a video while my car drives itself. Or maybe I’d say "go find a parking place." This certainly seems to be where Tesla, for example, seems headed with its autopilot feature.
But what are the societal goals for self-driving vehicles? These come down to safety and increasing system capacity. Society has little incentive to put human drivers out of work and on the dole.
Cars driven by computers are supposed to eventually be much better at driving themselves than human drivers, resulting in fewer accidents and deaths. If that driver is my 16 year-old son, Channing, we may already be there.
For government the clearest incentive for self-driving cars is to save money overall by putting more cars simultaneously on the road. Look at a Google’s worth of news stories about self-driving cars and you’ll hardly ever see this even mentioned, yet it is by far the biggest incentive of all. Which brings me to a story….
When we were shooting Triumph of the Nerds back in 1995 for PBS and Channel 4, series director Paul Sen came from London knowing only one person other than me in California -- his cousin who worked for the State of California as an engineer working on self-driving cars!
That was 23 years ago. Who knew this technology had been in the works for so long?
But Paul’s cousin’s tech wasn’t the self-driving tech we see today. His cars weren’t autonomous at all. They were part of a system that was, if not centrally controlled, certainly centrally monitored. The idea was to wire-up every highway and every car driving on that highway so they could work together to get as many passengers as possible from place to place. The selling points were simple: 1) cars would drive precisely one meter apart, increasing the number of simultaneous cars on the road and reducing energy consumption through NASCAR-style drafting to reduce drag and improve mileage, and; 2) every one of those cars would travel at the speed limit which might well be raised specifically for centrally-controlled driving.
Paul’s cousin told us that driving one meter apart would allow five times as many cars to share the highway, eliminating the need for new highway construction, saving billions. Cars would change lanes only to get on or off the highway or to go around obstructions like stalled cars or road repair crews. Everyone would drive everywhere at 80 mph, which was the rated design speed for the Interstate Highway System. Even driving at 80 mph the road would be safer because any collision would involve a differential acceleration of less than one meter, which was within the capability of the five mph bumpers of that era.
Even if the car ahead slammed on its brakes and you bumped into it, chances are there would be no damage.
Where are the self-driving proponents today telling us that they will save on highway construction, dramatically increase system capacity, and significantly reduce travel time? They don’t make any of those claims, citing only the possibility of improved safety as a societal benefit.
Ironically that system from 1995 could be implemented today almost for free using GPS, WiFi, and cellular data. No highways would need to be wired and even the need for servers would be minimal.
So why don’t we go back to that simpler idea of driving ourselves to the highway then letting the road take over until we exit? Well it misses the goal of autonomously driving on all kinds of roads, but more importantly it would be difficult to patent given the very work done those many years ago, now in the public domain.
Like most R&D today it comes down to controlling intellectual property and getting rich (or richer).
Now back to the idea of taking baby steps. Handling the highway part first would have saved enough money to more than pay for everything else. And we RIGHT NOW could be driving from San Francisco to Los Angeles at 80 mph while reading a book. On a door-to-door basis that would probably be just as fast as the $67 billion California High Speed Rail Project that won’t be ready until 2029.
While high speed rail saves energy on a seat-mile basis, the energy economics aren’t nearly as good if you count the energy used to build the system.
So why haven’t we been doing this incrementally? Well mass transit people like railroads and self-driving car moguls like monopolies so I guess that explains it. But to me it looks like both self-driving entrepreneurs and regulators aren’t doing their jobs right.