Driverless cars (DCs). The concept is everywhere these days, and according to many futurists the actual cars soon will be, too. Every tech outfit worth mentioning has a finger or two in the DC pie and several big-name consortiums already have prototypes rolling. A few cities have okayed test programs, and Colorado legislators, ever alert to the chance to lure more technology dollars to the state, are proposing a friendly set of regulations designed to make our admittedly deteriorating roads more attractive to robot rides. A bill being considered in the legislature would set state standards for testing of DCs and preempt Colorado cities from enacting more restrictive rules (Boulder, for example, is rumored to favor allowing only electric vehicles; each powered, one might suppose, by its own wind turbine). The most optimistic press releases have the technology highway-ready in five years or less.
As a lifelong technophile, I’m usually excited about this kind of forward leap. But as a pragmatic Dilbert-type, experience has taught me to be skeptical of anyone touting some revolutionary breakthrough that will forever change the way we get around. After all, I’m still waiting for my flying car. So let’s take a clear-headed look at the promise of the DC.
State Sen.Owen Hill is one of the sponsors of the Colorado DC initiative. In a recent interview, Hill hyped the safety angle of going driverless, allowing that DCs could prevent nearly all of the 40,000 highway deaths that occur each year by “removing human error” from the equation. This kind of statement could only come from someone blithely ignorant of the level of complication involved in building such a system, one created, initiated – and debugged – by humans. The human error factor won’t be eliminated. It will only be moved to another part of the process. And humans, at least when it comes to driving, are a lot more capable that we get credit for. Navigating an automobile through the real world is an extremely complex and variable undertaking, but 99% of human drivers handle it well 99% of the time. The explanation for that is our ability to learn from experience. Every mile we drive adds to our experiential library, and we have the unique ability to not just remember events but to absorb and reconstitute them to meet and deal with new, unprecedented situations. It’s called intelligence. We have it. Computers, at least for the moment, do not.
So what is called for in the DC is AI. Yep, artificial intelligence: the Holy Grail of technology, enabling a computer to do what 16-year-olds with learner’s permits have been doing since the 1920’s. When we drive a car, we (most of us, at any rate) think and reason. Computers in DCs execute their programming. Staying in a lane, keeping a safe distance from the car ahead, stopping at red lights (a novel concept), avoiding old ladies walking their Pomeranians, all these actions are relatively easy to program. Experienced drivers do them reflexively, often while texting, digging a Tic Tac from between the seat cushions, yelling at the kids or applying mascara. But sooner or later a situation will arrive that demands who or whatever is in control of a vehicle to make a split-second decision based on maybe 10 data variables. If the programmers have missed even one of these, tragedy will ensue. A human driver is equipped to take in all 10 and respond effectively. Computers are not there yet.
A completely computer-driven transportation system would be an unprecedented achievement, but we’re not talking about that. We are talking about a construct which will have to accommodate and integrate with human-piloted vehicles far into the future. DCs will be mixed with UPS trucks, yardcare pickups towing trailers, buses, street sweepers, snowplows, bike couriers and 20 years’ worth of used Civics and Corollas, all being driven, with varying degrees of incompetence and distraction, by humans. DCs will undoubtedly be programmed to provide utmost safety for their occupants, so they will take no chances in traffic. Real drivers will soon recognize, and try to take advantage of, this tendency. Imagine navigating rush hour with a few thousand super- cautious robot drivers clogging traffic lanes and an equal number of super-aggressive flesh-and-blood dolts doing their best to get ahead. Think that will make for error free mobility? The only way to totally eliminate human error is to eliminate humans. In other words, every vehicle on every street and highway not just under the control of its onboard computer but in constant contact with every other vehicle anywhere nearby and with a central traffic management system keeping track of the whole mess. Imagine a cross between a beehive and a Mars landing, all with a couple levels of redundancy. Such a system would use up most of the servers on the internet just to run Denver.
But assume for the sake of argument that we can create this system. Who will pony up the bucks to retrofit the millions of used cars already plying the streets and alleys of the country with tech that may cost three times what the cars are worth? What happens when a hailstorm wipes out the sensor arrays of several hundred cars while they are rolling down the Interstate? Not to mention vandalism. And who is Frank Azar going to go after when, not if, an accident happens? The owner of the car (who was in the back seat doing the Times crossword)? The manufacturer? Google?
These are some, but by no means all, of the problems to solve and the policy issues to decide, before we can become a society free of fender benders and road rage. DCs in their beta form may be on some roads, somewhere, maybe even in Colorado, within the next five years or so. But the day when you can climb into a true DC that will take you anywhere you need to go quickly, cheaply and uneventfully is probably a lot further away than prognosticators would have you believe.
Which is fine with me. I like driving. Besides, it’d be no fun flipping the bird at an empty driver’s seat.