Will Driverless Cars Be Safe?

photo from the DARPA Urban Challenge gallery

A car that drives itself. This technology sounds really scary, but ironically, that's what will ultimately make it safe. Consider elevators--the consequences of an elevator failure are so frightening that they tend to be over-engineered, with redundant safety systems that make them extremely reliable. Now, consider the complexity of navigating a moving vehicle among thousands of other moving vehicles. Also consider how paranoid engineers will be about integrating this technology with threats of liability lawsuits hanging over their heads. When the "blue screen of death" could literally mean death, it's easy to understand why the technology is being rolled out in stages. Many car models are already available with automatic, variable-speed cruise control and collision-avoidance systems. This interim stage of driving automation will allow developers adequate time to work out every last bug. During the 2007 DARPA Urban Challenge, a vehicle mistakenly parked itself in a carport where it didn't belong, and although this didn't result in catastrophic collision, it's easy to imagine how these bugs might play out in less tolerant environments, such as a busy freeway.

The real irony here is that we're criticizing the mistakes of machines, but with an estimated 93% error rate, humans aren't very good at driving, either: they're impatient and drive too fast, they're unpredictable, they text message on cell phones in traffic, and they let emotions influence their driving behavior. Computer-driven cars have the advantage in almost every category: instantaneous reaction time in emergencies, routine algorithms that determine when to brake or accelerate, sensors that are not limited by blind spots, and most importantly, the potential for networking. Your car will know before you will when the car ahead is about to make a right turn, or brake suddenly to avoid a collision. Features like these, once they've been thoroughly tested, may one day bring down the number of auto collisions, as well as significantly improve traffic flow, perhaps eliminating traffic jams. Maybe you'll even get an insurance discount when you switch on "auto-drive."

The only real questions remaining are these: how long is your commute, and what will you do with all of that extra time?