Robot Cars?

Robot Cars? In My Nevada?

Singularity Hub wrote an article last week about Nevada’s Assembly Bill 511: A proposed law setting down some guidelines for robot cars in the state. Actually, most of the bill seems to be directed to plug-in electric cars, giving them a distinctive decal and access to HOV lanes along with free parking at some public areas around Nevada. Section 8 covers autonomous cars, and directs the DMV to come up with licensing and operation guidelines for autonomous cars. Singularity Hub does a great job breaking the bill down (although it’s only a few pages, as written, and not too chock full of legalese).
Instead of rehashing what Singularity Hub said, I’ll approach the bill from a different direction: Will people actually want these cars?

The benefits are clear. If the car is driving itself, then there is little reason to worry whether the person in the car is distracted by conversation, or texting, or drunk. People can read the paper on the way to work, or watch a movie on a road trip across country. There are efficiency gains to be had as well. If owners can upload a list of locations, the car can optimize those locations to save time and fuel. With a network managing all the autonomous cars, each individual car can be routed through the most efficient roads, bypassing accidents. With complete control, computers can drive each car more quickly, and more safely, than humans can: NASCAR drivers already know about drafting (driving closely behind another car to take advantage of the displaced wind resistance) but Grandma Smith certainly doesn’t, and couldn’t perform the maneuver safely if she did. As cars become less individualized and more integrated through central hubs they begin to act more like a swarm, and can operate in tandem because the computer always knows what the car next to you is currently doing and about to do. There are safety advantages too. Computers can fly airplanes so aerodynamically unstable that the plane would fall out of the sky without computer control. Given that, managing a blowout or oil slick should be no problem for a computer controlled car.

In short, with a smart network cars should be able to maximize the roads to provide each person a better driving experience.

On the other hand, there won’t be much of a driving experience, will there?

I’m far from a driving junkie: Most times I just want to get where I’m going. All of my cars had automatic transmissions, including my newest Mustang. I promised myself that if I ever got a real sports car I’d get a manual transmission, but I hadn’t really felt the ‘thrill’ of a manual transmission. It just seemed like unnecessary work. Then I got my first motorcycle, and before too long I started to ‘get it.’ While I still enjoy driving my car, compared to a sportbike, driving a car is boring. Some of the experience has to do with the environment: The engine roaring, the wind enveloping you, the agility and acceleration of sheer horsepower available at a flick of the wrist. A computer ought to be able to control a motorcycle and the person riding it ought to have many of those same experiences.

The overall riding experience, however, will be very different just because the rider is no longer controlling the bike (or car.) It’s not -just- that a bike is fast and agile and that the engine roars: It’s that, for a short time, a rider really feels in control of the machine. The two merge, and the thoughts of the rider through small wrist movements and balance shifts translate into a crisp, quick turn. The rider gets a sense of control, and perhaps a little pride, from mastery over the machine. Computerized cars (or motorcycles) lose that sense of control because, by definition, autonomous cars will have a computer in control, and the passenger is literally just along for the ride.

I can imagine wanting an autonomous car for day to day driving. There is no excitement (hopefully), and no pride, in managing stop and go traffic. For about three weeks when I was 16 I took some pride in navigating stop and go traffic without incident, but since then driving in the city has mostly been a chore. I certainly don’t need (or expect) much excitement on the way to school or the office, and so a car that just gets me there works out nicely. But sometime I want to get out and ride. Sometimes I want to be the one in control, and hit some mountain roads for a burst of adrenaline.

Certainly some people will want to continue driving their cars, and that presents a new level of difficulty for autonomous cars. All of the integrated swarm analysis above depends on the computer being able to predict what the vehicles around each other will do: If a human is driving one of those cars the computer will not be able to predict that car’s actions. Drafting at 150mph is (perhaps) efficient if computers are controlling the entire chain, but dangerous if a person slides into the middle of that chain and loses control of their car. There are, I think, only two ways of handling the computer-human mix: Either people will be required to buy autonomous cars, or they will have to be separated from autonomous cars.

The first proposition seems unlikely. First, plenty of people (including, but not limited to, criminals) will want to drive sometimes. They bought their vehicles, and for as long as they run people want to drive them. While Congress could probably pass a law that forbids the sale of human controlled vehicles, enterprising folks will continue to make them (they should be substantially the same except for a few electronic components.)

The second proposition is more likely, but expensive. At minimum, a barricade in the middle of every road would need to be built, along with separate merging ramps. Some way of identifying that each car on that side of the barricade is autonomous would be needed as well, though a short transmission from each car should keep the honest people honest. None of that construction is going to be cheap, although it might provide a national recovery effort, New Deal style, as tens or hundreds of thousands of people are hired to modernize the roads in the US.

Barring either of those propositions, people will need to accept that accidents will happen at a more frequent rate than if computers were controlling all of the cars, but still probably a lesser rate than now when humans are driving (almost) everything.

As with most technology, there seems to be a spectrum of ideals with the major outliers marking the end points. At one extreme, people fear and dislike change, and are hesitant to adopt any new technology; especially one that involves a personalized activity like driving a vehicle. On the other extreme are the technophiles, eager to use the next new thing undeterred by early bugs. Because I imagine the vast majority of people often just want to get to their destination safely, I suspect that the technophiles will jump on automated cars, the bugs will get worked out and their safety will be proven over the next five or ten years, and then the majority will begin to replace their cars with automated vehicles. Some subset will never get them, including the elderly (who, generally, are resistant to technology) despite how useful automated cars would be to people reaching the end of their safe-driving ability.

Autonomous cars are coming, and I hope the human driving experience remains after they get here.