Robots, and Accidents, on the Roads
Are you a better driver than a robot?
Posted Mar 27, 2018
Humans are lousy drivers. We get distracted and bored. We drive when we’re tired and drunk. I’ve hoped that self-driving cars will be safer. Maybe the robots can save us from ourselves. Or maybe not.
My hopes for self-driving cars reflect the possibility that robots will correct many of our faults. I’ve often written about the danger of using your cell phone when driving (e.g., going on autopilot; hang up and drive). Robots shouldn’t be distracted by their cell phones. I’ve written about using rock-n-roll to stay alert when driving while tired. Robot drivers shouldn’t get bored or tired when driving long distances. I’m also optimistic that robotic, self-driving cars won’t drive after drinking — instead, self-driving cars may get people home safely. In many ways, robots should be better than us.
I recently rode as a passenger in a partially self-driving car. It was way cool. The new car has a fancy type of cruise control (I think you can find this on several different makes and models). Personally, I’ve used old-school cruise control. You set a speed, and the car holds exactly to that speed. But you must constantly watch the cars in front of you. When they slow down, you must adjust.
The new cruise control works better. You set a speed. Then the artificial intelligence built into the car tracks the cars in front of you. If those cars slow down, your fancy new car slows down too. It was amazing. The car would then return to cruising speed when the traffic in front cleared. You can even set your preferred following distance, to a certain extent. My friend still had to steer. But the car even helped with that! If he drifted out of his lane without using a turn signal, his car beeped at him. This wasn’t a fully self-driving, robot car. Instead, this robot car augments human control. Maintaining speed and following distance. Warning you if you aren’t staying in your lane. Pretty cool and very helpful on a freeway.
Of course, completely self-driving cars are already cruising on some roads. Maybe you’ve seen one? So far, the robots are only in limited locations. But I’m sure they’ll be coming soon to a road near you.
Unfortunately, I’ve questioned my hopes for robot cars this week. A self-driving car hit and killed a woman. As with any accident injury or death, I feel awful for the woman and her family. The accident happened in Arizona — a state that has actively encouraged the development of self-driving cars. The car was an Uber self-driving car with a human in the car for emergencies. The details of the accident aren’t completely clear yet. From the reports I’ve read, the woman was crossing the street in a place without a crosswalk. She was pushing a bicycle. She was also crossing from the left to the right, which would have given the car (and back-up driver) more opportunity to see and respond to her. I’ve also read a report that the road was badly designed so that it implied crossing to pedestrians, but not to drivers. One report is that the human back-up driver had no time to take over and respond. Some video from the car has been released, and you can find that on the internet (link to the video and a report). With respect to the video, keep in mind that the video is from a dash cam. That is not the information that the robot car was using. Nor is it what a driver would have seen — the human eye is substantially better than that camera. Responsibility hasn’t been attributed as I write this. But no matter who was at fault, one thing is clear: Self-driving cars are not perfect.
Driving is hard. You track lots of information about your car, the road, the signs, the signals, other vehicles, bicycles, and pedestrians. Although humans are lousy drivers, we are remarkably adept at tracking a complex visual world. We can integrate a lot of information incredibly quickly. Our problems are human ones. We get distracted, bored, tired, and drunk.
Robots may solve our human problems. They should be less likely to get distracted, bored, tired, and drunk. But robots have their own problems. Integrating large amounts of visual information is hard. Predicting what the pedestrian beside the road is going to do is particularly hard.
I nonetheless remain hopeful about the future of self-driving cars. I suspect the robots will eventually take over the roads. The transition will not, however, be accident-free. Perhaps during the transition, self-driving cars should focus on the simple conditions. Oddly, the simple conditions can be hard on humans. On long highway drives, we get bored and distracted and tired. Long drives could be the perfect time for our robots to take the wheel.
The long-term goal is for everyone to be safer as they get where they are going. Roads should be safe for drivers, passengers, bicyclists, and pedestrians. Hopefully, the robots will help.