As soon as new technologies come along, the public immediately dials up their natural fear of all things technologic, and start imagining all the things that can go wrong.  This technophobic trend has naturally been extended to the concept of the self-driving car.  As soon as the possibility of self-driving cars entered the public consciousness, people started imagining assuming that those evil robot cars would make buggy, counterintuitive decisions in every case, causing the deaths of hordes of innocent citizens on a daily basis.  (Totally unlike, say, the hordes of innocent citizens killed by human drivers on a daily basis.)

The current trend along these lines is to place self-driving cars in “ethical” dilemmas, to illustrate their inherent danger to their human occupants and obstacles… this Phys.org article is a standard version of this trend.

But just about every scenario imagined has been based on a false dilemma, a blatant exaggeration of possibilities that even a sixth grader should be able to see past.

typical illustration of the self-driving car's false dilemmaThe illustration above shows a typical scenario given a self-driving car: Upon the sudden appearance of a pedestrian or crowd in the street, does the self-driving car:

  • swerve to avoid the crowd, even if that means intentionally killing a pedestrian who is not in the road (and presumably not in immediate danger);
  • swerve to avoid a single pedestrian in the road, even if that means having an accident that injures or kills the car’s passengers; or
  • swerve to avoid a crowd in the road, even if that means having an accident that injures or kills the car’s passengers?

The question basically assumes there are no better solutions, and the car will be “forced to make a moral choice” to pick who it must injure or kill.

And that assumption is flat-out wrong.

To begin with, there’s no morality or ethics involved with a self-driving car’s actions to avoid collisions; they are designed to stop as quickly as possible—not to swerve like a maniac driver in a buddy-cop movie.  So it’s not going to be deliberating on the comparable value of life of individuals and “choosing” directions in which to swerve.

It must also be noted that in tests of self-driving cars so far—and this has included numerous tests of cars driving on public roads and executing cross-country trips—self-driving cars have already demonstrated exemplary driving prowess, and have covered thousands of miles without causing a single accident.  (So far, every documented accident that has involved a self-driving car has been determined to be the fault of another human driver.)

There are two primary reasons for this.  First, self-driving cars have comprehensive packages of sensors that give them a highly-detailed knowledge of their surroundings, to a degree much higher than that of the average human driver.  Thanks to modern sensors and computers, self-driving cars see more, see it better, see it sooner, and are more completely aware of their surroundings than a human driver is capable of.  They are also incapable of being distracted or losing focus, as human drivers regularly do.  If there is a danger, it is detected by the car well before a human driver would normally see a danger coming.

Second, the power of the computer means that self-driving cars can react to a danger in milliseconds… or a few thousand times faster than it normally takes a human driver to see a scenario, have it register on their brains as a threatening situation, and react physically.  And as a human’s quickest reaction is often a panicked one, regularly making bad impulse decisions in a crisis, the self-driving car is much more likely to make a good decision than a human driver.

In short: Self-driving cars will be better than you.  Which leads to the true, realistic solution of the self-driving car in the scenario posed above:

scenario_DThe self-driving car, having identified the danger far sooner than any human driver, has more than enough time to formulate a safe reaction without panicking, executes that reaction perfectly and stops, without causing an accident or hitting anyone. Duh.

And sure, there will be the incredibly rare situation when a car will not be able to react and avoid killing someone.  But even without bringing morality into the situation, the car will be better able to react and cause a minimal loss of life than any human driver could do in the same situation.  This is the reality of the self-driving car and its capabilities: Even in extreme scenarios, the self-driving car will be far safer than a car in the hands of the vast majority of flawed, distracted, slow, impaired and panicky human drivers.  Period.

And given this undeniable reality, it continues to amaze me when I hear people insisting that they can drive better than self-driving cars; people actually believe they can see sooner and react better and more effectively than computers, and our roads would be safer without computer-controlled cars on the road.  Without mincing words, it’s just plain delusional.  We will all be safer when computers do our driving for us; and the sooner we accept that, and do everything we can to accelerate the process of putting us all in self-driving cars, the better off we’ll all be.