Solving the driverless car problem Part 3

By Terry Freedman.

The problem referred to in the title may be summarised as: who ya gonna kill? The car is hurtling along the road when a child steps out in front of it. The car is faced with a dilemma: kill the child, or the pedestrian waiting to cross the road, who would be in the car's path should it swerve to avoid the child. Terry Freedman explains why this whole thing is a red herring.

Driverless car by Terry Freedman.

Driverless car by Terry Freedman.

Over the last couple of days both Lance Sharpe and I have been discussing this issue. Lance Sharpe believes that Economics and technology can solve the problem. I have argued that Sharpe’s analysis rests on faulty assumptions and hidden biases.

However, this is a non-problem – or should be. Here's why: if you are a good owner you will be continuously scanning the road ahead, behind and on both sides for potential hazards. This is the basis of advanced driving. To an advanced driver, nothing happens "suddenly" (surely the most overused word in road traffic accidents?). The child who runs out into the middle of the road didn't just beam down, as in Star Trek: they came from somewhere!

An advanced driver, seeing a child running along the pavement (sidewalk) sees a child who may at any moment run into the road. A pedestrian staring at a phone while walking presents another hazard. A bus stop suggests the possibility of someone running across the road without looking. A sign declaring the proximity of a school is another trigger.

In fact, every other person and object on the road is evaluated and assessed for its hazard potential. According to statistics I saw some years ago, drivers who took and passed the advanced driving test administered by the Institute of Advanced Motorists (IAM) had 80% fewer accidents than drivers who had not taken a course in advanced driving, Intriguingly, though perhaps not surprising, even drivers who took the test and failed had 50% fewer accidents, than motorists who hadn't taken the course at all.

If human beings can achieve this level of safety, how much more so should artificial intelligence be able to? As Matthew Avery, director of research at Thatcham Research told RoadSmart, the magazine of the IAM:

"The pedestrian/pram scenario is a red herring. As a driver you will probably never face this specific scenario; the frequency of this type of event is rare because you're a safe driver and you can anticipate danger. These cars can be programmed to do the same, but to process and react to situations even faster and more decisively than humans."

Artificial intelligence is a topic frequently covered in the  Digital Education newsletter. Subscribe now!