A sign of things to come
Tesla's "autopilot" is actually quite modest and it's easy to see how you might break the problem down and model it - multiple lanes of cars all going the same way, sensors that model the car's surroundings / lane markings, algorithms that maintain speed & distance, algorithms that mark opportunities to overtake, algorithms to avoid / brake hazards based on proximity, control steering / brakes / lights. It's complex no doubt but it can be modeled.
But it requires:
a) The computer is able to see all hazards, act in a predictable way and additionally only engage when the road and conditions are suitable. This is clearly not the case.
b) The car forces the driver's attention. Force the driver to hold the wheel with both hands. Force them to touch a pedal in a certain way. Monitor their head and posture. This is clearly not the case either.
It is the failure of a) and b) which causes accidents. A failure of a) is bad enough but without an attentive human, it's a guaranteed accident. This is a forseeable consequence of not forcing attention, i.e. bad design.
The funny part is Tesla's self drive solution is quite modest. The problems facing mostly or even fully automated cars are multiple factors higher. Perhaps reports of accidents might do something to allow a little bit of reality to creep into the hype about self drive vehicles.
Biting the hand that feeds IT
