Quote:
Originally Posted by
woodPixel
Ian, good post. It shows I've been subverted by their psyops! You are spot on. I wish I had your clarity of thought.
I've had a bit of practice.
I wrote the following about 18 months ago in response to an article on the programming of driverless vehicles, and how the AI code should respond to a range of scenarios. I think the tone of my last paragraph will give you a flavour of my attitude.
Real people make moral decisions based on gender, race, religion, a person’s attractiveness (good looking female vs fat lady), a person’s social standing and other factors. Anyone who denies that bias affects their decision making is a liar. The best you can hope for is that a person faced with a decision will be aware of their biases and will consciously attempt to negate that bias.
I find that all your scenarios at moralmachine.mit.edu unrealistic as they presuppose that the driverless vehicle will be travelling in an area with high pedestrian activity at a speed that, in the event of a crash into a solid object, will result in the death or very serious injury of the vehicle occupants. Alternatively, a crash into the group of pedestrians will result in the death of many of them. Go read the Swedish Vision Zero. In it you will find guidance that should result in a driverless car’s programming not allowing the vehicle to exceed 30 km/h in an area with significant pedestrian activity. At 30 km/h a vehicle / pedestrian crash is survivable for the pedestrian (and a vehicle impact into a solid object at that speed is 100% survivable for properly restrained vehicle occupants.)
Even reading the Wikipedia entry –
https://en.wikipedia.org/wiki/Vision_Zero – will get you started on a more appropriate path with your vehicle programing.
And the presumption that a pedestrian walking against a DON’T WALK can legitimately be killed is completely morally repugnant and reflects a too legalistic view of road use. People make mistakes, and in some cases ignore instructions or guidance. Those people should not be considered legitimate targets.