Re: Automate the Pedestrians.
I think that one thing which has been overlooked by all this is continual analysis of possibilities. A self-driving car could be programmed such that it is always looking for escape routes from potential accidents. Think chess: computers are good (when programmed correctly) at multi-step thinking. At the point of the child stepping out into the road, they could have analysed 1000's of possibilities and 1000's of possible reactions, already calculated the appropriate action and be ready to respond instantly should that situation occur.
I know that good drivers do this already, although they do so subconsciously for the most part, just as the best chess players think many moves ahead. However, the car has been programmed to do so, and doesn't stop doing so because they are tired, or have had an argument with their wife, or are trying to solve that problem they were stumped by at work. The car is also more likely to have spotted that kid before he runs out into the road, analysed it's actions and prepared appropriate reactions.
How to respond comes down to the software writers. In a purely logical view, they need to minimise the damage done. A child's life would have a value, as would the life of the occupants of the vehicles, and the lowest cost action available would be implemented. The real question is how things would be weighted in the algorithm. That's the hard part, with interesting and conflicting moral and social dilemmas involved.
Biting the hand that feeds IT
