Anthony Flack wrote: Mon May 15, 2023 4:24 pm

seby wrote: Sun Apr 23, 2023 8:54 pm For example, collision-avoidance protocols in the algorithms of autonomous vehicles are responses to the Trolly Problem (have fun looking this up).

Are they? I always thought this was a bit of a red herring for autonomous vehicles, which can't tell the difference between a doctor, a mugger or a pregnant woman. They're just going to do their best to stop without hitting any obstacle... as opposed to a human driver who would similarly avoid the trolley problem by panicking, throwing the car into a skid and wiping out everybody.

I think the trolley problem is misapplied to vehicles, probably because it's constructed as a problem with a trolley, as opposed to say, a problem of allocating limited medical resources. During the Covid epidemic, the ventilator problem was a far better real-world analogue to the trolley problem than anything involving vehicles.

Anthony Flack wrote: Mon May 15, 2023 5:25 pm Actually we know what the human driver's response is to the trolley problem which is that you instinctively save yourself, which is why the front seat passenger is the most likely to be killed in a car crash. The human driver tends to instinctively plow through the crowded mall rather than into the brick wall.

Would a computer driver take into account who's in your car? Would it even know? Should we program the car to head for the brick wall rather than the crowded mall? Would people submit to being driven around in an autonomous car knowing that it was programmed to do that? Would we end up making cars that drive into brick walls every time a cat runs out in front of it, ie. false positives do more harm than good? What if it's glitchy and unpredictable but still performs better than a human driver? Maybe a random choice would be better. It flips a virtual coin and we all take our chances. Why are we still driving cars?

Here in my car, I feel safest of all...