While these are interesting and important questions from a philosophical, legal and generally humanities angle, they are remote edge cases that aren't even very clear. The engineering is so, so far from making those unlikely choices. We'd be lucky if it makes out a white truck against a light, bright sky.
All right, all right, I realize it's going to get past those easy cases, but there are vast, broad, numerous edge cases that need engineering/statistical solutions before we even get to "do I run over the mom with the stroller or do I plow into the bus stop?" kind of made-up scenarios. We have little way of modeling how each target would move or react. Are we going to conduct experiments to see that? It's just a discussion that people find interesting, but it's more of an advanced engineering problem (in fact, it may not be a "well-defined" one, but I won't define that here). We have beginner and intermediate to get through first.
Quote:
Originally Posted by clyde
Have you guys been following the discussions about how cars should be programmed in terms of choosing what/who to hit when collisions are unavoidable?
There is still so much shit to figure out.
Assuming the threshold has been crossed that autonomous cars are good enough that they should be allowed to be used in the market, is it best to push the technology out before ethics and regulations catch up?
Does anyone want to sit on the jury for the first case of a manufacturer being sued when a baby gets hit in a crosswalk and the car chose to mow them down rather than swerve into oncoming traffic and risk the driver?
|