Quote:
Originally Posted by sirdaft1
Another interesting thought... how are the software engineers supposed to program the logic into the AV's in the case of an inevitable crash? For instance, what should it be programmed to do when a semi-truck swerves into the AV's lane and the choice is either: head-on crash with the semi, swerve into oncoming traffic and high chance of head-on crash, or swerve to the right where there are people walking on the sidewalk? Do they place a "life value" on the individuals involved in the scenario and take this into account in the logic? (ie: mother pushing baby in stroller has higher life value than convicted felon driving in AV???)
It may sound like is a highly unlikely event, but it's worst case scenarios like these that someone will have to program the logic on. And the more you think about it, the more it gets really messy.
|
That is my exact point I was making above without getting into the details like you did. A human making a mistake and possibly hurting or killing someone in a car accident can be forgiven in most cases, a machine making the same mistake is not a forgivable action.
The programing has to follow the same logic the person would do if they were driving which is, most everyone will put higher value on their own life. If you are not the cause of an accident (meaning you did nothing wrong) most likely you going to do everything can to prevent your death even if it means hitting someone else who was in fact the one at fault and killing them.
What most people do not understand is the fact a self driving cars is operating under a set of predictable control systems with known feedback mechanism and sets of rules, and they just placed it in a chaotic system with no rules or predicable behaviors with humans and asked them both to interact with one another in a predictable manor.