Are we holding autonomous vehicles to impossibly high standards?

[diplomat33]

Found this thought provoking video by the youtube channel Transport Evolved:

I feel like the video makes a good point. As long as we have a combination of some autonomous cars and some cars being driven by (bad) human drivers, accidents are going to happen. Expecting autonomous cars to be infallible is not realistic. And if we say that we will only accept fully autonomous vehicles when they are infallible, we could be waiting a very long time. So the question might be: if we expect that autonomous cars will be much safer than human drivers but won't be 100% perfect, what level of fallibility would still be acceptable?

 

[Daniel in SD]

I think they just need to reduce accidents relative to a human driver. Note that I didn’t say safer because I can imagine a situation where autonomous vehicle get in fewer at fault accidents than humans but way more accidents total. For example the autonomous vehicles being tested in California seem to be getting rear ended all the time.

 

[diplomat33]

I think they just need to reduce accidents relative to a human driver. Note that I didn’t say safer because I can imagine a situation where autonomous vehicle get in fewer at fault accidents than humans but way more accidents total. For example the autonomous vehicles being tested in California seem to be getting rear ended all the time.

Yeah, I do think we need to make a distinction between accidents where FSD is at fault and accidents where it is not at fault. If the FSD car correctly stops a red light for example and some idiot human driver rears end the FSD car because they are distracted on their phone, that's not something that the FSD car could ever prevent. We should not hold the FSD car responsible for situations like that. The FSD car should only be held responsible for situations where it makes a mistake and directly causes an accident.

 

[ewoodrick]

Probably over 1,000 people killed in Georgia since the first of the year.

Can autonomous vehicles do much worse?

 

[willow_hiller]

Yeah, I do think we need to make a distinction between accidents where FSD is at fault and accidents where it is not at fault. If the FSD car correctly stops a red light for example and some idiot human driver rears end the FSD car because they are distracted on their phone, that's not something that the FSD car could ever prevent. We should not hold the FSD car responsible for situations like that. The FSD car should only be held responsible for situations where it makes a mistake and directly causes an accident.

One shouldn't ignore accidents where the autonomous car isn't at fault, though. An unusually high number of not-at-fault accidents could indicate the vehicle isn't behaving in a way predictable to human drivers (e.g. rapid, unexpected braking).

But of course, that will only matter while there are still humans driving.