[imtringued]

This is just stupid. An accident is an accident and that is all there is to it, there are no ethics involved in a malfunctioning piece of equipment.

People hate ethics because of stupid crap like this. Because people have a hammer (trolley problem) and every thing turns into a nail but the point of the trolley problem isn't that there are trolleys and victims. You are falling into the trap of trying to make a thought experiment as real as possible, which is completely useless in practice. Comments like this give ethics a bad image and turn it into a sad joke.

Here is a real ethical dilemma: Early adoption of autonomous vehicles will cause deadly accidents during the development phase but the deployment of autonomous vehicles will save lives over the long term.

Should we accelerate the development of autonomous vehicles and accept early deaths in exchange for saving more people by deploying life saving autonomous driving earlier?

Should we delay the development of autonomous vehicles by two decades to avoid early deaths while letting people drive themselves and suffer an increased accident rate for 20 years that will ultimately cost more lives than the development of autonomous vehicles?

Tesla is following the first option. They are deploying technology that isn't ready to speed up the rate of progress.

Waymo is following the second option. Stay safe during development, even if it takes much longer for your product to actually decrease the fatality rate.

[_carbyau_]

I am not an ethicist expert. As many people are not in a great many things. But it doesn't mean I don't care.

My point is that there is more to an accident. And hell, there very much is ethics involved in a piece of malfunctioning equipment. Ask any air crash investigator.

No one blames a human, when faced with making millisecond decisions, for not doing much. We're simply not very capable in tiny timeframes.

But will people accept the same standard for a crafted program (through whatever means) who's very promise relies on constant vigilance and sub-millisecond decisions? And what are the decisions to be?

I like your wider perspective take too. "Move fast, break things" vs "Stay safe" is a development timeframe vs cost trade off.