Makes sense, if both party's lives are in equal danger. You don't want your car to put you in danger.
However, under many speed conditions, head on collision into a wall, or even a tree, will only nuke the car and leave the occupants unscathed. Much lower speed conditions will result in dead pedestrians. Then I'd vote for prioritizing pedestrian survival if the driver is at low risk.
Real "10 dead pedestrians or 1 dead driver" cases? Textbook ethics decision cases? Probably pretty unlikely compared to generic autopilot fails with worse overalll human life outcomes. Or the daily meatgrinder of human-triggered casualties. Ivory tower thinking.
Not discounting ethics but these seem to be more what-if stuff than real world issues. Once the basic autopilot tech is reliable enough in a more general sense.
I find the economics & ethics of what is a human life worth when it is a std manufacturer defect vs when it is an autopilot fail, if the autopilot results in less overall deaths much more interesting. Making mistakes is tremendously costly for car makers ( ex: Prius) . Will we hold autopilots to those higher standards, compared to human drivers, if their imperfections still kill less than human drivers whose liability is capped by car insurance companies? Will we instead say, it's Ok, Mercedes, that's another $200K because thats the max liability a human would have incurred, even though a particular recurring fail might have been corporate incompetence? Who will be insured by whom?
Airplane makers have particular international agreements that caps quite stingily. But their overall safety record is good, because of the root cause fault finding efforts deployed on plane crashes.