It is fairly clear (at least to me) that driverless cars should make no "moral" decisions at all. They should attempt to avoid accidents if possible most likely

It is fairly clear (at least to me) that driverless cars should make no "moral" decisions at all. They should attempt to avoid accidents if possible most likely restricted to emergency braking. Avoiding one accident in favor of a different accident is not attempting to avoid an accident. AI should not be involved or in any way evaluate the better of two accidents.