Comparing appropriate statistics

Crash statistics for human-driven cars are compiled from all sorts of driving situations, and on all types of roads. This includes people driving through pouring rain, on dirt roads and climbing steep slopes in the snow. However, much of the data on self-driving cars’ safety comes from Western states of the U.S., often in good weather. Large amounts of the data have been recorded on unidirectional, multi-lane highways, where the most important tasks are staying in the car’s own lane and not getting too close to the vehicle ahead.

Automated cars are rather good at those kinds of tasks – but then again, so are humans. The data on fully automated systems will naturally expand to cover more roads as states allow automated vehicles to operate more widely. But it will take some time before self-driving cars can cover as many miles in a year and in as many circumstances as human drivers presently do.
Autonomous cars are good at driving on clear, open highways in good weather – but so are people.

It is true that self-driving cars don’t get tired, angry, frustrated or drunk. But neither can they yet react to uncertain and ambiguous situations with the same skill or anticipation of an attentive human driver, which suggests that perhaps the two still need to work together. Nor do purely automated vehicles possess the foresight to avoid potential peril: They largely drive from moment to moment, rather than thinking ahead to possible events literally down the road.

To an automated vision system, a bus shelter full of people might appear quite similar to an uninhabited corn field. Indeed, deciding what action to take in an emergency is difficult for humans, but drivers have sacrificed themselves for the greater good of others. An automated system’s limited understanding of the world means it will almost never evaluate a situation the same way a human would. And machines can’t be specifically programmed in advance to handle every imaginable set of events.