I get that impression from the stats. They say the autopilot is ok if you keep an eye on it but in the real world people won't always - I think the guy who died was watching a movie.

On the other hand about 1 million people die each year in driving accidents so if by winging it a bit you can stop that a year earlier it's a million lives saved vs maybe <10 autonomous deaths in the research phase.