Obscured under " Develop a self-driving capability that is 10X safer than manual via massive fleet learning " is the cold hard truth that that learning will be paid for in lives - mostly of Tesla drivers possibly of others.

The rates of accidents in manual vs current version of autopilot may work out to be favorable - (and that is still under debate) - but there certainly will be people who will die (and have died) due to a premature roll-out of Autopilot and their trust of it. This is some bloody cold calculation