Sensors that prevent running into pedestrians are already being deployed. Collision avoidance is getting better and better.
Not driving into walls should be the next pri0 feature cars implement, but aside from that I see little actionable improvements being proposed here.
There were three concrete and actionable proposals at the end:
Ban all active lane keeping
Mandate geofencing of all series automation
Mandate Driver Monitoring Systems, to include both camera and capacitive touch
I still think it is silly to discuss the use of te term "autopilot" as the key mistake here rather than the choice to produce cars with auto-steering functionality implemented as series automation rather than parallel.
The incentives around Autonomous Emergency Braking is to avoid false positive at all costs because unnecessarily braking on a highway can easily injure or kill.
Avoiding false negatives is somewhat less important.
that's kind of funny, given that its primary purpose is a secondary actor on the human's false negative