This is a really difficult situation, in that lowering regulatory overhead would mean getting automated solutions to the public faster, and even if they weren't optimally safe they'd likely be vastly better than human drivers and the sooner we could reap that benefit the better. However, there is a substantial long-term risk to the viability of self-driving cars if any high-profile accidents do happen, because outside of Silicon Valley people generally trust themselves more than technology. So on the other hand it might be better to continue the dangerous status quo for longer, until regulatory framework can assure safety, so as to make the rollout of autonomous cars actually result in adoption of autonomous cars.