Very interesting article that covers a lot of my own software and engineering opinion on the matter.
I disagree that flipping the target to manufacturers will increase safety as you describe. If a corporation is liable for AI safety, then as with Wall Street, manufacturers will view penalties for people dying as a cost of doing business.
Billionaires like Musk are leading EV car companies. They will bribe those setting the penalities. The resulting penalties will be low. Autonomous driving will not improve much, because there is little penalty for not improving. Again, just like our fnancial markets.
Unless humans can go to jail for making bad AI decisions, AI decisions will continue to be bad. Case in point: 38 of NHTSA's 47 autonomous driving crashes were in Teslas.
Musk finally admitted that self-driving is more difficult than the thought. No shit, Sherlock.
Driving aids like lane detection will continue to improve, but will only ever be aids until the brain of an experienced and technically skilled driver can be duplicated electronically. That day is a long way off at least, and may never be possible.
