Quote:
Originally Posted by PokerHero77
Identifying hazards real-time (e.g. driving in fog, mountain driving, flood zones, etc.) is a lot easier said than done. And what to do once a hazard is sensed is another matter as well.
Once human only driving is essentially outlawed for the majority of drivers, then it will be very easy to regulate driving in a way that precludes a lot of these scenarios. For example, mountain route A is no longer navigable unless you live there, which means you would be able to earn a class 2 license allowing for human driving on mountain roads. Bad weather would basically shut down all AI autonomous driving until the weather is safe enough to proceed. More people unable to get to work, more people unable to get from point A to B, more people restricted from operating a vehicle, etc.
Some hazards yes (such as driving in blizzard) other hazards no (such as stalled car in middle lane of road).
The danger of a stalled car in the middle of the highway, unable to get to the shoulder often causes the biggest pile up and death tolls as the first cars with good visibility see it and swerve but ones after start slamming in to one another.
Networked AI would deal with that type of hazard much more efficiently even if the first car was unable to stop due to insufficient distance at recognition, but every car there after should easily avoid it.
Whatever the reason, every time I am forced to rapidly decelerate on a major highway, i am equally looking to see what the hazard is ahead as to why others are rapidly decelerating while anxiously checking my rear view to try and see if anyone is flying up behind oblivious. I am also looking if I have any options to escape should that rear hazard present. I am far more concerned about the person behind me inattention, than to my ability to deal with any hazard that exists in front of me.
But I agree with you that situations would present that AI would just shut down all vehicles and not operate until the situations improved.
For instance I am not sure that anything short of full sentient AI would be able to deal with the types of Canadian winter driving (winter squalls, black ice, fierce winds) as well as the better Cdn drivers learn to do thru experience. Dealing with squall like conditions and roads going from changing patches of exposed black ice thanks to winds, and then to better grip but snow covered constantly requires driving a lot by 'feel'.
One of the worst stretches of road in Canada for that is the highway that connects Calgary to Edmonton as you are going thru the Red Deer corridor. Wide open prairie fields on both sides of the highway that create fierce wind squalls and blinding snow, and ever shifting black ice patches that are exposed and then covered up and then exposed again as the snow shifts like the sands of a Dune.
I had to drive it each and every time, often several times a winter season, as none of my business partners felt comfortable enough driving in those conditions. The entire 3 hour drive, in good weather, would take 5 hours plus in storms like this, with cars littering the ditch the entire way. I actually really enjoy driving in the most challenging conditions (feeds my adrenaline junky brain) and I do so with safety being my highest priority as for me its a game of winning by beating the weather conditions and not letting them win, so i do everything to win and always have. No accidents yet despite more hours driving in those conditions than most people on this planet who are not professional drivers.
i struggle to see how all but the most advanced fully sentient AI would be able to navigate that and think it would just check out. I am sure chez would say 'and that is the right decision', to which I would not disagree for those extreme instances.