Self driving car companies either want to reduce all accidents or they don't. They either want to integrate with human driving or they don't. Humans occasionally do stupid things in traffic, we can all agree on this. If self driving cars are going to reduce accidents, they are going to have to compensate for the stupid things that humans do, just like we all do every time we go out on the road. I believe the worst outcome is coming true, self driving car companies are not feeling any responsibility to the others on the road. They would rather ignore any accident that is "the human's fault". They would rather behave unpredictably and even though a rational person sees from a statistical level, the more unpredictability they add into the system the more accidents there will be, they seem happy as long as the accidents are "not their fault". Well this is just not an industry that is demonstrating a willingness to be safer overall. So stop saying this is about making driving safer, it's not. It's about making the most money in the soonest time possible and we need to demand better for public roads.
Furthermore, defensive driving is a lot harder for these companies because defensive driving is very much about what they other human driver will do. So these companies are focusing on the much lower bar of eliminating their liability, which leaves humans to pick up the slack.
A self driving car should be better at recognizing if someone is tailgating them. It should be able to measure the distance to the car behind them and adjust driving accordingly, or pull over and let the person pass. But the sad truth is, autonomous driving companies are too focused on eliminating their own liability and less focused on reducing accidents altogether. Because reducing accidents altogether requires a lot of human psychology and they have no idea how to deal with that.
The fact of the matter is, there is a lot of psychology at play when driving. Automated car companies are going to find out very quickly that human driving will be disrupted if human psychology is disrupted. Putting it another way, automated cars are going to have to understand a great deal of human psychology if they want to drive with humans. I have been saying this from the beginning.
Go out and drive at a regular speed on a busy highway then slam on your breaks. Do this repeatedly. If it's not your fault, then it shouldn't be a problem, ever.
Even if a person leaves enough room, it is distracting and stressful to them when that room becomes suddenly less. This will make driving harder for them, which will make them more likely to get into an accident somewhere. Don't ignore human psychology.
It doesn't really matter until they can handle all conditions a human can. It's never going to be an accurate comparison until then anyway. Use Autopilot as a comparison if you want, which does mostly only work in the conditions you describe. Take all the automated solutions and combine them together if you want, they're still not even close.
You have never been on a road with ice ruts. On a road with ice ruts, it no longer matters where the lane was before the ruts were there, because missing the ruts is dangerous. Ice ruts can throw the car sideways at walking speed if you don't align the car with them. Furthermore, they won't line up with the lanes around corners either, so any car that is driving by map had better be willing to forgo that map and drive in the ruts or there will be enormous problems.
So they are working with a flawed implementation from the start, unless they have 1 google car for every construction zone in the world. It is not the job of construction crews or cities to be making 3D maps every time there is a change.
No that's not true. If the robot car was doing something that a human never would have, I blame the robot car; because technology companies should be striving to make them act human so that other humans don't get confused. But unless the robot car was doing something unusual in this case, it is truly the human's fault.
Laws were meant to decide who is at fault between two human drivers. They were never meant to decide the morality of introducing an unpredictable robot car into a human driving environment and expecting humans to 'just deal with it'.
There's a lot of practical little heuristics like this that humans use. Bots could also, but it may take years to include and tune them.
And just when they are done tuning for every situation involving another car on the road, there will be another accident with a car that was for some reason half on a sidewalk, and then they will need to start over again. There are so many possible edge cases I cannot fathom how it could ever work.
No they aren't. Put aside the fact that humans drive way more safe miles than unattended robot cars, the robot cars don't even drive in near enough conditions to make that conclusion.
And yes I realize I am going to a special place in Slashdot hell for this; but I am tired of seeing technology companies give humans the short end of the stick with this.
If robot car makers want to test in public roads, they should bear the burden of proving that any accident would also have occurred if a human was at the wheel. This means, if a robot car slams on the brakes because it is a robot, they should take ten humans and put them in the same situation and see how many slam on the brakes just as suddenly. If no one slams on the brakes, then the autonomous car company should be responsible. In the end it will be good for everyone, because robot cars are never going to be successful if they can't integrate with humans anyway.
And as in most problems in life, the final edge cases take the most time to resolve. Driving will take almost 100% accuracy to be realistic. It isn't about being as good or better than humans, it is just because of the serious consequences that occur when a 4500 pound machine makes a mistake in traffic.
Here are some statistics:
Humans drive 3.22 trillion miles in the US per year, and get into 16 million accidents. 30% of accidents are unreported, so there are really 20.8 million accidents.
That makes 154808 miles driven successfully in all conditions and in all places cars can drive, and includes both parties in the accident whether it is their fault or not. What is the best Waymo can do? 3700 miles in specific test situations?
Slashdot Mirror
Self driving car companies either want to reduce all accidents or they don't. They either want to integrate with human driving or they don't. Humans occasionally do stupid things in traffic, we can all agree on this. If self driving cars are going to reduce accidents, they are going to have to compensate for the stupid things that humans do, just like we all do every time we go out on the road. I believe the worst outcome is coming true, self driving car companies are not feeling any responsibility to the others on the road. They would rather ignore any accident that is "the human's fault". They would rather behave unpredictably and even though a rational person sees from a statistical level, the more unpredictability they add into the system the more accidents there will be, they seem happy as long as the accidents are "not their fault". Well this is just not an industry that is demonstrating a willingness to be safer overall. So stop saying this is about making driving safer, it's not. It's about making the most money in the soonest time possible and we need to demand better for public roads.
Furthermore, defensive driving is a lot harder for these companies because defensive driving is very much about what they other human driver will do. So these companies are focusing on the much lower bar of eliminating their liability, which leaves humans to pick up the slack.
A self driving car should be better at recognizing if someone is tailgating them. It should be able to measure the distance to the car behind them and adjust driving accordingly, or pull over and let the person pass. But the sad truth is, autonomous driving companies are too focused on eliminating their own liability and less focused on reducing accidents altogether. Because reducing accidents altogether requires a lot of human psychology and they have no idea how to deal with that.
In fact, just start slamming on your breaks every place you would normally apply them slowly. See how little of a problem it is.
The fact of the matter is, there is a lot of psychology at play when driving. Automated car companies are going to find out very quickly that human driving will be disrupted if human psychology is disrupted. Putting it another way, automated cars are going to have to understand a great deal of human psychology if they want to drive with humans. I have been saying this from the beginning.
Go out and drive at a regular speed on a busy highway then slam on your breaks. Do this repeatedly. If it's not your fault, then it shouldn't be a problem, ever.
Let's let millions of autonomous cars on the road right now without safety drivers and you'll see how good 40,000 really is.
Even if a person leaves enough room, it is distracting and stressful to them when that room becomes suddenly less. This will make driving harder for them, which will make them more likely to get into an accident somewhere. Don't ignore human psychology.
It doesn't really matter until they can handle all conditions a human can. It's never going to be an accurate comparison until then anyway. Use Autopilot as a comparison if you want, which does mostly only work in the conditions you describe. Take all the automated solutions and combine them together if you want, they're still not even close.
But if it chooses to drive in the ruts at a point before humans are, then it will miss the lane and hit another car.
You have never been on a road with ice ruts. On a road with ice ruts, it no longer matters where the lane was before the ruts were there, because missing the ruts is dangerous. Ice ruts can throw the car sideways at walking speed if you don't align the car with them. Furthermore, they won't line up with the lanes around corners either, so any car that is driving by map had better be willing to forgo that map and drive in the ruts or there will be enormous problems.
Humans are already 99.9% accurate.
And if the signs are just as snow covered as the road?
So they are working with a flawed implementation from the start, unless they have 1 google car for every construction zone in the world. It is not the job of construction crews or cities to be making 3D maps every time there is a change.
That, or get confused about what lane it is in and merge in front of a bus going faster than it, as a Waymo car did last year.
No that's not true. If the robot car was doing something that a human never would have, I blame the robot car; because technology companies should be striving to make them act human so that other humans don't get confused. But unless the robot car was doing something unusual in this case, it is truly the human's fault.
Sometimes human drivers are just responsible.
Laws were meant to decide who is at fault between two human drivers. They were never meant to decide the morality of introducing an unpredictable robot car into a human driving environment and expecting humans to 'just deal with it'.
There's a lot of practical little heuristics like this that humans use. Bots could also, but it may take years to include and tune them.
And just when they are done tuning for every situation involving another car on the road, there will be another accident with a car that was for some reason half on a sidewalk, and then they will need to start over again. There are so many possible edge cases I cannot fathom how it could ever work.
No they aren't. Put aside the fact that humans drive way more safe miles than unattended robot cars, the robot cars don't even drive in near enough conditions to make that conclusion.
And yes I realize I am going to a special place in Slashdot hell for this; but I am tired of seeing technology companies give humans the short end of the stick with this.
If robot car makers want to test in public roads, they should bear the burden of proving that any accident would also have occurred if a human was at the wheel. This means, if a robot car slams on the brakes because it is a robot, they should take ten humans and put them in the same situation and see how many slam on the brakes just as suddenly. If no one slams on the brakes, then the autonomous car company should be responsible. In the end it will be good for everyone, because robot cars are never going to be successful if they can't integrate with humans anyway.
But we are only out of control in the presence of robot cars. Hmmmmm.
And as in most problems in life, the final edge cases take the most time to resolve. Driving will take almost 100% accuracy to be realistic. It isn't about being as good or better than humans, it is just because of the serious consequences that occur when a 4500 pound machine makes a mistake in traffic.
Here are some statistics: Humans drive 3.22 trillion miles in the US per year, and get into 16 million accidents. 30% of accidents are unreported, so there are really 20.8 million accidents. That makes 154808 miles driven successfully in all conditions and in all places cars can drive, and includes both parties in the accident whether it is their fault or not. What is the best Waymo can do? 3700 miles in specific test situations?