Uber Halts Self-Driving Car Tests in Arizona After Friday Night Collision (businessinsider.com)
"Given that the Uber vehicle has flipped onto its side it looks to be a high speed crash," writes TechCrunch, though Business Insider reports that no one was seriously injured. An anonymous reader quotes their report:
A self-driving Uber car was involved in an accident on Friday night in Tempe, Arizona, in one of the most serious incidents to date involving the growing fleet of autonomous vehicles being tested on U.S. roads... Uber has halted its self-driving-car pilot in Arizona and is investigating what caused the incident... A Tempe police spokesperson told Bloomberg that the Uber was not at fault in the accident and was hit by another car which failed to yield. Still, the collision will likely to turn up the temperature on the heated debate about the safety of self-driving cars.
Conversations would be different if the uber car was at fault but not all accidents can be avoided.
I am reminded that when cars were first invented, there were laws put in place mandating that someone walk ahead of any self-propelled vehicle waving a red flag, for fear of scaring horses and making people uncomfortable.
I'm sure that in one hundred years this sort of reaction - blaming the software for an inattentive driver failing to yield - will be seen in exactly the same way.
The human made a mistake yes, but the self-driving car crashed into him. So now the question is whether a human would have done better in that situation.
It's a given there will be some instances where a human driver might have done better than a self driving auto. In the same vein, the possibility also exists that the human driver may have done worse in the identical situation.
If driver-less autos can perform appreciably better than humans do over a large enough sample size, they should then be considered a safe alternative... the only question is how much better they need to perform.
Happiness in intelligent people is the rarest thing I know.
Ernest Hemingway
But. If we concede that any human life = any other human life, and the widespread use of driver-less vehicles saves X accidents and Y highway fatalities over the same number of driven miles, it has saved more accidents and human lives than it lost.
If we set the bar at ZERO accidents a human could've avoided, well, that is an impossibly high standard; and self-driving vehicles should be shelved right now.
Happiness in intelligent people is the rarest thing I know.
Ernest Hemingway
I concede that companies should not profit from products that might kill people. That is all that I concede.
You might as well just say that companies shouldn't be allowed to sell anything. Read the safety labels on anything you buy these days.
How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
"Compatible", WTF? The human was at fault, the human was the incompatible piece in the situation. We can choose to go for "years and years" with this insane mixture and almost certainly will, but that doesn't change the fact that the human was at fault. Make that vehicle driverless and the accident never happened.
Well, now you're making a very different argument than the original "companies should not profit from products that might kill people." But I'll bite anyway. There are plenty of products that, though used correctly, can under some circumstances cause injury or death.
A very obvious one is medication. There are many medications that can have serious side effects, including death, when taken exactly as prescribed. We continue to use them because the benefits outweigh the risks.
You mentioned chainsaws. It is true that the majority of chainsaw accidents happen because of operator error. However, that doesn't mean that all of them do. The only way to completely eliminate the possibility of harm is to not use a chainsaw. But again, we continue to use them because the benefits are big enough.
There does need to be a standard for how safe autonomous vehicles need to be before we allow them on the roads. But setting that standard at "they need to never cause a death" is not only unrealistic, it is totally inconsistent with how our society deals with other potentially dangerous products.
How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
Most people try to pin the blame for an accident on a single cause. Most liability laws are based on this same (erroneous) concept.
Airline accident investigations are really good at demonstrating how an entire chain of events led up to the accident. And that any single factor happening differently could've prevented the accident. e.g. The Concorde crash was caused by (1) debris on the runway from a faulty repair on a previous plane, (2) failure of the Concorde's tires when it struck the debris, (3) failure of the undercarriage to withstand tire debris striking it from a blowout at take-off speed, (4) the manufacturer not making any procedures or provisions to recover from a double engine failure on a single side because it was considered so unlikely. Any one of these things doesn't happen and the Concorde doesn't crash.
Safety systems layer multiple accident-avoidance measures on top of each other. This redundancy means that only when all of those measures fail is there an accident. Consequently, even if the self-driving car was not legally at fault, that it was involved in an accident still points to a possible problem. e.g. If I'm approaching an intersection and I have a green light, I don't just blindly pass through because the law says I have the right of way. I take a quick glance to the left and right to make sure nobody is going to run their red light, or that there aren't emergency vehicles approaching which might run the red light, or that there's nobody in the crosswalk parallel to me who might suddenly enter into my lane (cyclist falls over, dog or child runs out of crosswalk, etc).
So even if the autonomous car wasn't legally at fault, that's not the same thing as saying it did nothing wrong. There may still be lessons to learn, safety systems which were supposed to work but didn't, ways to improve the autonomous car to prevent similar accidents in the future.
Automated cars just need to become as good as an average human who is paying attention, awake, and sober. That will make them better than 80% of the cars on the road. Good enough for mass use at that point.
Don't let the perfect be the enemy of the good.
Socialism: a lie told by totalitarians and believed by fools.