I believe he's alluding to the nature of the accidents. They're high intensity events which are more likely to be fatal (speeds were ~70mph). They're not fender benders when the autopilot fails.
The WaPo article addresses that and counts them as AutoPilot engaged crashes. It seems to disengage 1 second before many of the crashes. It is nearly impossible to draw any conclusions about anything from this article. Other than, people use AutoPilot and sometimes they crash using it.
The one interesting bit is where the damage happened on the cars and the hitting of stationary objects. Those seem like unforced autopilot errors.
> But that makes sense statistically, because those are thousands of ostensibly independent events
Not only that, but in the case of a plane or train crash, the passengers were completely powerless to do anything about it. So it's a death that's imposed on them by the event.
In a car, the driver has nearly all the agency to do something about it and not be part of a crash. Car crashes don't just suddenly happen to people on a random distribution.
Many of them are directly the fault of the driver, so you can opt of those (don't drive drunk, sleepy, etc). Many are the fault of other drivers but you could often opt out by paying attention (go watch youtube dashcam car crash videos, it's astonishing the number of crashes where the victim could've avoided it but doesn't do anything). Many are also due to lack of skill (e.g. car slides on an oil slick and driver doesn't know how to control it) which you can opt out by taking car control training. Some are from equipment malfunction which you can also opt out (maintain your car).
> Those "few occasions" would presumably be accidents without the intervention of the safety driver.
I'm not sure that's a safe presumption. The car could have done the right thing but not been confident enough in the planned course of action. Still not great, but not a presumed accident either.
>How does that number compare with humans? Well, regular people in the USA have about 6 million accidents per year reported to the police, which means about once every 500,000 miles.
Aren't accidents only reported to police if there is a hit & run, or some other criminal activity?
Seems like the wrong metric for comparison, given the way they define the self-driving "accidents" and that the majority of human fender benders are not reported.
>There's a ton of things that can go wrong in a car which can cause an accident.
And pretty much none of them ever do if the driver doesn't react exceptionally poorly. Even the spectacular stuff that the internet absolutely loves to hand wring about, like a wheel falling off for whatever reason, almost always results in the car coming to a controlled stop on the side of the road. The conversion rate between "failures" and "meaningful harm to anyone or anything is abysmal."
> Every crash that is not deliberate is an accident.
If you were to define the word "accident" as "everything not deliberate", then, sure ok. But that's not how the word is typically used.
There are vanishingly few real accidents in traffic. Nearly all crashes are also not intentional.
The vast middle is crashes due to driver incompetence or negligence. Not intentional, but totally avoidable if the driver had done their job (pay attention, be lucid, have suitable skills).
People like to call things "accidents" because it shifts blame to destiny. But reality is that nearly all crashes could have been avoided by the driver.
>But the question is what are we optimizing for? The fewest crashes? That's probably the right thing to do given crashes are bad
Another thing to take into account is that not all crashes are equally bad. I wouldn't be surprised if getting rear ended 10 times was safer than a single t-bone or head on collision
Well, my point was that most automobile collisions aren't 'accidents' in the true sense of the word, and most automobile related injuries (and deaths) occur when someone does something negligent. The vast majority of collisions are caused by some combination of being distracted, intoxication, speeding, or just being reckless. Deaths (or serious injuries for that matter) rarely occur at low speeds, it's most often in situations that involve speeding.
There's a lot of data[0][1][2][3] to back up the correlation between speed and deaths, that's the whole focus of vision zero[4].
> they tend to, as you mentioned, be worse accidents.
This reminds me of red light cameras. People love to complain that they only exist to make money because they statistically cause more accidents. And it's true, they do cause an increase in total number of accidents. But they decrease the number of serious accidents. You get more rear end collisions, but fewer T-Bone collisions which tend to be much more serious.
reply