This is tangential to your point, but in many (I'd guess most) parts of this country, that is not true. Unless you are under the influence, distracted driver's frequently kill others, and the cops don't press charges, chalking it up to "mistake".
Wasn't it proven that their system recognized the pedestrian and continued driving? That means business logic killed a person. Somebody, somewhere within Uber made the calculation that more miles of uninterrupted data was more valuable to them than public safety. If a jury could be convinced that I, as an individual, recognized a person and chose to purposely continue driving to save myself time, that's manslaughter.
Seems more like murder, of either the depraved heart or plain intentional kind.
So the question here has to be whether the creator of the machine acted reasonably in designing the machine as they did.
[1] This may be a fiction, but one at the heart of how we see each other morally accountable agents, and one that underlies all legal systems.
I'm wary of this line of reasoning. No programmer outside of a terrorist organization will ever intentionally design a machine poorly. If a civil engineer -- specifically a Professional Engineer in the U.S. -- signs off on an unsafe bridge, and that bridge later collapses and kills someone, you can bet your arse they're getting sued and possibly charged with a crime. Unfortunately, we have no real equivalent to this in software engineering/programming -- the concept of a Professional Engineer doesn't exist, and so no one can be held accountable for faulty designs.
Yes, this would bring software development to a grinding halt. I for one don't necessarily consider that a bad thing. It doesn't have to be for every project too -- after all, you don't need a PE to sign off before you build a shed in your backyard. Likewise, a web browser or video game probably ain't gonna need that level of accountability. But for a self-driving car? If someone isn't willing to be held accountable for the software failing, I personally feel the system shouldn't be allowed on the road.
If a self-driving car hurts someone, the people who created and operated it can be hauled into court and forced to justify their decisions. Jurors will hear the testimony of experts in the relevant fields and then decide who should bear the burden of the injury.
No, it's more like one year or less[1]
> I would bet a lot of people hit "something" that often
That's quite a claim... just from personal experience, I'd wager this is a very wrong assumption. Majority of people go most of their lives, if not their entire lives without hitting "something". More serious incidents are even more rare.
My collision insurance is $270 a year, which I think is fairly reasonable for NY. I was involved in an accident a couple years ago where the corner of my car was smashed in a bit, pretty much the definition of a "fender bender". It cost about $3500 to fix, and if I had been at fault, I would have had a $1000 deductible. Repairs could have been $2500 if I was more lucky in exactly what was bent. So, that suggests my insurance company breaks even if I go about 5-10 years between at-fault accidents, or 75,000-150,000 miles. In the aggregate, I assume they do estimate risk pretty accurately, so I expect that 5-10 years is normal. Whenever people mention what they are paying for insurance, it usually seems to be higher than mine.
That's almost surely because of your $1000 deductible. That's very high... mine is $100 for just one reference point.
> So, that suggests my insurance company breaks even if I go about 5-10 years between at-fault accidents
That would be a "break even" point, but insurance companies attempt to turn a profit, so this suggests they expect majority of their customers to not have an event every 5-10 years, but on average a longer interval. They also account for non-traffic accidents, such as vandalism, broken window, theft, etc... none of which involve the driver being "at fault" or hitting something, which was the original assertion by the GP.
The non-collisions aren't relevant because I quoted the price of my collision premium only and not comprehensive or liability. That was on purpose, to be consistent with the topic of "hitting things" and with the sort of fender bender I was talking about.
Profit is not very large relative to premiums, so "breaking even" is a reasonable simplification. For example Progressive (not my insurer, just the 10-K I was able to find the quickest) made about 6%.
Not at fault accidents are some other companies "at fault" accidents, so counting them is double counting and would inflate my estimate.
All I'm saying is that, without false precision, evidence suggests people hit things every 5-10 years on average. Having an accident every year or never in a lifetime is not normal.
Besides, you basically just described most teenage/young adult or inexperienced drivers.