You would need more than one rate:
- percentage of product owner operators who are killed by the product they own/operate per year
- number of other people killed by the median product owner per year
- inflation adjusted property damage (belonging to other people, or to the public/govt) caused by the median product owner per year
Regulating products based on the potential to kill, maim, or injure, is not a terrible idea.
It’s why we require more training of people who fly 747, then people who operate cars.
But if it was going to work, we’d have to do it without carve outs - if it only applies to some products, then it’s really just politics.
If Tesla’s auto pilot is really so safe that it needs little or no regulation, then by definition, regular cars are so dangerous that they should be banned or require much more regulation. But I only ever hear the first half of this argument, which makes me worry this is not really an argument about safety.
I keep using this as an example - the Therac mechines for radiotherapy undoubtedly saved lives. They also undoubtedly administered radiation treatment better, faster and more accurately than any manual operator could have done.
And yet, they all got recalled when we realized they "sometimes" administer a lethal dose of radiation by mistake. Or do you think they should have continued operating? What was the "acceptable maximum rate at which your products can kill people" for them? Because I'd argue it's zero. And it should be zero for Teslas or any cars that have something called "autopilot".
A surgeon who performed a voluntary operation which caused unpredictable complications leading to death shouldn't necessarily stop operating on other patients. There's a line that has to be drawn somewhere, I'm not going to draw it.
A self-driving car that kills less people per mile than a reasonably selected cohort of human drivers is probably a good thing.
Replace the surgeon with a robotic surgeon operating under some kind of autonomous mode and yes, I think every robot of its kind should be immediately pulled out of use.
>>A self-driving car that kills less people per mile than a reasonably selected cohort of human drivers is probably a good thing.
Hard disagree, and I honestly hate it when people make that argument. The number should be zero.
>>Or are you suggesting that we shouldn't use a technology that is vastly safer than human drivers, but still causes a nonzero amount of deaths?
Absolutely this one. And the key word here is "causes" - if the deaths are being caused by mistakes of the algorithm(and by extension - their creators) then every single one of these systems should be disabled and pulled out from sale until it can be addressed, in the same way a plane autopilot would be. I suspect we will disagree on this.
Perhaps look at it this way - when I buy an automatic pressure cooker, I need it to explode exactly zero times, not "less than manual pressure cookers". If my car drives into a concrete barrier because it thought it's actually a perfectly straight road g - I really really don't care that on average fewer people have died while using it than when driving themselves. It's unsafe and it should be forbidden from sale and use on public roads.
