I think the truth is somewhere between the 100x and 2x numbers. Here's why:
Let's say self-driving does reduce the number of deaths by half. Then there are different classes of people who ride in cars:
1. People who would have died without self-driving, and did in fact die with self-driving.
2. People who would have died without self-driving, and didn't because of self-driving.
If you leave it at that, you might think that of the 40K number you mentioned, with self-driving being 2x as safe we'd have 20K in class 1, and 20K in class 2. But that's not accurate. There is an additional class of people:
3. People who wouldn't have died without self-driving, and who died because of some failure of self-driving.
Since drivers are supposed to be ready to take control at any time, car manufacturers want public perception to be that there is no category 3 -- and with good reason, because:
The problem self-driving car manufacturers have is that people in category 2 are anonymous -- almost invisible. At the end of the year we can look at the totals and say, "yup, there are 20,000 people out there alive because of self-driving." But if the numbers are actually:
1. 19,000
2. 20,000
3. 1,000
Then people aren't going to be thinking about the 20,000 anonymous lives saved by self driving. They're going to be thinking of the 1,000 very obvious and named people in category 3, who would still be alive today were it not for the fact that their car drove them into a semi.
I don't know what ratio of category 2 to 3 is necessary for people to accept self-driving as a benefit, but I'm betting that 2x as safe isn't going to get category 3 low enough to make the case.
Yep -- I think that if self-driving cars prevented all accidents, but each year one self-driving car randomly drove at max speed into the side of a Walmart, people would reject that trade. I think of myself as fairly logical, and I'd have a hard time accepting that. The Trolley Problem is hard.
Let's say self-driving does reduce the number of deaths by half. Then there are different classes of people who ride in cars:
1. People who would have died without self-driving, and did in fact die with self-driving. 2. People who would have died without self-driving, and didn't because of self-driving.
If you leave it at that, you might think that of the 40K number you mentioned, with self-driving being 2x as safe we'd have 20K in class 1, and 20K in class 2. But that's not accurate. There is an additional class of people:
3. People who wouldn't have died without self-driving, and who died because of some failure of self-driving.
Since drivers are supposed to be ready to take control at any time, car manufacturers want public perception to be that there is no category 3 -- and with good reason, because:
The problem self-driving car manufacturers have is that people in category 2 are anonymous -- almost invisible. At the end of the year we can look at the totals and say, "yup, there are 20,000 people out there alive because of self-driving." But if the numbers are actually:
1. 19,000 2. 20,000 3. 1,000
Then people aren't going to be thinking about the 20,000 anonymous lives saved by self driving. They're going to be thinking of the 1,000 very obvious and named people in category 3, who would still be alive today were it not for the fact that their car drove them into a semi.
I don't know what ratio of category 2 to 3 is necessary for people to accept self-driving as a benefit, but I'm betting that 2x as safe isn't going to get category 3 low enough to make the case.