barrydgold@ca.rr.com
Date: Tue, 2 Oct 2018 23:41:06 -0700
On 10/2/2018 4:33 PM, RISKS List Owner wrote:
> CB (Carbon-based) 100VMT for 2016: 1.2 (~270M registered vehicles)*
> SB (Silicon-based) 100VMT for 2016: 3 (~100 registered vehicles)^
>
> This hypothetical statistic demonstrates a safety disadvantage for AVs. Not
> a likely selling point for consumers currently. Also, the AV sample size is
> at least 4 orders of magnitude smaller than the CB population.
That's for all SB vehicles. What happens if we segregate the statistics by
manufacturer? Are *any* of those fatalities due to Waymo vehicles? I haven't
tried a full statistical analysis, but what I remember from newspaper
reports is that at least 2 of those 3 are from Uber's SB Vehicles, not from
Waymo. This would _tend to_ suggest that Waymo is more careful than Uber in
designing their hardware/software systems to avoid accidents. And in testing
their vehicles: Waymo had SB vehicles "driving" around with a human (CB) in
the car to take over in case of emergency.
Unfortunately, given the numbers in Stein's posting, that would leave a sample size way too small. At least if we insist on measuring only fatalities.
I remember this arising a couple of years ago when I was reading analyses in
the newspapers, that said that fatal accidents were so rare (1.2 per million
VMT per the NHTSA figures Stein quotes) that it would take a long time to
accumulate enough VMTs on SB vehicles to know if they were "safe enough". I
was skeptical then, and I remain skeptical. You don't have to count only
fatalities, because fatalities correlate positively* with (1) non-fatal
injuries, and (2) non-injury collisions.
So if we look at injury accidents per 1E8 VMT for SB vs. CB "drivers", we should get a pretty good idea of whether SB "drivers" are better or worse than CB drivers.
* Not necessarily proportionally, although I suspect that fatalities are
proportional to non-fatal injuries.