Originally Posted by
genec
On the other hand, we now seem to demand that AVs be 100% error free... not just slightly better than human drivers or half as bad as human drivers... (like say 20,000 deaths a year), but no, AVs have to be nearly perfect. 99% perfect would still equate to 400 deaths a year. Will that be good enough? I doubt it. And certainly 90% better than humans would be 4000 deaths a year... no way that would be considered acceptable. But hey, 40,000 deaths by errant humans is "just an accident," right?
Congratulations, I think that may be the most strawman arguments and made-up statistics I have ever seen in a small paragraph. I'm pretty sure no one is satisfied with 40,000 deaths as being acceptable, we just right now don't accept that AI control is capable of reducing that number. You're acting like there's overwhelming evidence for that acceptance of AI control, and there just isn't.
We haven't gotten to the point where AV have demonstrated basic competence, let alone improving safety, and the issue for us as cyclists is that we will be dealing with a mix of AV and driver/operator controlled. I'm pretty sure none of us are intending to ride our AI controlled bicycles. (I'm also pretty sure that for the foreseeable future, the vast majority of motor vehicles are going to be driver controlled, so your "instant reduction" scenario is absurd on its face). My understanding is that AI is having trouble identifying cyclists and also generally anticipating the likely actions of non-AI operators--we don't function according to computer-generated algorithms.