I couldnāt wait any longer for an answer so I found one elsewhere.
I experienced it myself where I was having to intervene every couple miles with v11 at the beginning of the year, and by the end of the year with v13 I was going well over a thousand miles between interventions that were necessary to prevent an accident.
So if you drive 10k miles per year, and 1-2k between interventions to prevent accidents, then only 5-10 accidents per year.
I heard a great interview of a Tesla engineer speaking about the ānines.ā I could not find a transcript of the interview but this link contains a great summary:
I think basically what takes the long amount of time and the way to think about it is that itās a march of nines and every single nine is a constant amount of work, so every single nine is the same amount of work, so when you get a demo and something works 90% of the time, thatās just the first nine, and then you need the second nine, and third nine, fourth nine, fifth nine, and while I was at Tesla for five years or so, i think we went through maybe three nines or two nines. I donāt know, but like multiple nines of iteration, thereās still more nines to go, and so thatās why these things take so long [ā¦]ā
A very good explanation indeed! What is missing is whether this rate is with or without edge case simulation. The first two nines catch most driving experience while the problems are the expensive edge cases.