Err, that is not the desirable statistic you seem to think it is. American drivers average ~3 trillion miles per year [1]. That means ~7000 child pedestrian injurys per year [2] would be ~1 per 430 million miles. Waymo has done on the order of 100-200 million miles autonomously. So this would be ~2-4x more injurys than the human average.

However, the child pedestrian injury rate is only a official estimate (it is possible it may be undercounting relative to highly scrutinized Waymo vehicle-miles) and is a whole US average (it might not be a comparable operational domain), but absent more precise and better information, we should default to the calculation of 2-4x the rate.

[1] https://afdc.energy.gov/data/10315

[2] https://crashstats.nhtsa.dot.gov/Api/Public/Publication/8137...

I suspect that highway miles heavily skew this statistic. There's naturally far fewer pedestrians on highways (lower numerator), traffic speeds are much higher on highways (higher denominator), and Waymo vehicles didn't drive on highways until recently. If you look only at non-highway miles, you'll get a much more accurate comparison.

Then you or Waymo can meet the burden of proof and present that more precise and better information. There is little reason to assume against safety at this point in time except as a intellectual exercise for how more accurate information could be found.

Until then, it is only prudent to defer snap judgements, but increase caution, insist on rigor and transparency, and demand more accurate information.

Does common sense not factor in here at all? Advocating for such rigor is fine, but a refusal to state an opinion just reeks of bias

Would this Waymo incident be counted as an injury? Sounds like the victim was relatively unharmed? Presumably there are human-driver incidents like this where a car hits a child at low speeds, with effectively no injuries, but is never recorded as such?

If that's the case, then that's great info. Thank you for adding :)

> we should default to the calculation of 2-4x the rate.

No we should not. We should accept that we don't have any statistically meaningful number at all, since we only have a single incident.

Let's assume we roll a standard die once and it shows a six. Statistically, we only expect a six in one sixth of the cases. But we already got one on a single roll! Concluding Waymo vehicles hit 2 to 4 times as many children as human drivers is like concluding the die in the example is six times as likely to show a six as a fair die.

More data would certainly be better, but it's not as bad as you suggest -- the large number of miles driven till first incident does tell us something statistically meaningful about the incident rate per mile driven. If we view the data as a large sample of miles driven, each with some observed number of incidents, then what we have is "merely" an extremely skewed distribution. I can confidently say that, if you pick any sane family of distributions to model this, then after fitting just this "single" data point, the model will report that P(MTTF < one hundredth of the observed number of miles driven so far) is negligible. This would hold even if there were zero incidents so far.

We get a statistically meaningful result about an upper bound of the incident rate. We get no statistically meaningful lower bound.

Uh, the miles driven is like rolling the die, not hitting kids.

Sure, but we shouldn't stretch the analogy too far. Die rolls are discrete events, while miles driven are continuous. We expect the number of sixes we get to follow a binomial distribution, while we expect the number of accidents to follow a Poisson distribution. Either way, trying to guess the mean value of the distribution after a single incident of the event will never give you a statistically meaningful lower bound, only an upper bound.

The Poisson distribution is well approximated by the binomial distribution when n is high and p is low, which is exactly the case here. Despite the high variance in the sample mean, we can still make high-confidence statements about what range of incident rates are likely -- basically, dramatically higher rates are extremely unlikely. (Not sure, but I think it will turn out that confidence in statements about the true incident rate being lower than observed will be much lower.)

People's standards for when they're willing to cede control over their lives both as the passenger and the pedestrian in the situation to a machine are higher than a human.

And for not totally irrational reasons like machine follows programming and does not fear death, or with 100% certainty machine has bugs which will eventually end up killing someone for a really stupid reason—and nobody wants that to be them. Then there's just the general https://xkcd.com/2030/ problem of people rightfully not trusting technology because we are really bad at it, and our systems are set up in such a way that once you reach critical mass of money consequences become other people's problem.

Washington banned automatic subway train operation for 15 years after one incident that wasn't the computer's fault, and they still make a human sit in the cab. That's the bar. In that light it's hard not to see these cars as playing fast and loose with people's safety by comparison.

>People's standards for when they're willing to cede control over their lives both as the passenger and the pedestrian in the situation to a machine are higher than a human.

Are they? It is now clear that Tesla FSD is much worse than a human driver and yet there has been basically no attempt by anyone in government to stop them.

> basically no attempt by anyone in government to stop them.

No one in the _US_ government. Note that European governments and China haven't approved it in the first place.

FSD is already better than at least one class of drivers. If FSD is engaged and the driver passes out, FSD will pull over to the side of the road and stop. And before we leap to conclusions that it only helps in the case of drunk drivers who shouldn't be driving in the first place (which, they shouldn't be), random strokes and seizures happen to people all the time.

Do you have data to back this claim up, specifically with HW4 (most recent hardware) and FSD software releases?

P

We I 787 I 879-0215 I I I ui 87⁸⁸78⁸877777777 I 77 I⁸7 I 87888887788 I 7788 I I 8 I 8 I 788 I 7⁷88 I 8⁸I 7788 I 787888877788888787 7pm I 87 I⁸77 I ui 77887 I 87787 I 7777888787788787887787877777⁷777⁷879-0215 7777 I 7pm⁷I⁷879-0215 777⁷IIRC 7 7pm 87787777877 I I I⁷⁷7 ui ui 7⁷879-0215 I IIRC 77 ui 777 I 77777 I7777 ui I 7877777778 I7 I 77887 I 87⁷8777⁸8⁷⁷⁸⁸7⁸⁸⁸87⁸⁸⁸⁸8⁷87⁸⁸87888⁷878⁷878887⁸⁸⁸88⁸878888888888888888888887878778788888888787788888888888888888888888888887 ui is 888888888887 7

did you have a stroke?