We do know that Waymos are safer than human drivers

Gonna try a new thing:

The irony is almost too perfect. The Piper article's core thesis is that the Bloomberg piece "dances around the data rather than explaining it" - and then the comments dance around the article rather than engaging with it.

4d4m's "cherry-picked" accusation is particularly maddening because Piper's entire methodological point is that you can pick any of the available metrics - police-reported crashes, any-injury crashes, airbag deployments, serious injuries - and they all converge on the same 5-10x safety advantage. That's literally the opposite of cherry-picking. Cherry-picking is when you find the one metric that supports your conclusion. Finding consistency across all available metrics is robustness.

tim-tday just recites the crime scene example that Piper explicitly addresses in the article. The railroad tracks issue was a software bug that got patched. They're treating "I remember some headlines" as equivalent to "I have engaged with the statistical argument."

The "sensor package costs as much as a 4-bedroom home" is also just... not an argument about safety? It's not even an argument about cost-effectiveness since Waymo is a service, not a product you buy.

What's the actual function this serves for commenters? My read: engaging with the statistical argument would require either conceding the point or doing actual work to find methodological flaws. Handwaving about "cherry-picking" and listing failure anecdotes lets them maintain their prior (skepticism of corporate claims = smart) without that effort.

I think that even if they do end up being safer statistically, they will fail differently than humans, so that people learning of Waymo accidents will correctly say, "I NEVER would have done something that stupid!"

This argument makes me think Waymo find us consumers stupid - and I think Waymo hopes by saying it enough people will parrot it (like they did Tesla FSD self-published "stats".)

Of course Waymo claims its safer in cherry-picked data - but it's a silly claim. Any self-driving vehicle is NOT inherently safer than a human driver due to scope and capability.

A human can drive a car unassisted. However - a Waymo in a unmapped area or without a supervisory human teleoperator cannot run, and therefore a safety comparison is apples to oranges and as presented completely disingenuous.

No Waymo can get you out of an emergency situation. There is no Waymo running without these two conditions of a supervisory driver and limited geographic area it works in, and newsflash they tend to run in fair weather areas with high tolerances for drivers breaking common laws and creating nuisances, which Waymo doesn't seem to be counting. I think the societal costs are tracking and billing directly to those experimenting on us on the roads instead of closed testing environments...

Signed, a huge fan of Waymo, and of being objective. We're not that dumb.

Except when there’s a power outage. Of railroad tracks Or when they’re driving through active police crime scenes

Or massive connectivity outages Or massive cloud infrastructure outages

Also the sensor package costs as much as 4bedroom home in the Midwest.

Other than those things, they’re better.