“On roads suited to AI, upon which they have been practicing and testing extensively “
There, they must have not been able to fit that into the headline .
I believe there is a hole in that study somewhere, we just aren’t privy to it yet.
Well, not holes but significant caveats :
- testing has taken place on a tiny subset of American roads
- there is some history of dishonesty in the self driving industry
So, giving them the benefit of the doubt, Swiss Re are right - but there are externalities that constrain the value of the results.
"…some critics have questioned whether the cars have been on the road long enough to begin making fair comparisons to human drivers. For example, Waymo’s driverless cars have traveled roughly 4 million miles, but traffic-related fatalities are routinely measured per 100 million miles. Across America’s roadways last year, there were 1.37 fatalities for every 100 million miles traveled, totaling 42,795 lives lost in vehicle traffic crashes, according to the National Highway Traffic Safety Administration. Each year, human drivers log more than 3 trillion miles on the road.
“It’s interesting that we were able to show statistical significance with so little data,” Victor said. “That has to do with two things: one is that we have very robust baselines from Swiss Re, and the other is that we have incredible safety performance.”
Waymo did not release the raw data it used to come up with its findings, saying that information is “proprietary.”
“The information they released is the results of the study rather than all the details,” said Steven Shladover, a research engineer with the Berkeley’s Institute of Transportation Studies. “Of course, as researchers, we like to look more at the details and the methods for the analysis.”
Shladover said he was also “disappointed” the study grouped its findings for both San Francisco and Phoenix instead of sharing separate data points for each city. While Waymo’s study only offered combined figures, it examined different time spans for each city: Waymo’s Phoenix data spanned more than four years (March 2019 to Aug. 1, 2023), while Waymo’s San Francisco data incorporated less than two years’ worth of driving (March 2022 to Aug. 1, 2023), according to a Waymo spokesperson.
“It lumped them all together and those are quite different areas,” he said. “I think it would have been more interesting to see a comparison for San Francisco and then a separate comparison for Phoenix.”
Leaders at Waymo and Swiss Re said they plan to eventually submit their study to a scientific journal so that it can be independently reviewed, however, neither company would say when that might happen." Source
mmm, Swiss Re have been in the insurance game since 1863. I’d go with the idea they know what they’re doing.
Nestlé started in 1867. Guess that makes them good guys, eh?
No one said “good guys”, just that they know what they’re doing. And I’d say Nestle knows exactly what they’re doing.
When you can’t discern expertise from morality you need to reevaluate… something
Self driving doesn’t have to be perfect to be safer than humans driving.
No, but even if the number of deaths are significantly lower, the deaths that do occur can’t be of a type a human would have absolutely avoided (ex: mistaking a truck for a sun glare, as happened with a Tesla IIRC), if we want wide adoption…
Repeat after me:
Waymo is Google.
I live where water falls as a solid most of the time and I rarely see self driving vehicles. Hope to one day have that change because I hate driving
If they are comparing Waymo to taxi drivers, then I think that is a low bar, based on my personal experience.
I mean, we’ve been expecting this for a long time. Reaction time and consistency are not human strengths. It’s just that there’s been a really heavy tail of weird situations where humans actually have an edge. I’m not surprised that in carefully chosen spots, Waymo cars - which have a reputation for being the smartest - blow us out of the water.
