(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
. . . . . . . . . .



Pump the Brakes on the Waymo Safety Data [1]

['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.']

Date: 2023-12-22

Waymo put out a report that claimed it as six times safer than human drivers. I don’t think that it is impressive as it sounds.

Before getting into the details, though, we really should acknowledge that Waymo is doing this the right way. They are using public data, though there are limits I will discuss, and they are upfront with their assumptions. This is all good behavior and should be acknowledged. With that said, I am not certain the takeaway is as rosy as they want it to seem.

Let me start by saying I am not a research scientist, so it’s possible I missed material in the report itself. But no reporting I have seen, including the Waymo website itself, fully addresses these concerns.

First, the study is small by human standards. They have 7 million miles driven but human drivers go 100 million miles between fatalities. This is, then, very preliminary data and should be treated as such.

Next, the study makes assumptions to try and deal with human underreporting of accidents. Not everyone reports every fender-bender. We do have to take Waymo’s word that they are reporting all accidents. We know that other companies have lied and misreported this data, but I am unaware of Waymo being caught in such deceit. Per the report, Waymo tried to adjust for underreporting and the fact that they are working in geofenced, non-highway largely good weather environments. Other researchers have stated that their assumptions about human underreporting are too generous to Waymo. That, in turn, would have some effect on their numbers.

The desired takeaway from this report is that Waymo cars are 2-7 times safer. But as the head of the Insurance Institute for Highway Safety noted, we cannot necessarily extrapolate accident data to fatality data. Waymo cars may have failure modes that are not clear at this stage, given how little relative driving they have under their belts and how constrained the conditions they drive under. Remember, a Cruise car drug a woman for several feet after an accident. That is not something a human would do or would expect.

There also seems to be no adjustments for impaired drivers. Now, that may sound odd, but I am not sure it is. There is no immediate future in which self-driving cars are forced on every driver. So therefor, the people who drive drunk or otherwise impaired will still refuse to take a Waymo taxi the same way they refuse to take a human driven taxi. What matters in the real world is whether the human taxi is more or less or about the same safe as the Waymo taxi. Keeping impaired drivers in the mix makes answering that question harder and tilts the results toward Waymo’s desired outcome.

It seems that the report suggest — but does not prove — that Waymo cars in geofenced locations under highly favorable conditions can be safer than human driven cars. However, what I have described — vehicles running set routes in highly controlled conditions — also describes subways and light rail. Our question, then, is whether or not the money spent on Waymo is better than the money spent applying mass transit to the kinds of areas that Waymo can safely operate in. I believe mass transit would be a better way to spend that money, but I am open to being shown otherwise. I think we should be considering that question rather than just assuming Waymo-style self-driving is definitively the correct answer.

More importantly, I think, is that it seems clear that to really get a handle on how safe these cars are is going to require driving millions and millions of more miles. That, then, leads to this question: is it moral to force human beings to be the test subjects of a self-driving experiment they have not agreed to participate in? I think the answer is clearly no. There are other means of getting those millions and millions of miles, they are just expensive for companies like Waymo.

That, as they say, is not my problem. I would be even willing to go so far as to subsidize proper training away from the general population if the answer to the mass transit vs. Waymo question can realistically be Waymo. But the idea that we must allow these experiments to continue in the hope that self-driving cars will be viable outside a mass transit context is not, in my mind, a reasonable proposition.

Reports like these are meant to be reassuring, meant to encourage us to allow Waymo to continue to experiment on the general population. What they really show, I think, is just how little we actually know, and how much more research has to be done. I don’t think my family, or yours, should have to bear the brunt of the risks of those experiments.

[END]
---
[1] Url: https://www.dailykos.com/stories/2023/12/22/2213098/-Pump-the-Brakes-on-the-Waymo-Safety-Data?pm_campaign=front_page&pm_source=more_community&pm_medium=web

Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.