You know how I've been saying since 2016 that Tesla's comparison of raw Autopilot crash data to average human driver crash data is deceptive because it doesn't adjust for the safety of the operating domain (among other factors)?
proved it!
engrxiv.org/preprint/view/
Conversation
Replying to
In addition to creating an apples-to-apples comparison of crash data based on the operating domain, he also adjusted for owner age. The results effectively eliminate Tesla's claim that Autopilot meaningfully improves safety outcomes. This is huge.
15
76
273
Like any good academic, Goodall is upfront about the assumptions he makes in his analysis and the limitations of his study.
I find the study convincing and the assumptions reasonable, but then I've been calling this since 2016. Read it for yourself:
engrxiv.org/preprint/view/
4
8
119
Here's that 2016 debunking I mentioned, huge thanks to for collaborating and for assigning/editing it. It's been hard having to wait six years for this vindication, but better late than never!
2
7
78
Also gotta shout out for doing great work on this over : arstechnica.com/cars/2018/05/s
MIT was also all over this issue early: technologyreview.com/2016/07/06/158
Thanks to everyone else who stood up for the truth on this issue. Truth always wins out!
4
6
69
Really cool paper, thanks for posting that.
One question: why no confidence intervals on the chart?
1
4
Cheers! Good question, but not one I can answer I'm afraid. Hopefully (the paper's author) can provide some color on that.
2
The author assumes (incorrectly) that autopilot is used almost exclusively on highways and not in cities, and then extrapolates a conclusion from that faulty assumption. Although his math is impressive, its based on a guess by someone who clearly has never used autopilot.
2
3
The author goes through the limitations of his analysis in a very transparent manner in section 5.1. What you call his "assumption" is in fact based on MIT's naturalistic driving study. It's not perfect, but it's way more credible than your subjective feelings as a user.
1
43
Show replies
