The F150 accident was two young people manually driving their vehicle into a barrier off the road, likely while intoxicated, and not wearing seatbelts. Different scenario. Also, the media reported on both situations, did mean the tesla one was picked up nationally?
Do you know that they were manually driving the F150? Newer models have adaptive cruise control, pre-collision automatic breaking, a lane-keeping system, evasive steering assist, and they starting implementing self-driving features 2 years ago.
Exactly my point. The fact that it is unrelated without making any useful point IS the point. It has become so vogue and tiresome at HN to hit on everything Musk. Tesla remains the safest car on the road with advanced vehicle technologies:
"Tesla cars using driver-assist systems are significantly less likely to crash than Waymo/Transdev/GM's Cruise, according to NHTSA's data. Although the regulator's data only show the number of accidents such a conclusion is obvious with a little digging" [1]
This refers to "NHTSA Releases Initial Data on Safety Performance of Advanced Vehicle Technologies" (June 15, 2022) [2]. In that actual report, it misleads by showing that Tesla has the highest number of crashes of all SAE Levels 3-5 automated driving systems, when in fact that is only because they have the highest number of Level3-5 vehicles and drivers on the road and in fact they have the lowest number of accidents per thousand drivers per year. There is a good video discussion on this [3]
You won't hear any complaint about me that the reports and the articles about them show basic innumeracy (at least the NHTSA report says "Summary Incident Report Data Are Not Normalized").
But it's quite clear the report has other, deep issues, which prevent it from being useful: "it is important to note that these crashes are categorized based on what driving automation system was reported as being equipped on the vehicle, not on what system was reported to be engaged at the time of the incident."
I only want to see a report about crashes that occurred when the system engaged and the user was not given an opportunity to take over with enough lead time.
I don't think the companies are sharing that detailed information right now.
If it's terminal brain fart by an active driver, that sux. If it's an example of auto-driving having issues, or worse, a Darwin award from somebody uninterested in paying attention, that's a datapoint for the argument as to if/when we'll see auto-drive.
While it's possible to design fault-tolerant systems that are quite capable, I feel they're at a level of cost and effort that's not being used for current-day programming in pretty much any form. I know they've been working on computer vision since the 60's so I assume that's solved (enough) to trust its data.
Tesla gets clicks. Apart from that though, probably because Elon has been marketing them as having FSD soon for nearly a decade, plus numerous claims that autopilot is already safer than humans. Sure, not every Tesla has that but most people are more ignorant than you might think. The number of people I've met who think a Tesla can actually drive you hundreds of miles while you sleep in the back is alarming.
Schadenfreude, probably. No other car manufacturer encourages people to take their hands off the wheel, light a blunt, and watch the world burn (nb: egregious poetic license).
is there any other company with a similar technology (activated and working instead of the driving having any control) with a similar number of miles, and if so, what are their accident and death rates when their technology is enabled?
IIUC MB and other manufacturers have fairly sophisticated self-driving packages now but I don't know how they compare to Tesla, in terms of technology or marketing or user expectation.
My guess is: this is an easy article because it gets clicks because it feeds into a bunch of people existing assumptions about self-driving, and also Musk gets a lot of clicks too. I haven't seen many (any?) articles that actually attempt to interpret all the available data about self-driving car accidents in the US and draw any reasonable conclusions, in a good-faith way (rather than promoting a pre-existing agenda or going to cheap clicks).
Note that I'm mainly interested in the situation of accidents that occur during full hands-off automated driving where the system had not started notifying the user to pay attention and the weather and road conditions are not adverse. Having accurate estimates of accident and death frequencies that could be used to compare autonomous systems confidently would be great.
22 comments
[ 1.7 ms ] story [ 70.2 ms ] threadIsn't this a rehash without any further updates?
Probably dupes of:
https://news.ycombinator.com/item?id=32023337
https://news.ycombinator.com/item?id=32052267
[1] https://www.houstonchronicle.com/news/houston-texas/houston/...
i think you're on the right track with that one
"Tesla cars using driver-assist systems are significantly less likely to crash than Waymo/Transdev/GM's Cruise, according to NHTSA's data. Although the regulator's data only show the number of accidents such a conclusion is obvious with a little digging" [1]
This refers to "NHTSA Releases Initial Data on Safety Performance of Advanced Vehicle Technologies" (June 15, 2022) [2]. In that actual report, it misleads by showing that Tesla has the highest number of crashes of all SAE Levels 3-5 automated driving systems, when in fact that is only because they have the highest number of Level3-5 vehicles and drivers on the road and in fact they have the lowest number of accidents per thousand drivers per year. There is a good video discussion on this [3]
[1] https://twitter.com/EvaFoxU/status/1537149521955889153
[2] https://www.nhtsa.gov/press-releases/initial-data-release-ad...
[3] https://www.youtube.com/watch?v=SMpozSBXpN4&t=459
But it's quite clear the report has other, deep issues, which prevent it from being useful: "it is important to note that these crashes are categorized based on what driving automation system was reported as being equipped on the vehicle, not on what system was reported to be engaged at the time of the incident."
I only want to see a report about crashes that occurred when the system engaged and the user was not given an opportunity to take over with enough lead time. I don't think the companies are sharing that detailed information right now.
While it's possible to design fault-tolerant systems that are quite capable, I feel they're at a level of cost and effort that's not being used for current-day programming in pretty much any form. I know they've been working on computer vision since the 60's so I assume that's solved (enough) to trust its data.
IIUC MB and other manufacturers have fairly sophisticated self-driving packages now but I don't know how they compare to Tesla, in terms of technology or marketing or user expectation.
My guess is: this is an easy article because it gets clicks because it feeds into a bunch of people existing assumptions about self-driving, and also Musk gets a lot of clicks too. I haven't seen many (any?) articles that actually attempt to interpret all the available data about self-driving car accidents in the US and draw any reasonable conclusions, in a good-faith way (rather than promoting a pre-existing agenda or going to cheap clicks).
Note that I'm mainly interested in the situation of accidents that occur during full hands-off automated driving where the system had not started notifying the user to pay attention and the weather and road conditions are not adverse. Having accurate estimates of accident and death frequencies that could be used to compare autonomous systems confidently would be great.