Model X Crash on US-101 (Mountain View, CA)

[cmaster]

This is an interesting twist for a solution to a problem. When the user is at fault, blame the machine. It happens every where in life. This just ruins it for everyone else who wants to use it.

Just realized that I reversed the agencies in the second to last paragraph of my last post. I meant to say

"By contrast, while NTSB might recommend a new regulation here, it is more likely that, if they find that AP was a cause of the accident, they will find that the implementation of TACC+AC was insufficiently safe or overly prone to dangerous "normal misuse" (ie use on inappropriate roads, no-hands use, or inattentive use). In that event, NHTSA would consider whether TACC+AP is "defective." The likely result of such a defect investigation would be NHTSA arm twisting Tesla to either turn off some AP features, add more software-based limitations on use of features, or change the instructions for use of AP."

 

[Peteski]

Just realized that I reversed the agencies in the second to last paragraph of my last post. I meant to say

"By contrast, while NTSB might recommend a new regulation here, it is more likely that, if they find that AP was a cause of the accident, they will find that the implementation of TACC+AC was insufficiently safe or overly prone to dangerous "normal misuse" (ie use on inappropriate roads, no-hands use, or inattentive use). In that event, NHTSA would consider whether TACC+AP is "defective." The likely result of such a defect investigation would be NHTSA arm twisting Tesla to either turn off some AP features, add more software-based limitations on use of features, or change the instructions for use of AP."

All vehicles on the road are prone to misuse by their drivers. It's easy to get into any vehicle and drive it however dangerously or inappropriately that you feel like without the vehicle itself holding you back from your own stupidity. You can only really say that AP is "defective" if it actually malfunctions in some way while being used appropriately e.g. suddenly slams the brakes full on for no reason or suddenly tries to swerve off the road while following a clearly marked suitable road. Of course AP is prone to misuse and so are all other driving controls in a vehicle. So I don't think you can use that argument alone to limit AP functionality.

 

[Economite]

All vehicles on the road are prone to misuse by their drivers. It's easy to get into any vehicle and drive it however dangerously or inappropriately that you feel like without the vehicle itself holding you back from your own stupidity. You can only really say that AP is "defective" if it actually malfunctions in some way while being used appropriately e.g. suddenly slams the brakes full on for no reason or suddenly tries to swerve off the road while following a clearly marked suitable road. Of course AP is prone to misuse and so are all other driving controls in a vehicle. So I don't think you can use that argument alone to limit AP functionality.

That's not the standard NHTSA uses. From footnotes 19 and 20 of the NHTSA report on the Florida accident:

"19 An unreasonable risks due to owner abuse that is reasonably foreseeable (i.e., ordinary abuse) may constitute a safety-related defect. See United States v. Gen. Motors Corp., 518 F.2d 420, 427 (D.C. Cir. 1975) (“Wheels”).

20 Driver misuse in the context of semi-autonomous vehicles is an emerging issue and the agency intends to continue its evaluation and monitoring of this topic, including best practices for handling driver misuse as well as driver education."

https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF

 

[Barklikeadog]

That's not the standard NHTSA uses. From footnotes 19 and 20 of the NHTSA report on the Florida accident:

"19 An unreasonable risks due to owner abuse that is reasonably foreseeable (i.e., ordinary abuse) may constitute a safety-related defect. See United States v. Gen. Motors Corp., 518 F.2d 420, 427 (D.C. Cir. 1975) (“Wheels”).

20 Driver misuse in the context of semi-autonomous vehicles is an emerging issue and the agency intends to continue its evaluation and monitoring of this topic, including best practices for handling driver misuse as well as driver education."

https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF

Ah we will be at Idiocracy soon enough. Every commercial will have the standard disclaimer: Do not stick ______ up your ___.

Even car commercials!

 

[S4WRXTTCS]

This is an interesting twist for a solution to a problem. When the user is at fault, blame the machine. It happens every where in life. This just ruins it for everyone else who wants to use it.

I think it's fair when something about the design directly contributed to an accident.

A good example of that is the Jeep Grand Cherokee electronic shifter issue. Where the shifter was hard to use, and it didn't always go into park even if you thought you had put it into park. This UI/UX issue led to the fatality accident involving a hollywood actor.

The actor used it incorrectly, and the actor didn't use the parking brake. So is it Jeeps fault? Some would say no, but a lot would say yet. Because Jeep knew that it was badly designed.

Another example would be the recent Honda Odyssey fatality. Where a teenage kid got stuck between the rear seat and the rear gate. The seat wasn't correctly latched down. Was that Honda's fault or was it the users fault? Was it foreseeable? Probably not. I think most people would say it was a freak accident that Honda has no fault.

With the Model X fatality it's hard to say what the breakdown is. It seems like most owners like myself don't find Tesla at fault at least with the information we have so far. The fact that the user knew it didn't work very well in that area makes it extremely difficult to blame Tesla. Even when I don't like the implementation of AP.

 

[cmaster]

I do find it interesting that when a knife is used to kill someone, the company of knife isn't at fault. It's the person who is using the knife.

 

[Barklikeadog]

I do find it interesting that when a knife is used to kill someone, the company of knife isn't at fault. It's the person who is using the knife.

I admire your 'spin', you remind me of those who think free speech is universal etc etc

If that knife company marketed their knife as having safety features that would prevent cuts or eliminate stabbings... and the knife cut off a finger. Yea, that would
be a problem for the knife company.

 

[Economite]

Which ignores that many cars never had AS enabled yet they had all the AEB tweaks and they were in the bucket of higher air bag deployments.

Nhtsa could have divided up and presented the data in whatever way they found most relevant . They could have compared different versions of AEB. But they didnt. They choose to present it as cars without AS and cars with AS.

There is no basis to think that they would deliberalty or even ignorantly present data correlations that were misleading and had statistically relevant confounding factors.

But those anchored to a preconceived view of how they imagine AEB and LDW and AS to work, even when they haven't even used them for their own daily commute, will find it hard to comprehend.

NHTSA has now publicly disavowed Tesla's use of the 40% reduction in crashes statistic as a defense of AS.

Per Reuters:

"[NHTSA] on Wednesday contradicted Tesla Inc’s claim that the agency had found that its Autopilot technology significantly reduced crashes, saying that regulators 'did not assess' the system’s effectiveness in a 2017 report....The agency said on Wednesday its crash rate comparison 'did not evaluate whether Autosteer was engaged' and 'did not assess the effectiveness of this technology.' The agency described the Autopilot analysis in the 2017 report as a 'cursory comparison' of airbag deployment rates before and after installation of the feature to determine whether models with Autosteer had higher crash rates. Such a finding 'could have indicated that further investigation was necessary,' the agency said."

U.S. safety agency says 'did not assess' Tesla Autopilot effectiveness

 

[cmaster]

We can't get to the the details unless these questions are asked. Needed for clarification. Not about free speech I'd say.

A breakdown in communication is to stop asking the questions. Not quite the same as circular reasoning. Eventually, the discussion will motivate people to get to further details.

We all like to get very defensive when any questions go against our understanding.

I admire your 'spin', you remind me of those who think free speech is universal etc etc

If that knife company marketed their knife as having safety features that would prevent cuts or eliminate stabbings... and the knife cut off a finger. Yea, that would
be a problem for the knife company.

 

[bhzmark]

They aren't denying the data. Just stating that they didn't make the causal inference that Tesla did in the post.

That inference seems reasonable to me unless we learn that there was some vastly significant update to AEB at the same time that better explains the data. Having owned a car during that entire time, and paying attention to the reports of what was changing , I'm confident that there was not a better explanation.

People coming to this without that detailed knowledge may find it harder to make that inference.

NHTSA has now publicly disavowed Tesla's use of the 40% reduction in crashes statistic as a defense of AS.

Per Reuters:

"[NHTSA] on Wednesday contradicted Tesla Inc’s claim that the agency had found that its Autopilot technology significantly reduced crashes, saying that regulators 'did not assess' the system’s effectiveness in a 2017 report....The agency said on Wednesday its crash rate comparison 'did not evaluate whether Autosteer was engaged' and 'did not assess the effectiveness of this technology.' The agency described the Autopilot analysis in the 2017 report as a 'cursory comparison' of airbag deployment rates before and after installation of the feature to determine whether models with Autosteer had higher crash rates. Such a finding 'could have indicated that further investigation was necessary,' the agency said."

U.S. safety agency says 'did not assess' Tesla Autopilot effectiveness

 

[Az_Rael]

They aren't denying the data. Just stating that they didn't make the causal inference that Tesla did in the post.

That inference seems reasonable to me unless we learn that there was some vastly significant update to AEB at the same time that better explains the data. Having owned a car during that entire time, and paying attention to the reports of what was changing , I'm confident that there was not a better explanation.

People coming to this without that detailed knowledge may find it harder to make that inference.

I have a hard time understanding how Autosteer alone could be attributed to a 40% reduction in airbag deployments. Here is what they said in the preliminary report from the Florida crash: https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF

5.4 Crash rates. ODI analyzed mileage and airbag deployment data supplied by Tesla for all MY 2014 through 2016 Model S and 2016 Model X vehicles equipped with the Autopilot Technology Package, either installed in the vehicle when sold or through an OTA update, to calculate crash rates by miles travelled prior to (21) and after Autopilot installation.(22) Figure 11 shows the rates calculated by ODI for airbag deployment crashes in the subject Tesla vehicles before and after Autosteer installation. The data show that the Tesla vehicles crash rate dropped by almost 40 percent after Autosteer installation.

21 Approximately one-third of the subject vehicles accumulated mileage prior to Autopilot installation.
22 The crash rates are for all miles travelled before and after Autopilot installation and are not limited to actual Autopilot use.


So when did FCW, AEB and TACC get turned on during the initial rollout of AP1? I think Autosteer came last? To me it makes a lot more sense that the 40% includes the whole package, not just Autosteer.

 

[Barklikeadog]

I have a hard time understanding how Autosteer alone could be attributed to a 40% reduction in airbag deployments. Here is what they said in the preliminary report from the Florida crash: https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF

5.4 Crash rates. ODI analyzed mileage and airbag deployment data supplied by Tesla for all MY 2014 through 2016 Model S and 2016 Model X vehicles equipped with the Autopilot Technology Package, either installed in the vehicle when sold or through an OTA update, to calculate crash rates by miles travelled prior to (21) and after Autopilot installation.(22) Figure 11 shows the rates calculated by ODI for airbag deployment crashes in the subject Tesla vehicles before and after Autosteer installation. The data show that the Tesla vehicles crash rate dropped by almost 40 percent after Autosteer installation.

21 Approximately one-third of the subject vehicles accumulated mileage prior to Autopilot installation.
22 The crash rates are for all miles travelled before and after Autopilot installation and are not limited to actual Autopilot use.


So when did FCW, AEB and TACC get turned on during the initial rollout of AP1? I think Autosteer came last? To me it makes a lot more sense that the 40% includes the whole package, not just Autosteer.

An easy barometer is to look at autos without AP, but have AEB, FCW, blind spot, etc... and they are showing huge declines in serious accidents. an IIHS study had lane departure
showing a 21% drop in injury accidents, 24% drop for blind spot det. And this was just the general population. Includes elderly, teenagers, etc. Adjust for those factors and the numbers
are much higher.

 

[bhzmark]

I have a hard time understanding how Autosteer alone could be attributed to a 40% reduction in airbag deployments. Here is what they said in the preliminary report from the Florida crash: https://static.nhtsa.gov/odi/inv/2016/INCLA-PE16007-7876.PDF

5.4 Crash rates. ODI analyzed mileage and airbag deployment data supplied by Tesla for all MY 2014 through 2016 Model S and 2016 Model X vehicles equipped with the Autopilot Technology Package, either installed in the vehicle when sold or through an OTA update, to calculate crash rates by miles travelled prior to (21) and after Autopilot installation.(22) Figure 11 shows the rates calculated by ODI for airbag deployment crashes in the subject Tesla vehicles before and after Autosteer installation. The data show that the Tesla vehicles crash rate dropped by almost 40 percent after Autosteer installation.

21 Approximately one-third of the subject vehicles accumulated mileage prior to Autopilot installation.
22 The crash rates are for all miles travelled before and after Autopilot installation and are not limited to actual Autopilot use.


So when did FCW, AEB and TACC get turned on during the initial rollout of AP1? I think Autosteer came last? To me it makes a lot more sense that the 40% includes the whole package, not just Autosteer.

Discussed in detail earlier in thread. AEB and FCW were turned on early in 2015. So miles driven with AEB but not AS are Included in the higher crash rate and come from all cars prior to AS rollout in late October 2015, as well as those cars who did not buy AS ever but had mature AEB systems after October 2015.

AEB improvements are definitely a possible confounding factor, but I think the best causal explanation for 40% decrease in airbag deployments is a combination of both AS being available (and thus people using it) and maturing AEB. But since the maturing AEB was in some cars making up the higher rate, and the tweaks to AEB were not reported to be anything major, I think the AS addition is the most likely primary, or at least very substantive, causal factor for the decrease in air bag deployments.

 

[Az_Rael]

Discussed in detail earlier in thread. AEB and FCW were turned on early in 2015. So miles driven with AEB but not AS are Included in the higher crash rate and come from all cars prior to AS rollout in late October 2015, as well as those cars who did not buy AS ever but had mature AEB systems after October 2015.

AEB improvements are definitely a possible confounding factor, but I think the best causal explanation for 40% decrease in airbag deployments is a combination of both AS being available (and thus people using it) and maturing AEB. But since the maturing AEB was in some cars making up the higher rate, and the tweaks to AEB were not reported to be anything major, I think the AS addition is the most likely primary, or at least very substantive, causal factor for the decrease in air bag deployments.

There is TACC to consider as well. I would say a combo of TACC and Autosteer would be a big improvement over non-AP use. My experiences with TACC even without autosteer engaged indicate it catches sudden slowdowns faster than I can, giving me more time to react.

 

[Economite]

Here's a kind-of-good article about that statistic: Tesla's Favorite Autopilot Safety Stat Just Doesn't Hold Up

 

[bro1999]

Tesla’s Autopilot, suspected in fatal crash, has a history of swerves

 

[mongo]

Tesla’s Autopilot, suspected in fatal crash, has a history of swerves

A rehash of data available from TMC.
From article, regarding the main incident the article discusses:

“It was, in fact, in a construction zone, and the reason for the car swerving was valid,” said Holmes, a former tight end for the Miami Dolphins and now co-CEO of Wild Hair Media.

Thread on the incident from 10 months ago:
Autopilot Fail

 

[mongo]

Here's a kind-of-good article about that statistic: Tesla's Favorite Autopilot Safety Stat Just Doesn't Hold Up

And kind-of-bad, since the body of the article says the issue is that data is insufficient, not that the claim is definitely wrong:

Which is all to say, determining whether a new feature like Autopilot is safe, especially if you don’t have access to lots of replicable, third-party data, is super, super hard. Tesla’s beloved 40 percent figure comes with so many caveats, it’s unreliable.

When it [IIHS] tried to separate Model S sedan incidents after Autopilot was released, it observed no changes in the frequency of property damage and bodily injury liability claims. That indicates that Autopilot drivers aren’t more or less less likely to damage their cars or get hurt than others. But it did find a 13 percent reduction in collision claim frequency, indicating sedans with Autopilot enabled got into fewer crashes that resulted in collision claims to insurers.

The upshot is that Autopilot might, in fact, be saving a ton of lives. Or maybe not. We just don’t know.

It will be interesting to see next quarter's safety numbers.

 

[Kant.Ing]

Tesla should start its own auto insurance division for non-commercial vehicles. It's not cool I'm still writing my checks to TRUCK INSURANCE EXCHANGE!

 

[Economite]

And kind-of-bad, since the body of the article says the issue is that data is insufficient, not that the claim is definitely wrong:

By "kind-of-good" I meant reasonably well-written and thought out, not "good for Tesla" or "good for any particular argument." I call it a "kind-of-good" rather than "good" article because I thought the author somewhere misunderstood a couple of the complaints about the methodology.





It will be interesting to see next quarter's safety numbers.