This was the one where the media reported that because nobody was found in the front seat it must have been the fault of the self-driving feature. The NTSB found that none of the driver-assistance features were activated, and the driver was probably tossed into the backseat from the impact:
Although the driver’s seat was found vacant and the driver was found in the left rear seat, the available evidence suggests that the driver was seated in the driver’s seat at the time of the crash and moved into the rear seat postcrash. Specifically, residential security video showed both the driver and passenger getting into the front seats prior to driving away from the residence. In addition, the EDR data showed active accelerator pedal inputs consistent with driver activity in the 5 seconds prior to the impact with the tree, and that the driver’s seat belt was connected
at the time of the crash. Finally, the steering wheel examination conducted by the NTSB Materials Laboratory indicated an impact to the upper left quadrant, consistent with the driver loading the steering wheel during a frontal crash.
The entire journey appears to have lasted less than a minute (9:07pm is mentioned both as the time the car left the driveway, and as the time of the crash), so it's unlikely the driver jumped into the back seat as a stunt while driving.
I'm confused by this description. Are they saying the impact caused the driver to be ejected from the front seat to the rear? I'm not sure physics works that way. Are they saying the driver left the front seat and climbed into the rear so that he could blame autopilot when the police arrived on scene?
Worth pointing out that the most recent Tesla crash freakout was similarly thinly sourced. We have one line in a police report saying AP was in use, and no statements from the driver nor telemetry from the vehicle. And that was enough to drive thousand-comment threads right here on HN.
(And yeah, I'm on record saying that it's extremely unlikely autopilot commanded the lane change in that Bay Bridge accident. It just doesn't work like that, the blinkers came on simultaneous with the motion. That's a human driver for sure.)
> This was the one where the media reported that because nobody was found in the front seat it must have been the fault of the self-driving feature.
And the fact that this was media disinformation won't matter much, because the damage to the Tesla brand and the idea of self driving cars in the public's collective consciousness has already been done, and won't easily be undone by these newly revealed facts.
"A lie makes it halfway around the world before the truth even gets its boots on."
"Excessive speed" and "alcohol intoxication" -- two common causes of human driver failure.
Other causes of human driver failure include driving while tired or exhausted, falling asleep at the wheel, driving while angry or upset, getting bored while driving, using the car's infotainment system while driving, using a mobile phone while driving, getting distracted by passengers, smoking pot while driving, being a recklessly immature teenager (or a grown-up idiot), lacking the bare minimum of driving skills that every driver on the road is supposed to have, and so on. The list of causes of human driver failure is long. There are a lot of horrifically dangerous human drivers -- look around you next time you're on the road.
Notably, machines are immune to all these human failure modes.
I'm looking forward to the day in which cars drive themselves well enough to rid the roads of so much dangerous human driving.
Roadways have failure modes of their own, pavement fails. Vehicles have parts that fail. Pedestrians are 1/6 of all motor vehicle fatalities. Dangerous weather happens at least once a year in most of the country. People are allowed to and still enjoy riding motorcycles.
Meanwhile, excessive speed and intoxication account for more than 50% of all fatalities. You can solve this problem without having to wait for an AI car, it's probably morally justified to do even if you think those AI cars will arrive "any minute now."
Finally.. I am not looking forward to the day when a parent puts their kids in an AI car and does not accompany them to their destination. People will invent new failure modes for you.
> “Excessive speed" and "alcohol intoxication" -- two common causes of human driver failure.
Amazingly it’s worse than that. He was also on pills that would react with the alcohol to make things worse.
I’ll say I’m very surprised by the areal photo in the report. I was under the impression the driver (man or machine) had gone a real distance at speed before the accident. Which did lead some credence in my mind a driver assist system had been helping it along if the driver was slow to react/inattentive.
He could have pulled out of that driveway, hit the gas, and it happened. The trip had barely begun. You could do this crash easy in a 1960s sedan, no ADAS needed.
We already have self driving that works and it's called a train. Instead of dumping money into self driving we should improve infrastructure and motivate more people to take trains.
Me too, but that's not this day. I hope it comes soon, but I think it's important that car companies promise and deliver in a rock solid fashion like Mercedes is doing instead of grandiose near future predictions that everyone knows won't be met. Not commenting on this incident in particular, just the general landscape of self driving.
I'm not sure why they don't explicitly write in the report that the driver may have been trapped in the back seat, and was unable to locate or use the mechanical release to escape. While that wouldn't be a contributing factor to the wreck, it would be a contributing factor the the driver's death, and important for Tesla and their passengers to know about.
The report states:
The frontal impact with the tree resulted in a power loss of the car’s 12-volt system, which runs the non-traction power systems. During normal operation, the front door latches operate electronically with the pull of the interior lever. In the event of a 12-volt system power loss, the interior front doors open as usual using the interior door handles. The rear doors also have both electronic and mechanical latches; however, mechanically opening the rear door during a power loss requires additional steps. According to the owner’s manual, during a loss of 12-volt system power, a rear-seated occupant must locate a small cutout in the carpet beneath the seat cushions and pull the mechanical release cable tab toward the center of the vehicle to manually open the rear door. Inspection of the door latches and locking hardware was limited by postcrash fire damage.
Edit: Must add that this assumes the front door was jammed and couldn't be opened mechanically.
Not saying this as a remark to you specifically, just a fun anecdote.
Someone totaled my Model 3 on my birthday a few years ago and my front door was smashed. I smelled smoke and thought I had a battery fire (turns out it was smoke from the airbags). I crawled to the back seat and tried to open my door but it wouldn't open due to power loss. I ended up breaking the window because in my state of mind, I couldn't remember that the manual door override was available. People always say there is a manual backup, but even if it works, it's really hard to remember that during a terrible accident where time is an issue. Luckily I carry a knife in my car with one of those window breaking points.
> I'm not sure why they don't explicitly write in the report that the driver may have been trapped in the back seat, and was unable to locate or use the mechanical release to escape.
Because the NTSB cannot establish that conclusively given the almost total lack of physical evidence remaining after the vehicle fire.
It does not mean that a vehicle defect does not actually exist from the NTSB's point-of-view. It just means that the NTSB lacks the physical evidence to conclude anything there.
(I personally find the necessary "additional steps" required for exiting the vehicle from the rear seats in the event of a power loss troubling.)
More broadly, this NTSB report is being misinterpreted by many here.
The "probable causes" (not "causes") and "lessons learned" (tellingly distinct from the NTSB's more traditional "safety recommendations") established in the final report are exactly tailored to the physical evidence and physical facts that did remain independently of Tesla (the company) - and no more.
The actual title of the HN submission is, in my view, inappropriately written by excluding the word "probable".
That sounds pretty terrifying in an emergency situation. What is the reason for the rear doors working differently? I can't picture someone referring to their user manual or looking under the seat to open the door when the car is on fire.
I don't think they have sufficient evidence one way or the other to draw that conclusion. The driver's cause of death was BFT with thermal damage and smoke inhalation; I don't know they have reason to believe he was conscious and trying to escape the vehicle.
Almost 43,000 people died in car-related deaths in the US in 2021 (and millions more worldwide). I hope we can stop focusing on the small number of deaths from this one car company that has self-driving features.
No consumer can purchase any vehicle that is capable of "self-driving" or "driving itself" today.
Tesla's vehicles are not capable of self-driving and, at all times, the human driver is driving the vehicle as both Autopilot and FSD Beta are partial automated driving systems that require a human driver fallback at all times.
The attentiveness required of the human driver with a partial automated driving system is equivalent (on a systems-level) to if the vehicle was not equipped with any automated driving system at all.
Same with fires. It seemed like every tesla fire was reported by the news for a while, meanwhile conventional vehicles are at least 5x more likely to start on fire.
> the car accelerated from 39 mph to a top speed of 67 mph 2 seconds before the final tree impact, which occurred at about 57 mph. The application of the accelerator pedal ranged from 8% to 98% during the 5 seconds of recorded data, and there was no evidence of braking.
My hot take is that it shouldn't be possible to accelerate that quickly unless you actually apply the pedal to 100%.
Also, I think the accelerator and brake pedals should have a different feel. Maybe they should be made of different materials so there is clear tactile feedback.
my hotter take is that there should be acceleration and velocity limits that are geo-gated and you shouldn't be able to go more than 30% over the speed limit in urban areas.
The whole idea of being drunk and getting in a car with the intention of driving as fast as possible seems so stupid that I've never in my life even had the smidgen of a thought to do it, even if especially drunk. People really do have different brains.
OT but the NTSB really wants to take the fun out of driving via technology (from the crash report)...
>Recently, the NTSB recommended that NHTSA require all new vehicles to be equipped with passive vehicle-integrated alcohol impairment detection systems, advanced driver monitoring systems, or a combination thereof, which are capable of preventing or limiting vehicle operation if driver impairment by alcohol is detected.
>A vehicle technology-based solution, such as intelligent speed adaptation (ISA), can reduce speeding. The NTSB has recommended that NHTSA incentivize passenger vehicle manufacturers and consumers to adopt ISA systems by, for example, including ISA in the New Car Assessment Program.
I didn't notice a closing sarcasm tag, so I'm assuming you're serious about your "fun" being infringed upon. Too bad. You have no right to drive intoxicated on public streets. Nor to drive at speeds over the posted speed limit. If you want to race, take up track racing. If people didn't act like fools and idiots while driving, there'd be no need for this type of technology.
> Although the driver’s seat was found vacant and the driver was found in the left rear seat, the available evidence suggests that the driver was seated in the driver’s seat at the time of the crash and moved into the rear seat postcrash. Specifically, residential security video showed both the driver and passenger getting into the front seats prior to driving away from the residence. In addition, the EDR data showed active accelerator pedal inputs consistent with driver activity in the 5 seconds prior to the impact with the tree, and that the driver’s seat belt was connected at the time of the crash. Finally, the steering wheel examination conducted by the NTSB Materials Laboratory indicated an impact to the upper left quadrant, consistent with the driver loading the steering wheel during a frontal crash.
If the driver was indeed wearing a seatbelt, how does that person end up in the left back seat in a frontal crash? Now I have questions about that seat belt.
Especially considering this comment in the report: In addition, the EDR
data showed active accelerator pedal inputs consistent with driver activity in the 5 seconds prior to the impact with the tree, and that the driver’s seat belt was connected at the time of the crash.
So there was a driver pushing the accelerator, wearing a seatbelt (assumption) and ending up in the rear seat behind the drivers' seat? Did the driver simply plug in the seatbelt without putting it on? Or did the driver survive the crash, get into the rear seat, and was unable to get out in time?
It probably was the later. It seems probable he was attempting to exit using the backdoor because he was unable to operate the front due to it's loss of function. Horrifying situation to visualize if that is indeed what happened.
I think they picked it up because it might have involved autopilot, the driver's seat was empty, and the data recorder had been destroyed in the crash.
But should the P100D or any other car noted for its sudden acceleration even be allowed to exist? We're putting GPS radios on electric scooters so they can't exceed 10MPH in school zones, meanwhile drunk bozos can tap 1000HP if they want.
[+] [-] sowbug|3 years ago|reply
Although the driver’s seat was found vacant and the driver was found in the left rear seat, the available evidence suggests that the driver was seated in the driver’s seat at the time of the crash and moved into the rear seat postcrash. Specifically, residential security video showed both the driver and passenger getting into the front seats prior to driving away from the residence. In addition, the EDR data showed active accelerator pedal inputs consistent with driver activity in the 5 seconds prior to the impact with the tree, and that the driver’s seat belt was connected at the time of the crash. Finally, the steering wheel examination conducted by the NTSB Materials Laboratory indicated an impact to the upper left quadrant, consistent with the driver loading the steering wheel during a frontal crash.
The entire journey appears to have lasted less than a minute (9:07pm is mentioned both as the time the car left the driveway, and as the time of the crash), so it's unlikely the driver jumped into the back seat as a stunt while driving.
https://www.ntsb.gov/investigations/AccidentReports/Reports/...
[+] [-] brewdad|3 years ago|reply
[+] [-] ajross|3 years ago|reply
(And yeah, I'm on record saying that it's extremely unlikely autopilot commanded the lane change in that Bay Bridge accident. It just doesn't work like that, the blinkers came on simultaneous with the motion. That's a human driver for sure.)
[+] [-] anonporridge|3 years ago|reply
And the fact that this was media disinformation won't matter much, because the damage to the Tesla brand and the idea of self driving cars in the public's collective consciousness has already been done, and won't easily be undone by these newly revealed facts.
"A lie makes it halfway around the world before the truth even gets its boots on."
[+] [-] cs702|3 years ago|reply
Other causes of human driver failure include driving while tired or exhausted, falling asleep at the wheel, driving while angry or upset, getting bored while driving, using the car's infotainment system while driving, using a mobile phone while driving, getting distracted by passengers, smoking pot while driving, being a recklessly immature teenager (or a grown-up idiot), lacking the bare minimum of driving skills that every driver on the road is supposed to have, and so on. The list of causes of human driver failure is long. There are a lot of horrifically dangerous human drivers -- look around you next time you're on the road.
Notably, machines are immune to all these human failure modes.
I'm looking forward to the day in which cars drive themselves well enough to rid the roads of so much dangerous human driving.
[+] [-] akira2501|3 years ago|reply
Meanwhile, excessive speed and intoxication account for more than 50% of all fatalities. You can solve this problem without having to wait for an AI car, it's probably morally justified to do even if you think those AI cars will arrive "any minute now."
Finally.. I am not looking forward to the day when a parent puts their kids in an AI car and does not accompany them to their destination. People will invent new failure modes for you.
[+] [-] MBCook|3 years ago|reply
Amazingly it’s worse than that. He was also on pills that would react with the alcohol to make things worse.
I’ll say I’m very surprised by the areal photo in the report. I was under the impression the driver (man or machine) had gone a real distance at speed before the accident. Which did lead some credence in my mind a driver assist system had been helping it along if the driver was slow to react/inattentive.
He could have pulled out of that driveway, hit the gas, and it happened. The trip had barely begun. You could do this crash easy in a 1960s sedan, no ADAS needed.
[+] [-] P_I_Staker|3 years ago|reply
A lot of our driving in unusual conditions require our experience and reasoning about the world.
[+] [-] commiepatrol|3 years ago|reply
[+] [-] SketchySeaBeast|3 years ago|reply
[+] [-] jo6gwb|3 years ago|reply
The report states:
The frontal impact with the tree resulted in a power loss of the car’s 12-volt system, which runs the non-traction power systems. During normal operation, the front door latches operate electronically with the pull of the interior lever. In the event of a 12-volt system power loss, the interior front doors open as usual using the interior door handles. The rear doors also have both electronic and mechanical latches; however, mechanically opening the rear door during a power loss requires additional steps. According to the owner’s manual, during a loss of 12-volt system power, a rear-seated occupant must locate a small cutout in the carpet beneath the seat cushions and pull the mechanical release cable tab toward the center of the vehicle to manually open the rear door. Inspection of the door latches and locking hardware was limited by postcrash fire damage.
Edit: Must add that this assumes the front door was jammed and couldn't be opened mechanically.
[+] [-] xeromal|3 years ago|reply
Someone totaled my Model 3 on my birthday a few years ago and my front door was smashed. I smelled smoke and thought I had a battery fire (turns out it was smoke from the airbags). I crawled to the back seat and tried to open my door but it wouldn't open due to power loss. I ended up breaking the window because in my state of mind, I couldn't remember that the manual door override was available. People always say there is a manual backup, but even if it works, it's really hard to remember that during a terrible accident where time is an issue. Luckily I carry a knife in my car with one of those window breaking points.
I can't imagine trying to do that drunk.
[+] [-] adamjcook|3 years ago|reply
Because the NTSB cannot establish that conclusively given the almost total lack of physical evidence remaining after the vehicle fire.
It does not mean that a vehicle defect does not actually exist from the NTSB's point-of-view. It just means that the NTSB lacks the physical evidence to conclude anything there.
(I personally find the necessary "additional steps" required for exiting the vehicle from the rear seats in the event of a power loss troubling.)
More broadly, this NTSB report is being misinterpreted by many here.
The "probable causes" (not "causes") and "lessons learned" (tellingly distinct from the NTSB's more traditional "safety recommendations") established in the final report are exactly tailored to the physical evidence and physical facts that did remain independently of Tesla (the company) - and no more.
The actual title of the HN submission is, in my view, inappropriately written by excluding the word "probable".
[+] [-] climb_stealth|3 years ago|reply
This reads like a terrible design.
[+] [-] mjevans|3 years ago|reply
[+] [-] shadowgovt|3 years ago|reply
[+] [-] kdamica|3 years ago|reply
Relevant: https://en.wikipedia.org/wiki/Man_bites_dog
[+] [-] adamjcook|3 years ago|reply
Tesla's vehicles are not capable of self-driving and, at all times, the human driver is driving the vehicle as both Autopilot and FSD Beta are partial automated driving systems that require a human driver fallback at all times.
The attentiveness required of the human driver with a partial automated driving system is equivalent (on a systems-level) to if the vehicle was not equipped with any automated driving system at all.
[+] [-] KyleBerezin|3 years ago|reply
[+] [-] diebeforei485|3 years ago|reply
My hot take is that it shouldn't be possible to accelerate that quickly unless you actually apply the pedal to 100%.
Also, I think the accelerator and brake pedals should have a different feel. Maybe they should be made of different materials so there is clear tactile feedback.
[+] [-] lamontcg|3 years ago|reply
[+] [-] mjfl|3 years ago|reply
[+] [-] hnburnsy|3 years ago|reply
>Recently, the NTSB recommended that NHTSA require all new vehicles to be equipped with passive vehicle-integrated alcohol impairment detection systems, advanced driver monitoring systems, or a combination thereof, which are capable of preventing or limiting vehicle operation if driver impairment by alcohol is detected.
>A vehicle technology-based solution, such as intelligent speed adaptation (ISA), can reduce speeding. The NTSB has recommended that NHTSA incentivize passenger vehicle manufacturers and consumers to adopt ISA systems by, for example, including ISA in the New Car Assessment Program.
[+] [-] greedo|3 years ago|reply
[+] [-] ninjagoo|3 years ago|reply
If the driver was indeed wearing a seatbelt, how does that person end up in the left back seat in a frontal crash? Now I have questions about that seat belt.
Especially considering this comment in the report: In addition, the EDR data showed active accelerator pedal inputs consistent with driver activity in the 5 seconds prior to the impact with the tree, and that the driver’s seat belt was connected at the time of the crash.
So there was a driver pushing the accelerator, wearing a seatbelt (assumption) and ending up in the rear seat behind the drivers' seat? Did the driver simply plug in the seatbelt without putting it on? Or did the driver survive the crash, get into the rear seat, and was unable to get out in time?
[+] [-] brentm|3 years ago|reply
[+] [-] tgsovlerkhgsel|3 years ago|reply
[+] [-] nashashmi|3 years ago|reply
[+] [-] sp332|3 years ago|reply
The full list is here https://www.ntsb.gov/investigations/Pages/Investigations.asp... You can see it's been a while since since they investigated a Tesla.
[+] [-] sidcool|3 years ago|reply
[+] [-] mensetmanusman|3 years ago|reply
[+] [-] jeffbee|3 years ago|reply
[+] [-] sokoloff|3 years ago|reply
[+] [-] aaron695|3 years ago|reply
Two people killed in fiery Tesla crash with no one driving https://news.ycombinator.com/item?id=26852399 736 comments
NTSB: Tesla Model S in Crash Couldn't Have Been Using Autopilot - https://news.ycombinator.com/item?id=27112010 199 comments
Cops “almost 99.9% sure” Tesla had no one at the wheel before deadly crash https://news.ycombinator.com/item?id=26871001 53 comments
No one was in driver’s seat in fatal Tesla crash https://news.ycombinator.com/item?id=26866754 85 comments
Texas police to demand Tesla crash data as Musk denies autopilot use https://news.ycombinator.com/item?id=26869962 18 comments
[+] [-] foobarbecue|3 years ago|reply
[+] [-] foobarbecue|3 years ago|reply
[+] [-] brookst|3 years ago|reply