Model 3 traveling at high speed failed to make a 90 degree turn. I don’t believe Autopilot makes turns, you need FSD for that.
intercst
5 Likes
Details on crashes and investigations include:
-
Federal Investigations: The National Highway Traffic Safety Administration (NHTSA) reviewed over 460 crashes tied to Tesla’s driver-assistance technology, resulting in more than 50 injuries and over a dozen confirmed deaths. Ongoing probes continue to scrutinize the systems following severe accidents and fatalities.
-
Safety Tracking: According to statistics tracked by industry databases such as Tesla Deaths, there are at least 772 total reported fatalities in Tesla vehicles, with 65 of those verified specifically as involving Autopilot or FSD
.
While advocates note that Tesla’s active safety features record fewer crashes per mile than the national average for human drivers, critics and agencies like the NHTSA highlight critical safety gaps and concerns over how the system detects obstacles in low-visibility environments. You can track ongoing investigations and federal recalls via the NHTSA Recalls & Investigations Portal
Wendy
3 Likes
Just put this in context, there are around 40,000 motor vehicle fatalities annually in the US. Those 65 cases involving auto pilot or FSD don’t necessarily assign fault to the system, only the claim it was active at the time. 63 of those cases involve the legacy autopilot and 2 are linked to FSD. Also, there is no discernible data yet on evasive action taken by FSD that has prevented auto accidents or performed better than a driver likely would have in a similar situation.
4 Likes
What is the ratio of Tesla miles driven to non-Tesla miles driven?
I’m not against self-driving cars. I think that in 20 years people will be marveling that humans were ever allowed to drive cars, given the horrible safety record.
But it’s early days.
Wendy
5 Likes
How many of those fatalities were a person sitting in their easy chair in their living room at home?
JimA
3 Likes
And this context too: U.S. National Highway Traffic Safety Administration (NHTSA) has tracked zero fatalities caused by Waymo fully driverless autonomous systems.
3 Likes
The data being tracked for Tesla only indicates the system was active at the time of the accident not necessarily that it was the proximate cause. This is the problem with the data. It doesn’t tell us much. Waymo also only operates in small defined lower speed areas. Robotaxi has caused zero fatalities as well.
2 Likes
Automakers are continually doing the arithmetic on how many customer deaths they can tolerate. The number is typically set at the level that maximizes excessive Executive Compensation.
One of the early tranches of Elon’s $1 Trillion Tesla pay package triggers when there are 10 million unsupervised FSD subscriptions in effect. He can afford to settle a lot of wrongful death lawsuits.
intercst
5 Likes
It’s not how he operates. He’s pushing to solve Autonomous driving. The fact that he is incentivized doesn’t mean he is going to take unnecessary risks. Tesla is a lightning rod for civil litigation. It cannot afford autonomous driving that risks human injury on even a small scale.
Not so.
Waymo has been operating with L4 autonomy on multiple highways for some time - even in cities like Miami.
They also cover a large part of the Bay area, most would not call that a small area.
Overall, a web search says Waymo operates with L4 autonomy on over 1400 square miles in 11 cities (about 4 million miles per week and growing).
5 Likes
That’s true.
Tesla used to say that their scaling of L4 autonomy was limited by regulators.
But on the last earnings call Tesla said that their scaling was limited by safety - because their L4 driver isn’t safe enough, they cannot scale it up with many more vehicles and mileage and instead must keep the deployment limited to small marketing/test demonstrations with few vehicles doing little mileage.
4 Likes
Not any longer. The State of Texas allows Robotaxi operators to “self-certify” L4 autonomous operations. The EPA has also given Tesla the green light for unlimited production of CyberCabs sans steering wheels and pedals. That was the only impediment to large scale implementation at the Federal Level.
As soon as Elon decides he can accept the lawsuit risk, he can deploy CyberCabs in any quantity he likes in Texas (and I’m pretty sure in Nevada, too.)
intercst
I stand corrected on your other points.
2 Likes
Waymo also suspended some operations over its handling of flooded roads.
Waymo expanded from about 5 to 11 cities this year, including some highways.
But they got some growing pains to go with that scaling: handling construction on highways and handling flooded roads.
Autonomous vehicle expansion is a region-by-region, scenario-by-scenario grind, it’s not a nationwide overnight software update, at least not today.
2 Likes
Dense urban environments present the steepest challenges for autonomous vehicles. San Francisco is the second most populated area in the United States. Humans communicate intentions with eye contact and subtle body language, its difficult to anticipate sudden jaywalking or darting cyclists; especially in San Francisco where pedestrians have the right of way. Pedestrians and cyclists in The City expect vehicles to yield right of way in most cases, and operate so.
2 Likes
I’m kind of marveling that humans were ever allowed to drive cars right now.
2 Likes
There is business risk too. If the public thinks Cybercabs are unsafe they won’t get in them. The Tesla robotaxi business is almost non-operational. There are only a handful of operating vehicles that are only available in daylight hours, in good weather, and in small geomapped areas.
That says to me Elon don’t think Tesla robotaxis are safe, at least not currently with HW4. However, there are HW4 Cybercabs rolling off the assembly lines right now with no place to go.
5 Likes
One line of thought: “If Tesla is mass-producing a vehicle model with a dedicated taxi form factor (maybe with no steering wheel!), then a massive rollout of L4 autonomous Cybercabs must be imminent!”
Not my line of thought.
Web search says there are 102 in a lot at the factory. And some more units scattered around the country for testing.
They certainly do present challenges. However, the lack of fatalities may be a function of the dedicated routes and the lack of high speeds or just simply the flipping between human and car controlled driving. The point is Waymo’s lack of fatalities or that fatalities occur in Tesla’s doesn’t provide sufficient evidence of a cause and effect relationship with FSD and autopilot engagement.
I would agree that Waymo may be farther along in solving autonomy. However, it has a scaling problem compared to Tesla. If Tesla solves it, in what is a low margin business, it seems the clear winner.
Further, there is zero evidence that Musk is simply waiting to a point where a certain number of human losses are acceptable. He is aggressive about timelines but it has never been at the expense of putting human’s at risk.
Ahem. “Full Self Driving” was released to the public where both occupants of that car and of any car which passed them became a beta-tester. Tesla resisted for years the calls to rename the option because it wasn’t, in fact, “full” or “self driving”.
Finally, and belatedly, California authorities stepped in and required a name change, which Tesla grudgingly made: “Full Self Driving - Supervised.” Really rolls right off the tongue, eh?
Additionally, they have fought to keep accident records sealed, claiming “proprietary technology”, even though doing so would let safety regulators in other domains make decisions that could aid the industry in the quest.
10 Likes