- Joined
- Jul 13, 2023
- Messages
- 1,533
- Reaction score
- 2,567
- Location
- Tahoe City, CA
- Cars
- Toyota Highlander
He's not joking. Unfortunately....
Surely, you’re joking.
...
He's not joking. Unfortunately....
Surely, you’re joking.
...
Calling it “full self driving” and still not having changed the name in the US is solid evidence that “prioritizing safety” was not their first concern.He's not joking. Unfortunately.
Look at tesla. They didn't release hands free until v12.5 Before V12 they literally were in beta you needed a safety score of 100 to even get access to the software.Evidence, please.
The rate of improvement for robotaxi is evident in how fast they went from no driving, to unsupervised, to the ability to handle 1000 cars in a single city. The rate of improvement comes from the fact that waymo has year's old issues (after doing unsupervised for 10 years) that they seem to not be able to solve. It's because for years they simply buried these Issues behind HD maps.Tesla is not “improving at a faster rate.” If you want to make this assertion, you’ll need to justify it with proof.
People like making this comparison but it’s not actually valid, as accidents are not equally comparable. Do robotaxis make mistakes and get into crashes humans wouldn’t? If so, how much of that is acceptable?
Without the answer to that, simply counting the number of crashes does not make a case, as drivers are not equally skilled.
For the *average* driver, it is almost certainly better. For *good* drivers, it may not be yet.
And the problem, of course, is everyone thinks they’re a good driver.![]()
And you could get a safety score of 100 trivially by faking it. Ask me how I know.Look at tesla. They didn't release hands free until v12.5 Before V12 they literally were in beta you needed a safety score of 100 to even get access to the software.
No, they did not release them because they didn’t work yet. Not because of safety.They did not release parking spot to parking spot driving until v14. They did not release robotaxi features, minimum risk maneuvers, etc. until v14 because if you enable these things with v12, people start to fall asleep using a dangerous software.
That is definitely not what Tesla did, and not what it does. It does not care, in my experience, where you are looking.You make sure your software shuts off enough that nobody abuses it.
Yes.They had for 5 years a technology that could drive around, and they had an autopark system. Do you honestly think it would have been hard to build a system that would have done a parking spot to parking spot drive?
You are incorrect, and it is not as simple as that.Of course not. You just drive around and stitch it with the autopark software.
Cool. That’s because, I reiterate: it didn’t work.Tesla never did that. In fact they never even used the already working autopark feature at any time while using FSD. They redid the parking system from scratch.
That is not the case. They added them slowly because it took them time to build.Who beat them to it? The chinese. The chinese were doing parking spot to parking spot years before tesla did. They even have systems you can record drives in parking garages and the car can navigate them. Tesla never did that. They focused on adding the features slowly and safely.
These are all assumptions based on zero data or evidence. Please provide some if you are going to lob accusations.Now rivian will not have a safety level close to FSD so what they going to do? They are going to shortcut to feature parity with FSD because demo drives are the only way to market an inferior system.
They literally just hosted an event where they made a car do a flip with compressed air. They exclusively care about superficial, surface level performance, because it’s what gets publicity.Tesla never cared about superficial, surface level performance where you can do a parking spot to parking spot drive with no intervention.
Okay. I’m willing to agree to disagree.They cared about accident avoidance, fast reaction time, and fixing the few features they had because that's what scales to unsupervised driving.
Yes but at least the made an attempt to gatekeep it to safe usersAnd you could get a safety score of 100 trivially by faking it. Ask me how I know.![]()
No. They could have made the car park. Back when FSD 12 came out the autoparking feature worked well and the car would get into a parking lot, stop near the spaces and you could click on one of them with autopark. It's easy to make the system drive parking spot to parking spot if that is your only purpose. They just did not do it because there is no point to doing that. You'd have to fix it later and there's no point pretending you have something that you don't have.No, they did not release them because they didn’t work yet. Not because of safety.
False. The reason they added features slowly is a combination of two factors. #1 is safety because you don't add features only to have to redo them the right way.Cool. That’s because, I reiterate: it didn’t work.
That is not the case. They added them slowly because it took them time to build.
Rivian released universal hands free which works on any road, has zero accident avoidance, zero awareness of other cars, and does not stop for stop signs, traffic lights, etc.These are all assumptions based on zero data or evidence. Please provide some if you are going to lob accusations.
Not for FSD. Tesla never cared to implement features that make the car get better. They stopped fixing mapping issues after a certain point, because fixing mapping issues artificially improves performance where your core driving abilities do not improve.They literally just hosted an event where they made a car do a flip with compressed air. They exclusively care about superficial, surface level performance, because it’s what gets publicity.
Okay. I’m willing to agree to disagree.
Are you?![]()
Lol. Doing it fast is not evidence that they are doing it right or well.The rate of improvement for robotaxi is evident in how fast they went from no driving, to unsupervised, to the ability to handle 1000 cars in a single city.
I agree. A friend of mine died in an accident like this when their stopped car was hit by a truck from behind.
Classic case. Look how many people believe they would have avoided this. No chance.
You don't scale if you don't have a tech that works. Tesla is not stupid. They know what issues they have. The reason you don't scale is one death gets you shut down for good.Lol. Doing it fast is not evidence that they are doing it right or well.
Guess the answer was no.Okay. I’m willing to agree to disagree.
Are you?![]()
As of October 2025, there have been sixty-five reported fatalities, fifty-four of which NHTSA investigations or expert testimony later verified and two that NHTSA's Office of Defect Investigations determined as happening during the engagement of Full Self-Driving (FSD) after 2022.You don't scale if you don't have a tech that works. Tesla is not stupid. They know what issues they have. The reason you don't scale is one death gets you shut down for good.
As of October 2025, there have been sixty-five reported fatalities, fifty-four of which NHTSA investigations or expert testimony later verified and two that NHTSA's Office of Defect Investigations determined as happening during the engagement of Full Self-Driving (FSD) after 2022.
The end.
Calling other members ignorant is not an appropriate tone for this forum. Its especially bad when you are saying something that is very close to what they are saying. Any further rude remarks will be deleted.That shows your ignorance. As of 2026, there are only 3 deaths that have been caused by FSD directly running into something that was not some other external factor (like someone stepping on the accelerator pedal, disengaging FSD and crashing, etc.). tesladeaths.com is a great source for this
There are 13 deaths per billion miles.
Of these two deaths, one of them was back when tesla ran autopilot on freeways and they hit a pedestrian on the freeway obscured by sun glare. This is not an FSD failure but simply running older neural networks designed for the old AP1 computer. The level of sun glare in that video was nothing
The 1st death that happened, the guy set his speed to 70mph on a mixed speed highway (back when you could do that) when the speed had dropped to 45mph and a semi truck pulled out in front of him|
The most recent they hit an elderly man walking across the freeway at 70mph at night, the car dodged him but the corner still clipped and he died. This was likely unavoidable.
Across many billions of miles, there is basically no evidence of any version of FSD post E2E neural networks (12.5.4 where most people had access to FSD) ever having a serious defect with seeing stationary vehicles at night.
Waymo is involved in 3 deaths already, which shows that deaths at this scale are just unavoidable. The most recent death they may have even hit someone who was not already dead.
Congrats, you’re officially the first person on this forum I have ever actually wanted to ignore, and I have an unusually high bar, as most others here (I suspect) will attest to.That shows your ignorance. As of 2026, there are only 3 deaths that have been caused by FSD directly running into something that was not some other external factor (like someone stepping on the accelerator pedal, disengaging FSD and crashing, etc.). tesladeaths.com is a great source for this
There are 13 deaths per billion miles.
Of these two deaths, one of them was back when tesla ran autopilot on freeways and they hit a pedestrian on the freeway obscured by sun glare. This is not an FSD failure but simply running older neural networks designed for the old AP1 computer. The level of sun glare in that video was nothing
The 1st death that happened, the guy set his speed to 70mph on a mixed speed highway (back when you could do that) when the speed had dropped to 45mph and a semi truck pulled out in front of him|
The most recent they hit an elderly man walking across the freeway at 70mph at night, the car dodged him but the corner still clipped and he died. This was likely unavoidable.
Across many billions of miles, there is basically no evidence of any version of FSD post E2E neural networks (12.5.4 where most people had access to FSD) ever having a serious defect with seeing stationary vehicles at night.
Waymo is involved in 3 deaths already, which shows that deaths at this scale are just unavoidable. The most recent death they may have even hit someone who was not already dead.