- Joined
- Jul 13, 2023
- Messages
- 1,525
- Reaction score
- 2,554
- Location
- Tahoe City, CA
- Cars
- Toyota Highlander
He's not joking. Unfortunately.
Out of Spec roadtrips Gravity and DDP2 has some issues
- Thread starter Bdr100
- Start date
- Joined
- Nov 19, 2021
- Messages
- 16,312
- Reaction score
- 22,425
- Location
- Cupertino, CA
- Cars
- Air DE-P, Gravity DE
- Air DE Number
- 241
- Gravity DE Number
- 013
- Referral Code
- Q1BTN5Y3
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.
That’s pretty hard to argue against, since the Netherlands has forced them to change the name, and they’ve still chosen not to in the US.
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.
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. You make sure your software shuts off enough that nobody abuses it.
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? Of course not. You just drive around and stitch it with the autopark software.
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.
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.
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.
Tesla never cared about superficial, surface level performance where you can do a parking spot to parking spot drive with no intervention. They cared about accident avoidance, fast reaction time, and fixing the few features they had because that's what scales to unsupervised driving.
Last edited:
Also worth mentioning that FSD v14 as it is in consumer cars is arguably an "illegal" software at this point. There is nothing within the spirit of how SAE J3016 is written that does not make FSD a level 4 software.
At that point it is defined as a level 4 test software which is illegal to use in many states.
The way the SAE J3016 was written is level 2 softwares were intended to be lane centering and then they would become level 3 as the technology matured.
it was never intended that a "self driving" feature ever be anything other than level 4 which matures to unsupervised and then level 5 as the odd becomes nearly unlimited
At that point it is defined as a level 4 test software which is illegal to use in many states.
The way the SAE J3016 was written is level 2 softwares were intended to be lane centering and then they would become level 3 as the technology matured.
it was never intended that a "self driving" feature ever be anything other than level 4 which matures to unsupervised and then level 5 as the odd becomes nearly unlimited
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.
You can just see in the latest FSD releases how quickly tesla is able to address simple issues by pushing out to 100s of thousands of cars and verifying the issues are fixed.
Waymo on the other hand is not running freeway for exactly the same issue they had 5 months ago.
In this video among many other recent ones you can see that waymo does not see speed bumps and slow down for them. They premap all of them. What is it about this that gives you any confidence they can avoid debris on the freeway if they can't slow down for a speed bump without premapping them?
have you been in the new waymo zeekr vehicle? it drives so poorly and it's embarrassing that they ever released that to the public. After this many years they remove a few sensors and the car drives like that? It's so jerky. How does that give you any confidence that they are going to drive on 2 new models, with entirely different car design and sensor placement.
Last edited:
- Joined
- Sep 27, 2026
- Messages
- 140
- Reaction score
- 73
- Cars
- Lucid Air Pure
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.
Classic case. Look how many people believe they would have avoided this. No chance.
- Joined
- Nov 19, 2021
- Messages
- 16,312
- Reaction score
- 22,425
- Location
- Cupertino, CA
- Cars
- Air DE-P, Gravity DE
- Air DE Number
- 241
- Gravity DE Number
- 013
- Referral Code
- Q1BTN5Y3
And you could get a safety score of 100 trivially by faking it. Ask me how I know.
No, they did not release them because they didn’t work yet. Not because of safety.
That is definitely not what Tesla did, and not what it does. It does not care, in my experience, where you are looking.
Yes.
You are incorrect, and it is not as simple as that.
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.
These are all assumptions based on zero data or evidence. Please provide some if you are going to lob accusations.
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?
- Joined
- Nov 19, 2021
- Messages
- 16,312
- Reaction score
- 22,425
- Location
- Cupertino, CA
- Cars
- Air DE-P, Gravity DE
- Air DE Number
- 241
- Gravity DE Number
- 013
- Referral Code
- Q1BTN5Y3
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.
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.
#2 is because it is very illegal. FSD was always pushing the boundaries of legal because they were developing a level 4 test software under the guise of level 2
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.
When did tesla do that? Never.
You can read the room all over reddit. There are people who believe that gap in autonomy between rivian and tesla is small. They will see a software with feature parity as more/less "the same thing."
During the first autonomy day back in january, rivian faked their demo drives by only turning right on green and turning left at a dotted line intersection. When the car failed to do its few tasks and turned right on red, the car turned into traffic as if it was not there. They also took over any time a pedestrian or cross traffic crossed the street. Basically the car did lane centering and cruise control disguised like the car could drive around. Zero understanding of traffic and it merged into the turning lane not through an understanding of traffic but because you can enter the turning lane for free. And you expect a company that does this and lies about autonomy progress to care about safety?
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?
There was an FSD beta tracker which showed that various versions of FSD v12-14 showed no increase in safety, while the edge case safety improved by orders of magnitude every release. Shows you how useless that website really was
Lol. Doing it fast is not evidence that they are doing it right or well.
- Joined
- Sep 27, 2026
- Messages
- 140
- Reaction score
- 73
- Cars
- Lucid Air Pure
I agree. A friend of mine died in an accident like this when their stopped car was hit by a truck from behind.