You can install our site as a web app on your iOS device by utilizing the Add to Home Screen feature in Safari. Please see this thread for more details on this.
Note: This feature may not be available in some browsers.
Hey all!
Looking for some feedback. Which layout would you prefer to see?
View attachment 684057
View attachment 684058
View attachment 684059
View attachment 684060
Comments like these tell me you don't understand the enormity of what v9 is accomplishing. They're rewritten the way the car understands the world from the ground up. That it's better out of the the gate than the previous instance tells you that it's a rock-solid foundation to build upon in the near future via machine learning.This is the HUGE improvement we were expecting? Really?
The graphics look better but not impressed by its driving.
I think I bought some snake oil.
Nice drive. I don’t think it was actually aiming for the white parked car, it was moving to avoid the oncoming car and probably wouldn’t have hit the white car, but you were 100% right to override. IMO, all beta testers should be overriding when they aren’t comfortable. On that note, in situations where the car is taking too long for your liking to complete a turn or maneuver through an intersection, I think you should be applying the accelerator yourself (with your hands on the wheel and ready to take over steering as well if necessary) and just note in the video that the go pedal was applied. From everything I’ve heard, the Tesla team treats the accelerator as an intervention and reviews any use of it during a Beta run just like any other intervention.Thanks all. In the meantime, here is a video with layout 3
My vote is for Layout 1Hey all!
Looking for some feedback. Which layout would you prefer to see?
View attachment 684057
View attachment 684058
View attachment 684059
View attachment 684060
What's under-appreciated about V9 is that Tesla has essentially solved vision and is now just working on gathering more diverse data and labeling everything they can.
Why I think they've solved vision:
1) 3D environment and objects within are extremely stable, and V9 only shows what it sees with very low latency. For example, it only needs to see a small slice of a car to know it's there, its orientation, how far it is, and how fast it's moving.
2) V9 sees brake lights and associates them with the correct cars. Even when it only sees 1 brake light, it correct assumes that both the car's brake lights are on.
3) Visualization shows that V9 can see cars in all their various orientations, smoothly.
4) V9 can see and makes visual inferences (is it moving? is parked? should I be concerned? etc.) on ~40+ objects at a time
Karpathy mentioned recently that he's narrowing in on a labeling workflow that consistently produces better and better results. V9 is a demonstration of how powerful this workflow and "4D" video labeling is.
Although we see a lot of V9 "fails," I think we're very close to an update where everything seems to click, and we get surprisingly good performance.
I may not understand the enormity of this grand v9 but I do see a car that drives in circles, eats speed bumps for lunch and can't make left turns. When you listen to how this was all talked up like some huge improvement it looks like a total dud to me. I do not see this ever being released in its present shape to the general public. It will be sometime next year before that happens if not the year after for sure.Comments like these tell me you don't understand the enormity of what v9 is accomplishing. They're rewritten the way the car understands the world from the ground up. That it's better out of the the gate than the previous instance tells you that it's a rock-solid foundation to build upon in the near future via machine learning.
Tesla vision is incredibly impressive. It's a great proof of concept.
It is _not_ ready.
It drives into featureless textured objects, and can't see curbs very well. Also it apparently can't see objects low to the ground like potted plants.
This is a big problem because we need to be able to trust FSD to not drive into solid objects, and TV hasn't proven that it is capable.
No accidents because either the Tesla driver or other vehicle intervened.Disagree. I think it's ready for supervised testing by non-Tesla affiliated people. 8.2 was a lot worse, and no accidents happened per Elon.
It may be a enormous accomplishment in terms of the transition from radar to vision but as an end user I don’t care much about that aspect of it. I’m looking purely at how it drives between 8.2 and 9. And not being able to see big objects at this point is not confidence inspiring.Comments like these tell me you don't understand the enormity of what v9 is accomplishing. They're rewritten the way the car understands the world from the ground up. That it's better out of the the gate than the previous instance tells you that it's a rock-solid foundation to build upon in the near future via machine learning.
Do you have any data proving this besides Elon tweeting it? It makes many of the same mistakes as before, which isn't really the way a full rewrite works. Why did they rewrite it?They're rewritten the way the car understands the world from the ground up.
As is the design that the Tesla driver intervene!No accidents because either the Tesla driver or other vehicle intervened.
Have to disagree here. At 4:25 you have correctly pointed out the near collision, which 100% would have been a high-speed collision. That alone negates anything positive about the drive. We cannot tolerate this, there is nothing impressive that can make up for such an incident.(..)
4:25 - car requires intervention on unprotected right turn. Was taking the turn wide with a car approaching which could have resulted in a collision.
All in all, mostly an impressive and incident free vid. Mostly confirms my thought that we might be able to get a wider Beta release by the end of the year, albeit one with user confirmation required for more complex maneuvers like unprotected turns.
(..)
I would be concerned of who gets to test these builds on the public roads. I believe in the right hands, the development team can get very valuable data and accelerate advancement that they wouldn't get otherwise. But in the wrong hands, someone will likely get hurt badly or killed. What if someone decided to just push the envelope and decided not to intervene at all? We all know that's not the right way to operate, but it will happen if wrong people got their hands on it.As is the design that the Tesla driver intervene!
People lobbing these handfuls of of crap hoping something sticks are certainly predictable.
Exactly. If the failures are fatal, it is unacceptable. I feel like I'm taking crazy pills: are people really saying that a device that is documented to cause fatal mistakes within 24 hours of a very limited release is ok to release to the public (??)Have to disagree here. At 4:25 you have correctly pointed out the near collision, which 100% would have been a high-speed collision. That alone negates anything positive about the drive. We cannot tolerate this, there is nothing impressive that can make up for such an incident.
Exactly. If the failures are fatal, it is unacceptable. I feel like I'm taking crazy pills: are people really saying that a device that is documented to cause fatal mistakes within 24 hours of a very limited release is ok to release to the public (??)
I know it's disappointing. But if you find yourself excusing these problems, please reconsider your intentions and potential subconscious biases. I'm rooting for Tesla too, but we have to stay grounded in reality.
Once FSD can prove that it won't drive straight into solid objects, and demonstrates a high level of safety when it fails (maybe dropping into a failsafe autosteer-only mode, or pulling over to the side, etc), then a wider release is more appropriate. I'm not saying it has to be perfect. It just has to not kill people.