Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Poll: when will FSD Beta v9.x go fleet-wide?

When will FSD Beta v9.x go fleet-wide?


  • Total voters
    111
  • Poll closed .

shrineofchance

she/her, they/them
Feb 10, 2021
82
88
Canada
You have 7 days to vote. You can change your vote until then. Votes will be displayed publicly.

LRYoU4t.jpg
 

shrineofchance

she/her, they/them
Feb 10, 2021
82
88
Canada
I have had the same thought about visual driver monitoring being a pre-requisite for releasing FSD Beta (although I sincerely hope they don't offer hands-free driving initially and instead wait for a few billion miles of data to come in).

But I have no idea how far along Tesla is in getting visual driver monitoring ready for public release; no more than I know when FSD Beta v9.x will be ready for public release.

Fantastic video from verygreen for those who don't want to go digging through the other thread:

 
Last edited:

diplomat33

Well-Known Member
Aug 3, 2017
7,014
8,041
Terre Haute, IN USA
Did Elon actually say that the FSD City Streets Beta would be available fleet wide?

I think it is implied in "the button". Elon said that the button would allow anyone (with FSD) to download FSD Beta if they want to and agree to the terms. So that would be the same as making FSD Beta fleet wide, I would think.
 

diplomat33

Well-Known Member
Aug 3, 2017
7,014
8,041
Terre Haute, IN USA
I have had the same thought about visual driver monitoring being a pre-requisite for releasing FSD Beta (although I sincerely hope they don't offer hands-free driving initially and instead wait for a few billion miles of data to come in).

Yeah, I think camera driver monitoring is basically a pre-requisite for FSD Beta going wide. Tesla can't release something like FSD Beta with just the wheel torque. It would be too risky because of the likelihood of an accident when someone uses FSD Beta and is not paying attention. When testing is limited, Tesla can keep the risk low and kick people out if they are not paying attention. But when FSD Beta goes out to hundreds of thousands of drivers, the odds of someone not paying attention at the wrong time and getting into an accident is way higher. And if Tesla pairs the camera driver monitoring with the torque for extra redundancy, they can better enforce driver attention and keep risk low.

I do think there is great potential too in the camera driver monitoring to make a great product. When FSD Beta gets even better and we can go hands-free, imagine how great that will be. Yes, the system would still be L2 but having a hands-free system that can drive you in most cases from A to B, on city streets and highways, with no disengagements, would be fantastic. And it would definitely put Tesla ahead of the pack in terms of consumer car driver assist systems. The key will be safety and making sure that the system does not get abused because it would just take a couple accidents to shut down the whole thing.
 

EVRider-FL

Member
Aug 18, 2015
194
95
South Florida
I think it is implied in "the button". Elon said that the button would allow anyone (with FSD) to download FSD Beta if they want to and agree to the terms. So that would be the same as making FSD Beta fleet wide, I would think.

Elon also said that he would increase the number of beta participants from the current 1,000 to about 10,000, and then eventually scale out to more people. I interpreted that to mean "the button" will only be available to the additional owners who are invited to participate. But I hope to be proven wrong. 🙂
 

Microterf

Member
Sep 18, 2020
169
198
Ohio
Elon also said that he would increase the number of beta participants from the current 1,000 to about 10,000, and then eventually scale out to more people. I interpreted that to mean "the button" will only be available to the additional owners who are invited to participate. But I hope to be proven wrong. 🙂
If only there was a tool he could use to communicate to everyone with some more details
 
  • Funny
Reactions: HelixSpiral

Durwood101

Member
Jun 26, 2020
42
35
Escondido
If a cabin camera is a required thing, guess all the early adopters would basically be without any FSD Beta or otherwise without the ability for Tesla to detect whether a drivers eyes were focused on the road. It seems a little difficult to think Tesla would retro install any cameras inside exsiting fleet for Model S and Model X owners.......and although I do think its the right thing to do, something tells me going that route pretty much excludes any pre-date 2021 Model X or S owners. I cant imagine anyone who paid for FSD being asked to wait any longer because of the camera needing to be in the car...that would be a bitter pill to swallow for alot of us.
 
  • Informative
Reactions: pilotSteve

shrineofchance

she/her, they/them
Feb 10, 2021
82
88
Canada
It seems a little difficult to think Tesla would retro install any cameras inside exsiting fleet for Model S and Model X owners
My guess is that this is exactly what will happen. People install dashcam accessories all the time; a cabin camera seems like one of the least difficult things to retrofit onto a car. (Compare to the amount of trouble Tesla is going through to retrofit HW2 cars with the HW3 computer.)
 
  • Like
Reactions: Hbrink

shrineofchance

she/her, they/them
Feb 10, 2021
82
88
Canada
AwOrILc.png


At first, it is puzzling why Tesla would remove radar. Especially so since it's been reported that Tesla was developing its own in-house radar hardware.

After giving it a bit of thought, the best guess I can come up is that the new neural network architecture in FSD Beta v9.0 is incompatible with camera-radar sensor fusion for two reasons:

First, it processes all camera input as 360-degree panoramic video frames, and Tesla doesn't have 360-degree radar.

Second, and more importantly, the ratio of deep learning research published on processing camera images vs. processing radar signals is probably like 10,000:1 or 100,000:1. A hard research problem that Tesla has been working on is how to get a neural net to compare sequential video frames, thereby incorporating temporal information into its predictions (rather than treat every still image as belonging to its own universe. It may just now be possible to do this with camera images, but I'm fairly certain it's still impossible to do this with radar signals.

In other words: my guess is that Tesla has made significant progress in computer vision, and that progress relies on a neural network architecture that is incompatible with camera-radar sensor fusion.
 

AlanSubie4Life

Efficiency Obsessed Member
Oct 22, 2018
9,294
11,112
San Diego
To be clear on the "funny" - I just think it's funny that a single word response from Elon, for something that may not happen, results in so much speculation about the reasons for this "change". But, to speculate: My thinking is that they aren't incorporating radar for City Streets right now because it's too complicated to deal with at the moment (possibly for some of the reasons you mention). In other words, I agree with what you said about the current state of affairs, though longer term saying radar is “incompatible” seems like too strong a statement. I guess we'll see if their progress in computer vision remains incompatible with radar inputs long term. I don't think there's necessarily any reason it can't be incorporated in some situations, but it's also not really necessary; certainly for City Streets it seems less useful than on the highway.
 
Last edited:

diplomat33

Well-Known Member
Aug 3, 2017
7,014
8,041
Terre Haute, IN USA
AwOrILc.png


At first, it is puzzling why Tesla would remove radar. Especially so since it's been reported that Tesla was developing its own in-house radar hardware.

After giving it a bit of thought, the best guess I can come up is that the new neural network architecture in FSD Beta v9.0 is incompatible with camera-radar sensor fusion for two reasons:

First, it processes all camera input as 360-degree panoramic video frames, and Tesla doesn't have 360-degree radar.

Second, and more importantly, the ratio of deep learning research published on processing camera images vs. processing radar signals is probably like 10,000:1 or 100,000:1. A hard research problem that Tesla has been working on is how to get a neural net to compare sequential video frames, thereby incorporating temporal information into its predictions (rather than treat every still image as belonging to its own universe. It may just now be possible to do this with camera images, but I'm fairly certain it's still impossible to do this with radar signals.

In other words: my guess is that Tesla has made significant progress in computer vision, and that progress relies on a neural network architecture that is incompatible with camera-radar sensor fusion.

Elon's tweet about removing production radar is incredibly ballsy IMO. If FSD Beta gets into accident and it is revealed that the camera vision made a mistake, and radar would have prevented the accident, there will be loud cries for Tesla to put radar back into cars.
 
  • Like
Reactions: Dan D.

About Us

Formed in 2006, Tesla Motors Club (TMC) was the first independent online Tesla community. Today it remains the largest and most dynamic community of Tesla enthusiasts. Learn more.

Do you value your experience at TMC? Consider becoming a Supporting Member of Tesla Motors Club. As a thank you for your contribution, you'll get nearly no ads in the Community and Groups sections. Additional perks are available depending on the level of contribution. Please visit the Account Upgrades page for more details.


SUPPORT TMC
Top