Canada - FSD rollout

[TJVan]

I've had beta now for a couple of weeks and I try to use it on most drives, I've put a couple thousand kms on it. A big part has been training me to let it do it's thing, at the start I would take over too often. It took me a long time to start to trust it and it's really surprised me. Some situations are downright dangerous, but there aren't many. I've now had several drives with limited interventions and when I do mostly it's just to give a little power through the corners or cancel lane changes into parking lanes.
It's frustrating at times and amazing at others. It seems to have good days and bars days fun enough!
I'm just really happy to be part of this, honestly who thought we'd have self driving cars!!!!

 

[BobX]

Has anybody received the camera upgraded that is required for FSD beta? This is an upgrade for all Teslas that have the FSD package done at no charge by Tesla. Recently I upgraded to MCU2 and almost immediately received an invitation for the safety score program which I did. However, I am not sure that it is worthwhile pursuing the safety score since without the camera upgrade, no FSD beta is possible. The Service Center said they have no inventory and provided no timeline. It seems pointless to be pursing a safety score if Tesla isn't able to supply the upgraded cameras.

 

[Zas]

Has anybody received the camera upgraded that is required for FSD beta? This is an upgrade for all Teslas that have the FSD package done at no charge by Tesla. Recently I upgraded to MCU2 and almost immediately received an invitation for the safety score program which I did. However, I am not sure that it is worthwhile pursuing the safety score since without the camera upgrade, no FSD beta is possible. The Service Center said they have no inventory and provided no timeline. It seems pointless to be pursing a safety score if Tesla isn't able to supply the upgraded cameras.

Yup, I have a legacy X from 2017 and got the notification on the app of the camera upgrade ..scheduled the appointment for last week..but it was later postponed till the end of the month with them saying the parts hadn't arrived yet.

 

[thesteve]

Finally at 200km with 100 score. After multiple false fcw and two resets it’s so hard to keep 100. Last fcw because of a parked cement truck not even sure why. Anyway going to let it sit. Anyone have any ideas when more beta testers will be added?

 

[ecarfan]

I have a legacy X from 2017 and got the notification on the app of the camera upgrade

I have a mid-2017 X and paid for FSD (it was $3K at the time). I have that documented in a Tesla printout of my car’s configuration at purchase. I have not yet received a notification for a camera upgrade.

Curiously my Tesla app shows that I can upgrade to FSD for $6K even though I’ve already purchased it. I have not yet approached Tesla about getting FSD. It’s too early in the software development for me to find it useful. But later this year I’m going to ask for it.

 

[Zas]

I have a mid-2017 X and paid for FSD (it was $3K at the time). I have that documented in a Tesla printout of my car’s configuration at purchase. I have not yet received a notification for a camera upgrade.

Curiously my Tesla app shows that I can upgrade to FSD for $6K even though I’ve already purchased it. I have not yet approached Tesla about getting FSD. It’s too early in the software development for me to find it useful. But later this year I’m going to ask for it.

I would bring that to their attention now, as I see on the app that I have FSD...so if you have paid for it, it should show on yr app.

 

[pcons]

I have a mid-2017 X and paid for FSD (it was $3K at the time). I have that documented in a Tesla printout of my car’s configuration at purchase. I have not yet received a notification for a camera upgrade.

Curiously my Tesla app shows that I can upgrade to FSD for $6K even though I’ve already purchased it. I have not yet approached Tesla about getting FSD. It’s too early in the software development for me to find it useful. But later this year I’m going to ask for it.

I would bring that to their attention now, as I see on the app that I have FSD...so if you have paid for it, it should show on yr app.

You should definitely reach out to them now to have it corrected. It took a while for FSD to show up in my account despite having paid for it (after delivery). This was 3 years ago and there were no FSD options available (Navigate on Autopilot hadn't even been released yet if I recall...or maybe I had it because of EAP, but this was well before any HW3 upgrades were available so I wanted to make sure I didn't have any red tape in the way). I advise you to do the same....the 6k upgrade you see in the app is probably a bug because when you bought, you likely bought EAP, and the upgrade cost from EAP to FSD is not the same as the upgrade cost from AP to FSD (in current cars)....

 

[pcons]

For those of you still waiting...some light at the end of the 'Elon time' tunnel....who knows how long it will take, but if you don't have to suffer the pain of the safety score it might be worth it

Tesla's Full Self-Driving Beta will become available to all paying Americans and Canadians

Elon Musk is on record saying that the Full Self-Driving Beta feature will be an option to buy for every Tesla owner in the US or Canada later this year. Currently, not only do North American Tesla drivers need to pay the US$12,000 for the FSD software, but they also have to have a 99/100 Safety...

www.notebookcheck.net

 

[TrevTremaine]

How many?! I wonder how many Canadians are involved thus far... Tesla has now more than 100,000 people in its Full Self-Driving Beta program – is an accelerated rate of improvement coming?

 

[Darscot]

How many?! I wonder how many Canadians are involved thus far... Tesla has now more than 100,000 people in its Full Self-Driving Beta program – is an accelerated rate of improvement coming?

The number of users in the beta will not corelate to any increase in development. When your training a neural network data is key but it has to be good data. I don't see any way the data from the fleet is being filtered. I know in my driving sometimes intervene because the car is going to kill me, plenty of times I don't intervene but the car is a complete disaster. With no way for any of the data coming in to be filtered as good or bad its not useful. They can use some of the data to help with how the car sees. They can filter it by location and then compare it against the data they get from their lidar cars (I do find it funny they they use lidar as the standard for good). They need some standard for validation. You have to teach a neural network and this is like trying to teach someone math and never going over the answers. I am sure they have the employee fleet for validation but all this about the size of the beta is just marketing. It's not really a beta the whole point of a beta is feedback and there is virtually none coming in from the customer fleet.

 

[Vawlkus]

I have a mid-2017 X and paid for FSD (it was $3K at the time). I have that documented in a Tesla printout of my car’s configuration at purchase. I have not yet received a notification for a camera upgrade.

Curiously my Tesla app shows that I can upgrade to FSD for $6K even though I’ve already purchased it. I have not yet approached Tesla about getting FSD. It’s too early in the software development for me to find it useful. But later this year I’m going to ask for it.

My early 2017 does not show the camera upgrade yet (where was it located in the app?), but I do show FSD in my cars stats, both in the app and on the MCU.

 

[ecarfan]

My early 2017 does not show the camera upgrade yet (where was it located in the app?),

I was only saying that my Tesla app shows I can upgrade to FSD for $6K (though my purchase documentation shows I paid for it in 2017) not that my app shows a “camera upgrade” is available to me.

 

[drtimhill]

The number of users in the beta will not corelate to any increase in development. When your training a neural network data is key but it has to be good data. I don't see any way the data from the fleet is being filtered. I know in my driving sometimes intervene because the car is going to kill me, plenty of times I don't intervene but the car is a complete disaster. With no way for any of the data coming in to be filtered as good or bad its not useful. They can use some of the data to help with how the car sees. They can filter it by location and then compare it against the data they get from their lidar cars (I do find it funny they they use lidar as the standard for good). They need some standard for validation. You have to teach a neural network and this is like trying to teach someone math and never going over the answers. I am sure they have the employee fleet for validation but all this about the size of the beta is just marketing. It's not really a beta the whole point of a beta is feedback and there is virtually none coming in from the customer fleet.

Do you have any information to back this up? Tesla have stated quite clearly that they are using data from the beta testers.

 

[Darscot]

Do you have any information to back this up? Tesla have stated quite clearly that they are using data from the beta testers.

Only that I have trained a neural network and use machine learning professionally. It is not some magically thing that just learns by itself. You have to teach/train it to do what you want. You can't just throw information at it and expect it to learn and give you positive results. A funny story told to me about machine learning when I was a student. I don't know how true it is but it is a good analogy. There is a military contract and they are using machine learning to identify military targets. They go out in the morning set up all these civilian non valid targets. Then later on they go out and set these military valid targets. They use all this information to train the network and validate it against the data and is super accurate at identifying valid military targets. So they go out and start testing it in real world situations and it complete fails. Then they realize it never learned what a good or bad target was it just learned the time of day and lighting. In the morning with early light it thought everything was civilian, later past noon everything was military. They may be collecting some data from the beta testers but it is going to be very limited. You need data that is validated as good and bad. It is not so different than teaching anything you have to be able to say do this don't do this. I am sure they are using the data in some limited capacity. As I said the whole point of a beta is to have people test and validate the system. If you don't have the feedback its pointless. They are just letting customers run beta level software the customers are not actually part of the beta testing and validation.

 

[drtimhill]

As I said the whole point of a beta is to have people test and validate the system. If you don't have the feedback its pointless. They are just letting customers run beta level software the customers are not actually part of the beta testing and validation.

Indeed .. and they are. And they are using real beta testers as inputs to the training dataset. My point was you claimed that Teslas was not using beta tester input at all, and it was all just a marketing exercise. Explaining how NN training works (I'm aware of that) is not relevant. So, again, why do you think they are not using the beta testers feedback?

 

[Darscot]

Indeed .. and they are. And they are using real beta testers as inputs to the training dataset. My point was you claimed that Teslas was not using beta tester input at all, and it was all just a marketing exercise. Explaining how NN training works (I'm aware of that) is not relevant. So, again, why do you think they are not using the beta testers feedback?

I think you are just just splitting hairs based on language. I myself with this beta software have never been asked to provided feedback or been given any parameters around testing. I am sure they have employees and internal testing that does this job but the vast majority of customers running the "beta" are just driving around seeing what it does. There is no testing or validation going on and no feedback provided. The car makes mistakes constantly and intervention is purely based on the individual and how poor a driver they are willing to accept. Why do you believe that this is some how going to exponentially improve the software or is this just a case of because Elon said so?

 

[M109Rider]

I think you are just just splitting hairs based on language. I myself with this beta software have never been asked to provided feedback or been given any parameters around testing. I am sure they have employees and internal testing that does this job but the vast majority of customers running the "beta" are just driving around seeing what it does. There is no testing or validation going on and no feedback provided. The car makes mistakes constantly and intervention is purely based on the individual and how poor a driver they are willing to accept. Why do you believe that this is some how going to exponentially improve the software or is this just a case of because Elon said so?

I’m not sure I’m following your point, but the car is collecting data from every beta tester. It mostly collects data when the car is disengaged, or the driver hits the video button.
This data is used by the programming team to further the development.
Are you saying the data from beta testers cars are not being used ?

 

[Darscot]

I’m not sure I’m following your point, but the car is collecting data from every beta tester. It mostly collects data when the car is disengaged, or the driver hits the video button.
This data is used by the programming team to further the development.
Are you saying the data from beta testers cars are not being used ?

My only point was it is not going to massively accelerate the software like people are predicting. As a software engineer that uses this technology in a different field I do not see any actual evidence to support many of the claims being made. If you have ever been involved in an actual software beta or done any machine learning you can see there is a great deal marketing being thrown around here. I am not saying the data is not being used I am saying it will have very limited impact in the software development process. In my industry I hear the exact same claims made publicly all the time and everyone actually doing the work always has a good chuckle. Even the article posted is very skeptical.

 

[NeilOnCa]

My only point was it is not going to massively accelerate the software like people are predicting. As a software engineer that uses this technology in a different field I do not see any actual evidence to support many of the claims being made. If you have ever been involved in an actual software beta or done any machine learning you can see there is a great deal marketing being thrown around here. I am not saying the data is not being used I am saying it will have very limited impact in the software development process. In my industry I hear the exact same claims made publicly all the time and everyone actually doing the work always has a good chuckle. Even the article posted is very skeptical.

There are ways large number of testers can help in an automated manner. For example, Tesla could have 2, or even more, software versions running different ML models out with testers at same time. They do not need explicit feedback from testers. However, by simply looking at the interventions per 100 km driven across different versions, they can identity the better model among all. Here with a large number of testers covering more real world scenarios would definitely give this judgement more statistics confidence and significance. This is just one example that how beta testers can help.

 

[drtimhill]

I think you are just just splitting hairs based on language. I myself with this beta software have never been asked to provided feedback or been given any parameters around testing. I am sure they have employees and internal testing that does this job but the vast majority of customers running the "beta" are just driving around seeing what it does. There is no testing or validation going on and no feedback provided. The car makes mistakes constantly and intervention is purely based on the individual and how poor a driver they are willing to accept. Why do you believe that this is some how going to exponentially improve the software or is this just a case of because Elon said so?

First, of course testers have not been given parameters. And, if you understand NNs as you state, you should know why. If Tesla provide guidelines they may risk biasing the gathered data by causing drivers to focus in unexpected ways (as per your earlier example about time of day). As it is, the sample set may still have issues given all testers have undergone a screening process during signup.

Second, the car reports back after disengagements and any time you press the report button on the screen, so yes, Tesla are indeed gathering data from beta testers .. and they have asked for email feedback if that is also necessary. I dont know what percentage of beta testers are actually giving feedback (or even using the beta), but Tesla will simply (and probably have) adjust the beta test pool until they have adequate data coming from the overall pool.

And of course the car makes mistakes .. that's why its in beta. And I dont think anyone said anything about "exponential" improvements. I simply asked you to provide some justification for your claim that Tesla are not using data from the beta, and its just a marketing stunt. So far, you haven't done this, and have instead gone off on more or less irrelevant tangents. So, do you have anything to substantiate your claim, or were you just making stuff up (like you accuse Elon of doing)?