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

Wifi uploads have stopped

This site may earn commission on affiliate links.
Anyone else notice the autopilot nest brain has stopped requesting videos?
My M3s used to upload at least 1GB/day, sometimes 2GB.
They haven't uploaded a single video for the past 30 days (just little 20MB pings).
Has AP stopped "learning" while they re-write this code for HW3?
I would hope that since they have halted AP learning, they are close to releasing the "3d video foundational rewrite" soon, otherwise we will not see any improvement to the current 'gen 2' system for some time.

Or, maybe I'm just not driving in areas that trigger campaign uploads? Anyone else tracking this and see uploads continuing?
 
Anyone else notice the autopilot nest brain has stopped requesting videos?
My M3s used to upload at least 1GB/day, sometimes 2GB.
They haven't uploaded a single video for the past 30 days (just little 20MB pings).
Has AP stopped "learning" while they re-write this code for HW3?
I would hope that since they have halted AP learning, they are close to releasing the "3d video foundational rewrite" soon, otherwise we will not see any improvement to the current 'gen 2' system for some time.

Or, maybe I'm just not driving in areas that trigger campaign uploads? Anyone else tracking this and see uploads continuing?
Mine is still uploading.
B580AC51-22FD-48C6-8E58-3E1BF9758DDC.jpeg
 
Last 2 posts obviously jokes.
But for real, anyone?
Doing a little more research, GreenTheOnly is also noticing same thing / lack of firmware changes since end 2019, but his suspicion is an update on the UI side not AP side, I think.
Mine wasn’t a joke. That’s what my car has uploaded over the past week. Although I’m not sure what’s it’s uploading, seems normal to me.
 
  • Like
Reactions: coconutboy84
From what I’ve read, your car only uploads camera data if it matches criteria that Tesla is currently looking for. If there has indeed been a lull in uploads to Tesla from the fleet it could be just that they’re currently looking for fringe cases that many would not encounter. It could be anything really, no way to tell just based on how much data our cars have been uploading recently.
 
From what I’ve read, your car only uploads camera data if it matches criteria that Tesla is currently looking for. If there has indeed been a lull in uploads to Tesla from the fleet it could be just that they’re currently looking for fringe cases that many would not encounter. It could be anything really, no way to tell just based on how much data our cars have been uploading recently.
Definitely; they only upload what tesla is looking for at the moment, called “campaigns”. But my point is, I haven't triggered it in over 30 days, just wondering if anyone has at all. At first, I thought the same as you, they must be looking for only edge cases. But noone has uploaded anything for over a month as far as I can tell. So I think they have literally paused AP learning for now.
 
Is that a joke?

Yeah maybe, but your auto wipers, auto high beams, and AP and FSD have all been "constantly improved" over a number of years by the data collected from the "massive Tesla fleet" being fed into the Tesla "neural network."

I don't know about you, but if they still can't even figure out after 3 model years of data going into this amazing neural network when it's time to wipe the windshield, I think they need to try a new marketing gimmick.
 
Last edited:
Yeah maybe, but your auto wipers, auto high beams, and AP and FSD have all been "constantly improved" over a number of years by the data collected from the "massive Tesla fleet" being fed into the Tesla "neural network."

I don't know about you, but if they still can't even figure out after 3 model years of data going into this amazing neural network when it's time to wipe the windshield, I think they need to try a new marketing gimmick.
Are you using double quotes to imply there hasn't been any improvement, there's isn't a massive fleet of vehicles, and that tesla doesn't use neural network computation systems and models for their vehicles' automation features?

I'm not sure if you're trolling, uninformed, or just one who dons a tinfoil hat. If it's either of the latter two, you might consider reading up a bit. Or better yet, spend an hour going through a tutorial to see a basic example of how that particular tool in the software "toolbelt" can be applied to solve problems. Autonomous driving isn't a trivial problem to solve. It's pretty silly to dismiss a particular approach to solving it as a "marketing gimmick" just because there aren't leaps-and-bounds improvements like we'd all love to see.

Deep Learning with PyTorch: A 60 Minute Blitz — PyTorch Tutorials 1.4.0 documentation
 
  • Like
Reactions: Matsayz
Are you using double quotes to imply there hasn't been any improvement, there's isn't a massive fleet of vehicles, and that tesla doesn't use neural network computation systems and models for their vehicles' automation features?

I'm not sure if you're trolling, uninformed, or just one who dons a tinfoil hat. If it's either of the latter two, you might consider reading up a bit. Or better yet, spend an hour going through a tutorial to see a basic example of how that particular tool in the software "toolbelt" can be applied to solve problems. Autonomous driving isn't a trivial problem to solve. It's pretty silly to dismiss a particular approach to solving it as a "marketing gimmick" just because there aren't leaps-and-bounds improvements like we'd all love to see.

Deep Learning with PyTorch: A 60 Minute Blitz — PyTorch Tutorials 1.4.0 documentation

I don't need to read about it, I actually own a Tesla, and whatever they are doing isn't working. I heard they are completely rewriting FSD, so apparently my observations are spot on.
 
  • Like
Reactions: afadeev and bijan
I don't need to read about it, I actually own a Tesla, and whatever they are doing isn't working. I heard they are completely rewriting FSD, so apparently my observations are spot on.
Yeah. People that don’t require reading usually do have the most informed opinions. I love your insight that software shouldn’t ever require iteration or design changes. Reading is for peasants, for sure.
 
  • Disagree
  • Like
Reactions: Fernand and afadeev
Mentioned in a HW3 upgrade thread - I think Tesla is pushing for first release of FSD by end of Q1. That would allow the company to recognize the FSD revenue from every car with FSD and HW3.


Do you mean for the post March 2019 buyers? (since those would be the only ones who had HW3 from the factory- but were also the ones "only" promised stoplight/stopsign and in-city-driving both starting at level 2)

Because they can't recognize the revenue for the pre-march-19 buyers until they've got the much fuller FSD they promised those folks done and 0% chance that happens by end of Q1.

(honestly given the total re-write going on I'd be pretty surprised if they even got the in-city nav at L2 done by then)
 
TLLMRRJ said:
I don't know about you, but if they still can't even figure out after 3 model years of data going into this amazing neural network when it's time to wipe the windshield, I think they need to try a new marketing gimmick.

Are you using double quotes to imply there hasn't been any improvement, there's isn't a massive fleet of vehicles, and that tesla doesn't use neural network computation systems and models for their vehicles' automation features?

I think he means that claiming to use a tool, and producing positive results using said tool, are two different things.
The former does not guarantee the latter.


Autonomous driving isn't a trivial problem to solve.

Indeed.
But not for Tesla, which claims to have achieved "feature complete" FSD in 4Q 2019, and has partially recognized FSD revenue in that quarter for that achievement.

You need to either embrace the hype in full and worship the FSD / auto-wipe / auto-high beams perfection, or accept a doze of reality and skepticism.
Feel free to choose either path.

a