Daniel in SD
(supervised)
The only thing they have to prevent is running a red light or stop sign. It seems like your idea does nothing to make that less likely.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla looking to recognize more FSD revenue for 2020 Q1
- Thread starter Mardak
- Start date
What about traffic lights with no Green Light is flashing yellow goes to solid yellow to red back to flashing yellow Say the light is red it goes to green but there is a School bus stopped on the shoulder with it's lights flashing. You are in the far left lane on a multi lane road. When the car get to the school bus does it stop or will it keep going
Daniel in SD
(supervised)
I doubt they will do anything for school buses in the "Recognize and respond to traffic lights and stop signs" feature. That will have to wait until the "automatic driving on city streets" feature is released later this year.
I think the only safe way to implement stop light and traffic light detection is to have it only stop if the car detects that you're going to run the red light or stop sign. That way people won't try to rely on it.
Knightshade
Well-Known Member
The main downside I see in your idea is that makes it REALLY REALLY look like a safety feature entirely, not a self-driving feature at all... cue the arguments about tesla saying they won't charge for safety features
My assumption would be that the system would initially be overly cautious: treat any situation where there is no explicit green light as a stop.
This would suck in some parts of the world where a flashing red or amber light is used for "proceed with caution", but you have to start somewhere.
This would suck in some parts of the world where a flashing red or amber light is used for "proceed with caution", but you have to start somewhere.
Daniel in SD
(supervised)
I think that would require HD mapping and Tesla doesn't use HD mapping!
Seriously though that might be reliable enough. If you know where every stop light and stop sign is you can use that as a prior in your decision about whether to stop. If the vision system can't determine the status of the light with extremely high confidence then it could stop.
Yeah, I know. Also, it would be even more upsetting to people than emergency lane departure assist which Tesla has either turned off or gotten to work so well that no one notices it any more.
Ha, but hey do use maps, every intersection should have a traffic control device..Paranoid mode would be stop for all intersections unless green is highly 99% detected, or driver presses accelerator (in which case override if stop/red is highly detected).I think that would require HD mapping and Tesla doesn't use HD mapping!
Daniel in SD
(supervised)
It would also increase sales of rear bumper components!
The more I think about it the less likely I think this feature will be released. I was wrong about Smart Summon though...
It is a heck of a headache isn't it???It would also increase sales of rear bumper components!
The more I think about it the less likely I think this feature will be released. I was wrong about Smart Summon though...
I sort of like being on the sidelines.
Daniel in SD
(supervised)
Me too, I didn't buy FSD.
Maybe they can realize some revenue by fixing the ping-ponging they seem have introduced but only for FSD buyers.
Oh, I ordered FSD... but my Cybertruck isn't coming for at least a year, so I've got that going for me.
My sideline comment was in reference to not having to make or implement the plan. Don't even want to armchair quarterback/ coach on this (much)
Same here, just got HW3 installed last Saturday and I've appreciated a much different autopilot experience.
It feels like the car processes faster and makes fast adjustments. When there's an abrupt change in road direction, it no longer overshoot followed by a ping pong movement. It now just seems to dampen the overshoot and slowly drift back to the center of the lane.
Only time I've experienced slight ping pinging since HW3 retrofit is when lanes appear/disappear in an a repeated pattern.
It feels like the car processes faster and makes fast adjustments. When there's an abrupt change in road direction, it no longer overshoot followed by a ping pong movement. It now just seems to dampen the overshoot and slowly drift back to the center of the lane.
Only time I've experienced slight ping pinging since HW3 retrofit is when lanes appear/disappear in an a repeated pattern.
Looks like Tesla is more aggressively collecting data for stop signs and/or traffic lights. With the maps/navigation update (5.5GB) yesterday, I tested red light and stop sign warnings, and the car uploaded 0.5GB of data from just a little bit of driving, which seems likely to include video/images.
This was with 2020.12, and there's some warning behavior differences from when stop sign warning first became available. It looks like the car now needs to be going fast enough to trigger the warning, e.g., 20mph seems to more consistently trigger than 15mph. (Whereas with the initial release of stop sign warning, I could drop the speed to 5mph and very slowly roll towards the line and trigger the warning.) This has a side benefit of those who reduce the set speed to 0mph for stop lines won't need to hear the warning beeps anymore.
This was with 2020.12, and there's some warning behavior differences from when stop sign warning first became available. It looks like the car now needs to be going fast enough to trigger the warning, e.g., 20mph seems to more consistently trigger than 15mph. (Whereas with the initial release of stop sign warning, I could drop the speed to 5mph and very slowly roll towards the line and trigger the warning.) This has a side benefit of those who reduce the set speed to 0mph for stop lines won't need to hear the warning beeps anymore.
diplomat33
Average guy who loves autonomous vehicles
That's good news because Tesla definitely need to be using more data if they want to accelerate FSD. After all, the whole advantage that Tesla is supposed to have over the competition is the large data that they get from hundreds of thousands of cars driving all over the US and the world, picking up edge cases etc... So they better actually use that data.
willow_hiller
Well-Known Member
Most people here are also following other threads about the topic, but I thought maybe we could discuss the revenue implications of the release of the new FSD features here?
FSD stopping at stoplights is in early access, and videos are already appearing of it: The Tesla Show on Twitter
Does Tesla recognize FSD revenue on a car-by-car basis (until the individual customer receives their update)? Or if the update has started to roll out by March 31, could this additional revenue be all recognized under Q1 by virtue of the software being available?
FSD stopping at stoplights is in early access, and videos are already appearing of it: The Tesla Show on Twitter
Does Tesla recognize FSD revenue on a car-by-car basis (until the individual customer receives their update)? Or if the update has started to roll out by March 31, could this additional revenue be all recognized under Q1 by virtue of the software being available?
I am wondering before this quarter / this update..... what portion of the deferred revenue are they recognizing already ? I know they started recognizing some portion when they released smart summon, and then maybe some again when they released FSD visualizer? and maybe there was some other reasons I'm forgetting about?
so like right now many people have paid $5-7k for FSD.... is Tesla recognizing 50% of that? less?
Knightshade
Well-Known Member
For pre-march-2019 EAP buyers they're recognizing $0.00 of it... .because they've delivered 0 new functionality to those buyers.
IIRC folks reported they did recognize about 30 million for smart summon in Q3/19, but that would only be for 3/19 and later buyers who got no summon features at all without paying for FSD.
FWIW according to Teslas 10k (page 73) for end of 2019 we see:
Deferred revenue on auto sales (this is supercharger access, internet connectivity, AP and FSD features, OTA updates, etc)-
Start of 2018- 476 million
Added in 2018- 532 million
Recognized in 2018 112 million
End of 2018- 883 million
Start of 2019- 883 million
Added in 2019- 880 million
Recognized- 300 million
End of 2019- 1472 million
So in 2018 they recognized about 11.1% of total for the year... in 2019 it was about 17% of total for the year... (I'm not sure that's really a super useful measure, but since they don't break it out in more detail it's what we've got)
A lot less than 50% at any rate overall
At least from 2019 Q3 earnings call, Zachary Kirkhorn said $30 million was recognized for Smart Summon in US, and "As we expand Smart Summon to additional markets and release new features, we will continue to recognize additional deferred revenue."
But still unclear if it's based on vehicles installing the update or just that the update is available. But recognizing revenue definitely seems to require having functionality available to the region.
Similar threads
