-
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
Knightshade
Well-Known Member
Naah... midnight in California is still hours away! plenty of time to "officially" release an update
Well that ship came and went...
mikes_fsd
Banned
It would have been dumb on Tesla's part to recognize any revenue in Q1.
Right now, they can get away with "bad quarter because of Covid-19". I for one would take (pull forward if needed) some more charges in to Q1 and push out recognizing revenue to Q2 or Q3.
Just a thought.
Right now, they can get away with "bad quarter because of Covid-19". I for one would take (pull forward if needed) some more charges in to Q1 and push out recognizing revenue to Q2 or Q3.
Just a thought.
Knightshade
Well-Known Member
Yup- makes sense to me.
Take the hit you know you're gonna take anyway from low deliveries due to the end of Q push getting nerfed by the virus... then hope you can make Q2 look really good if the lockdowns lift before June.
Makes sense, but it will be hard for them to have a good Q2 as well....
I'd hate for them to wait to release until Q3... just for these reasons
mikes_fsd
Banned
I think they could make Q2 look much better if they pull some of the CapEx expenses into Q1 and recognize a bigger chunk of AP/FSD revenues.
diplomat33
Average guy who loves autonomous vehicles
“I am hoping to roll out traffic lights and stops to wide US release in a few weeks and probably a worldwide release in Q3 (so many variations in each country!). Very important to make sure this is done right.” Elon
Tesla is aiming to release Autopilot for intersections in just weeks in the US, will take months for other markets - Electrek
Looks like Tesla is trying to get traffic light and stop sign response out in the US pretty soon.
Tesla is aiming to release Autopilot for intersections in just weeks in the US, will take months for other markets - Electrek
Looks like Tesla is trying to get traffic light and stop sign response out in the US pretty soon.
Knightshade
Well-Known Member
To be fair- it's not only "looked like that" since last year- they literally promised it (PLUS automatic driving on city streets) by end of 2019 and that didn't happen either.
diplomat33
Average guy who loves autonomous vehicles
True but the difference is that this time, stopping at stop lights is actually in early access.
Knightshade
Well-Known Member
I mean, we know they're working on these things- if they keep throwing out target date after target date, ONE of em will eventually be right
@Knightshade @NHK X @diplomat33
So I figured that this wide release of this 'intersection stopping' would come much quicker than the wide release of smart summon, because it is far easer to to test and validate in shadow mode than smart summon is. With smart summon, that can only be validated (and get only get test data) from Early access people, and early access people using it often. And a lot of cases would come up that they would not have expected.
With intersections... they can get training data and validation data from the whole fleet everywhere and they don't need to be doing anything special, just going through intersections. So they already have over a year of training and validation data... and they can very easily collect more from both the EAP people and data from the whole fleet. Say they train a new model... they could roll it out in shadow mode, and have millions or tens of millions of intersections validated in a very short time period.
However, of course if they simply are having trouble training a model that meets the performance they need.... this could mean it could take a long long time. But I don't think that's the case.
However, again, seeing verygreens most recent comments on twitter about this early access update, it seems it is a very early release and seems they may have lots to do still.
I do think late June is a realistic possibility though. (who knows though, could be sooner, or much much later)
So I figured that this wide release of this 'intersection stopping' would come much quicker than the wide release of smart summon, because it is far easer to to test and validate in shadow mode than smart summon is. With smart summon, that can only be validated (and get only get test data) from Early access people, and early access people using it often. And a lot of cases would come up that they would not have expected.
With intersections... they can get training data and validation data from the whole fleet everywhere and they don't need to be doing anything special, just going through intersections. So they already have over a year of training and validation data... and they can very easily collect more from both the EAP people and data from the whole fleet. Say they train a new model... they could roll it out in shadow mode, and have millions or tens of millions of intersections validated in a very short time period.
However, of course if they simply are having trouble training a model that meets the performance they need.... this could mean it could take a long long time. But I don't think that's the case.
However, again, seeing verygreens most recent comments on twitter about this early access update, it seems it is a very early release and seems they may have lots to do still.
I do think late June is a realistic possibility though. (who knows though, could be sooner, or much much later)
diplomat33
Average guy who loves autonomous vehicles
Yeah, Green's reference to getting rear ended makes me think that Tesla still needs to optimize the actual stopping part. As I see it, traffic light response requires 3 pieces:
1) Detect the presence of a traffic light
2) Detect if the light is red or yellow or green.
3) If the light is red or yellow, slow down in a safe manner in order to stop at the stop line. When light is green, accelerate in a safe way.
So it is more than just training the NN with data to recognize red lights. #3 is tricky. If you brake too hard, you will stop short and risk getting rear ended. But if you brake too lightly, you can overshoot the stop line and risk getting hit by cross traffic. So you need to brake just right. Tesla may still need to be working on getting that part right.
June seems fair.
Agreed, Green described them as ghost brake type slow downs, would be interesting to see what the issue was. Or instead is it too much precaution on Tesla’s part while waiting for a response from driver?
While Green has said shadow mode doesn’t exists, I think he did mention this has been around for quite some time.
I guess risk profile is much greater for this feature, although it seems it should have come out sooner. Either way, I do hope it comes out better baked than smart summon did.
Knightshade
Well-Known Member
Sure- but the RISK OF HARM of a mistake at 3 mph in a private parking lot where the failsafe is just stopping dead when unsure is vastly vastly lower than blowing a red light or stop sign at speed through an intersection on a public road (or even of mistakenly slamming the brakes at speed on a public road at a green light).
My impression wasn't that it "didn't exist" just that it wasn't nearly as fancy, complex, and advanced as Tesla would prefer you think it is.... deep dive here-
green on Twitter
also 3B) If it's yellow, decide if braking to stop for it, or accelerating to get through it, is the correct decision based on distance from light, current speed, time of yellow
True. I did consider this.
However the risk of blowing a red light already exists... I think this only decreases that risk even if it isn't performing optimally. And I don't think they would implement it to slam on brakes... they probably set thresholds and alert if can't brake on time or not. The RISK of HARM that is brought up by the potential to reduce speed and come to a stop at a green light is quite low IMO. However, I agree the RISK of HARM for smart summon was also quite low.
Elon is also promising robo taxis this year/next year (allowing for some COVID slippage) so that means he must be planning on having all the features done without requiring interventions. If Tesla is being allowed partial revenue recognition, then he can declare "feature complete" and argue that some extremely fraction of a percentage should be withheld for government approval.
Knightshade
Well-Known Member
He was also planning for them to drive cross country without interventions 3 years ago.
He's... optimistic.
Well.. no.
For the pre-march-2019 folks he's delivered roughly 0% of feature complete FSD prior to the stoplight thing, and you'd have a hard time convincing any court (or the IRS) the stoplight thing is more than a small fraction of what's promised.
Now- for the post-march-2019 folks there's 7 listed features... in theory 5 of em were already largely 'delivered' before now, and this is the 6th... so if they wanted to recognize 6/7ths of all that they could probably get away with that.
What features were in pre-March 2019 FSD list that Tesla removed at that point? I don't think anyone pre-March 2019 is getting anything different from what March 2019 onward will get (i.e. everyone is going forward with the new h/w platform and re-written FSD software). Are you thinking there is a "bait and switch" where what pre-March 2019 FSD purchasers thought they were getting is different than what post March 2019 purchasers are getting?
Similar threads
