Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
The article says Luminar reported $2 million revenue from Tesla. I googled the cost of lidar used in autonomous vehicles and numbers between $100 to $500 per piece come up. Even if $2 million is spent on the high end, say $500 per piece units that makes 4000 units which I think is more than what could be used for panel alignment and other things. Just sharing my math... read into it as you wish
And that was only in the first three months of this year.
 
  • Like
Reactions: OldManCan
On my drive this morning was excitedly daydreaming to see NO NAGS and thinking how far "Deep Eyes" must have come ........and then my "Deep Rain" wipers started dry wiping. Nothing like a jolt back into reality. 😲 🤣
Totally worth it to save a few bucks on rain sensors, even at the cost of stealing CPU cycles from the FSD computer.

Yes especially entering a city. When the speed limit changes from 55 down to 35 it seems to stay at 55 way too long enough time to get a ticket is there anyway of adjusting this so that it will slow down faster?
Turn off Auto Max. Then you can set the max speed via automatic offset (% or absolute offset from the current speed limit) and manual real time adjustments via steering wheel buttons/scrollwheel. Even on Auto Max, the UI responds to the scrollwheel inputs but does not make it obvious what the changes are to the max speed (if any).

And as a reminder, similar things had been said when they added back a HD radar to Model S/X, but nothing had been done with it (while FSD has improved drastically since then).
Aren't they being used for reporting ground truth data back to Tesla for training.
 
The article says Luminar reported $2 million revenue from Tesla. I googled the cost of lidar used in autonomous vehicles and numbers between $100 to $500 per piece come up. Even if $2 million is spent on the high end, say $500 per piece units that makes 4000 units which I think is more than what could be used for panel alignment and other things. Just sharing my math... read into it as you wish
Others reported the Luminar units are $1k each and Tesla data gathering vehicles have been captured using 8 of them, so that's only enough for about 250 vehicles.

Edit as a sanity check, Verge which broke the story seems to be claiming each unit cost about $1000.
 
Last edited:
Others reported the Luminar units are $10k each and Tesla data gathering vehicles have been captured using 8 of them, so that's only enough for about 250 vehicles.

I found this other news article on the topic which quote some figures from Luminar


"According to Luminar, individual lidar sensors cost around $1,000, including software. Did Tesla buy 2,100 lidars for its vehicles? "
 
  • Like
Reactions: OldManCan
I found this other news article on the topic which quote some figures from Luminar


"According to Luminar, individual lidar sensors cost around $1,000, including software. Did Tesla buy 2,100 lidars for its vehicles? "
Yes, I just linked that in edit, actually the math is correct when using $1k not $10k, I remembered incorrectly. At $1k that is only 250 vehicles if each uses 8 of them.
 
  • Like
Reactions: OldManCan
Yes, I just linked that in edit, actually the math is correct when using $1k not $10k, I remembered incorrectly. At $1k that is only 250 vehicles if each uses 8 of them.
If you use 8 lidars - do you then ditch Vision completely? You probably need fewer than that if you leverage the existing Vision technology.

Wikipedia says that Waymo's lidars cost $7,500 per vehicle in 2017.
 
My guess is that Tesla was using the LIDAR to validate the Occupancy Network output. The Occupancy Network itself was in theory trained in an unsupervised fashion using only the "pixel-flow" from multiple camera views. But given it was outputting volumetric predictions, the only way its accuracy could be measured would be with a sensor like LIDAR.

That would also fit with Elon's tweet that they're no longer using them for the reason they originally bought them. The Occupancy Network has been completely replaced with V12's end-to-end approach.
 
Version 13 will probably be unsupervised and will be a 3rd option that costs more
While FSD supervised replaces enhanced autopilot
I didn’t purchase FSD Supervised. I purchased FSD Beta with the understanding the beta tag would be removed when complete and not replaced with another tag.
 
Not sure I agree with Chuck’s conclusion here. They’ve obviously never gated a release on the success of this turn.

It appears they are trying to model ideal driver behavior for new training data. I could see arms moving on the steering wheel and different driving behavior.

On the plus they are more closely approaching the stop sign and more quickly creeping for better view. On the down side they all appeared to be instructed to turn right before turning left even when unneeded. I expect that over trained crap to once again adversely influence almost all left turns. FSD - the more things change the more it stays the same.
 
I didn’t purchase FSD Supervised. I purchased FSD Beta with the understanding the beta tag would be removed when complete and not replaced with another tag.
It doesn't matter what they name it, or call it, what you purchased is the same thing: A L2 driver assistance feature. (Unless you bought it early on when they described it as a L4 feature.)
 
  • Like
Reactions: Matias and Pdubs
Specifically, it seems to have started overreacting to the "stop sign ahead" signs, abruptly slowing down immediately before or immediately after them, then continuing slowly for a while until reaching the actual stop sign. Has anyone else observed a similar behavior?
Yes, all the time. I think there was quite a discussion about this upthread.

Concerning getting rid of nags completely, I'd be surprised. Nags are a good thing, and stop my mind from wandering.