mikes_fsd
Banned
Latch on to the "important" stuff, why don't we.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD rewrite will go out on Oct 20 to limited beta
- Thread starter diplomat33
- Start date
Cheburashka
Active Member
Why are they wasting their time with testing Lidar if they've already determined that it's a crutch?
It's the nature of science. You can't claim, you have to prove. It makes perfect sense to me.
mikes_fsd
Banned
Look I like lightbulbs... Let's switch subjects.
But to answer your question any way:
For several reasons.
When your lead AI guy (Karpathy) publicly says "we see that the psudo-lidar approach with cameras is producing good depth accuracy... and the gap between camera approach and lidar is quickly shrinking" you will probably want to quantify that.
Or to validate the actual measurements of the camera NN's to the cm level accuracy of lidar?
Or 1000 other reasons to test.
At the end of the day, they are not embarrassed or ashamed to have a Tesla roll down the road with lidar equipment in plain site, so I would like to see why they would be embarrassed by the fact that they are looking at better radar.
Maybe using it as a ground truth
I just want the car to be able to park itself with me inside, for starters. Not much point in navigating the parking lot if you can’t park in the furthest spot from the store between two lines with no adjacent vehicles (everyone knows you never park adjacent to other vehicles
We need a “Tesla Parking Mode” to park far away from the store, between lines, in a space, calculated using urinal rules to avoid door dings...
M109Rider
Active Member
So the “Beta” version of FSD is out “one” day, and you’re already stating there is little evidence they will deliver reliable FSD.
LOL.
Are you always this much of a pessimistic guy in life, or is this just what it looks like, and you really just don’t “WANT” Tesla to deliver reliable FSD.
AlanSubie4Life
Efficiency Obsessed Member
There's a whole continuum of possibilities here. For example (and this is just ONE possibility of many), Tesla might want to start software development for a radar sensor that currently does not exist, or is available in very low volumes only, but they know (or are defining!) the specs for.
They could take a LIDAR unit (and maybe it is a LIDAR unit with memory, etc., or other features they want - totally speculating), and hack it to emulate the "better" radar unit, and software development can continue. They're extremely similar sensors in many respects, though obviously LIDAR has better resolution, etc., etc., and they aren't the same.
Obviously this would be a development path separate from the main production path, but it seems a perfectly good reason why Tesla could have LIDAR mules out there with no long term plan to actually use LIDAR.
We already know they're going to improve the sensors (for reasons you have mentioned yourself, amongst others), so I'm not sure why this is so controversial. Do you want to wait for the new sensor to be available to start developing software that takes advantage of the new capabilities? That would be silly.
In the end, we're just people on the internets, and there are a lot of perfectly reasonable reasons Tesla would be muling LIDAR with no intention of using LIDAR long term, and you can even interpret these reasons as being "pro-Elon," if you would like.
One thing is certain; cars that are produced in the next 1, 2, or 3 years will be more capable than cars produced today. There's really no doubt. Will it mean they are the only ones that can do L4/L5? I have no idea.
mikes_fsd
Banned
Was there supposed to be doubt of this?
I know some folks expect technology to stand still.... but that is a silly expectation and with Tesla that is just downright dumb.
The idea was that the core approach (camera's, radar and ultrasonics) are the Tesla way.
Not -
- only 720p cameras produced in 2016
- radar from Bosch or some other vendor
- ultrasonics that can only see 2 lanes over.
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
You seemed confused about the disagrees involving "the desire of people at Tesla for LIDAR" and subsequent remarks.
I was just explaining one way in which it could be perfectly reasonable for people at Tesla to want LIDAR installed in Tesla vehicles.
It's no big deal, and I'm sure engineers at Tesla are very familiar with asking for and having LIDAR added to Teslas.
It just doesn't really matter. At all. It's just a forward-looking sensor anyway so obviously it's not going to make or break anything for L4/L5 by itself. But as discussed, it may very well be important to get LIDAR installed on the test vehicles (might make or break the development schedule!).
And to make it perfectly clear, obviously I'm not talking about production vehicles here.
Anyway, looks like we crossed streams on my prior post.
Last edited:
It will be awesome to see a dog in the driver’s seat with its head stuck out the window. ‘If dogs could drive’ by Gary Larson
AlanSubie4Life
Efficiency Obsessed Member
I didn't get the impression anyone made that claim, but maybe I misinterpreted or missed something.
PR? That would be Elon Musk Twitter & Third Row Twitter acolytes
mikes_fsd
Banned
Yes, sorry, there is about 5 pages worth of context here. You would have to go back all the way to my post with green's tweet about the hi-res radar and specifically the comment that he tweeted in that post.
"The vendor appears to be Arbe Robotix and their website has a bunch of material that many of you would find resembling output of a sensor that Tesla cannot include due to perceived loss of face."
A bit like how Fox News has 1 out of 20 shows with a Progressive bent, so they can claim they are balanced
Daniel in SD
(supervised)
I suspect they do have a strong bias against using Lidar due to a perceived loss of face. Running mules with Lidar makes perfect sense to train a vision based depth neural net, I don't see why they would be embarrassed by that. I don't think anyone said they would be embarrassed by using high resolution Radar. They might not want people to know about any sensor improvements because they want to sell the cars they are making today not the ones they'll make in the future. The only contradiction I see is that high resolution radar is a crutch in exactly the same way Lidar is crutch. I'm pro-crutches, hopping around on one leg is a much worse solution (and unsafe!).
AlanSubie4Life
Efficiency Obsessed Member
Yeah, I saw that (I think I was the first to post about that Twitter thread). So I had the benefit of reading that tweet without any surrounding context/opinions. At the time, I didn't interpret it to mean that Tesla was "embarrassed by the fact that they are looking at better radar." (And I still don't!) I viewed it as an editorial comment that "Tesla doesn't want to include LIDAR in production vehicles, for loss of face reasons." I don't necessarily agree with that being the reason (there are plenty of other perfectly good reasons), but I agree that they don't want to include LIDAR in production vehicles.
mikes_fsd
Banned
Daniel in SD
(supervised)
The sensor that Tesla cannot include due to a perceived loss of face is Lidar, not Radar.
mikes_fsd
Banned
Similar threads
- Question
