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

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

This site may earn commission on affiliate links.
Like the guy below :D

Dana.png


Joe Fomenko on Twitter
 

Attachments

  • D.png
    D.png
    58.9 KB · Views: 73
Or a section of NN trained to detect the aspect change of an approaching pothole edge vs the normal planar surface. Vision might work better against water filled potholes also...
Not saying how big or how much programming, but a NN could easily outperform a human for pothole avoidance and do it safer. It would take into account nearby vehicles, speed, road conditions etc to decide if to avoid it at all, or what adjustments to make.
 
That is ~not~ what I said, and you know that...

I just noticed that several on this board seemed to be excited about this particular analyst's opinion on the stock, and I am just trying to point out that this particular guy may not be the one you want to hitch your wagon to...

He was also a big fan of Solar City's prospects back in the day too... Just sayin....
David Tamberbutthole of GS has the same track record tho
 
32 3s in Düsseldorf Germany, first deliveries Feb 19th

Dropbox - M3s in DUS - Simplify your life

VINs:

195866, 195880, 195890, 195906, 196599, 200085, 200296, 200829, 200854, 200857, 201052, 201054, 205984, 205987, 206004, 206041, 206081, 207243, 207311, 207313, 207435, 207467, 207475, 207496, 207535, 207600, 207652, 207760, 207777, 207853, 207923, 207981


file.php

file.php

TSLAQ take on that: Tesla is hiding unsold cars now in Europe
 
That is ~not~ what I said, and you know that...

I just noticed that several on this board seemed to be excited about this particular analyst's opinion on the stock, and I am just trying to point out that this particular guy may not be the one you want to hitch your wagon to...

He was also a big fan of Solar City's prospects back in the day too... Just sayin....

I like to think that people learn from their mistakes, as appears to be the case here.
 
  • Like
Reactions: Thumper
A little won't do. It has to be 99.999...% correct.
FSD can't take any risks, so it has to be sure.

I'm not entirely disregarding the possibility that a solution can be found. I just kept thinking that often times we drive based on memory of what we saw when the snow was not there... For example, a stop sign that is entirely covered by a sticky snow. If you drive there often, you know it's there. Location of turn lanes, etc.
Just took my daughter for her 2nd ever drive to a parking lot after the fresh snow. To do some doughnuts and have some fun. So, there are curbs in certain places around the lot and in the middle, which I can recognize by the slightly raised snow levels or trees/poles sticking out. Couple of times she drove right over those curbs, b/c she's a new driver and not used to paying attention to those things and staying away. I just saw how this can be a challenge to an FSD - if it is not sure that it's going to leave the road and end up in the ditch, then that risk should not be taken.

The other concern is related to cameras being obstructed as @bdy0627 mentioned.

I noticed that during a snowfall I start getting sounds related to an obstruction ahead and it is actually showing it on the screen ahead of the car when nothing is there, so that kind of confusion is not helpful for achieving 99.999%. Maybe this can be improved. But an ice buildup would certainly put cameras out of commission. Need to add heaters there.
Just as you have a memory of the road a FSD car will have the memory of every other FSD car that shares the same hardware platform. It will know what happened to a car 1 minute in front. The memory will be compared with what the car sees and positioning data before every single manoeuvre.

Main cameras are behind the windscreen so they will be kept clean at all times. Radars see through fog and snow. Cameras see through the dark. If a camera is covered with snow the car can let you know so you can wipe it before setting off. Having direct heating around some cameras might help to.

Snow will be problematic but far from impossible to deal with just as it is for us mere humans. I'm looking forward to "donut mode" in my FSD Tesla.
 
Just as you have a memory of the road a FSD car will have the memory of every other FSD car that shares the same hardware platform. It will know what happened to a car 1 minute in front. The memory will be compared with what the car sees and positioning data before every single manoeuvre.

Main cameras are behind the windscreen so they will be kept clean at all times. Radars see through fog and snow. Cameras see through the dark. If a camera is covered with snow the car can let you know so you can wipe it before setting off. Having direct heating around some cameras might help to.

Snow will be problematic but far from impossible to deal with just as it is for us mere humans. I'm looking forward to "donut mode" in my FSD Tesla.
YA but, as a Swede, in the meantime don't forget about the Semmel mode! :p
 
  • Funny
Reactions: AndyH

BTW., 85,337 VINs registered in Q1 so far means the following:
  • Even if Tesla stops filing for new VINs today, with the "simple 85% rule" this would suggest a Q1 production target of 72,500 Model 3's.
  • But there's still 48 days left from Q1, and in Q4 Tesla registered another ~9,000 VINs in the remaining 48 days of the quarter. If Q1 ist similar, that means ~94,300 VINs in Q1, which gives a Q1 production target of 80,150 Model 3's.
  • But if we use my full estimation method that worked well in Q4, then it's going to be 97,900 VINs, which gives a Q1 production target of 83,200 Model 3's.
So after Q4 Model 3 production of 61,400 this is a pretty bullish increase for Q1 in the +18-37% range, and most of those units are expected to be high ASP configurations like in Q3, but using Q1 production efficiencies (!).

The usual caveats apply: Tesla might depart from the pattern. Tesla didn't give any Q1 production guidance.

I'll keep updating my estimate as the quarter progresses.
 
Actually, a good radar could do so. You need a wavelength on the scale of some large fraction of the potholes' size. A weak return means "smooth" and a strong return means rough, aka "you have a pothole". You also need sufficient angular resolution. An easy way to do this would be with dual-antenna interferometry (one antenna on the left side of the car, one on the right), to create a large virtual aperture without needing a physically large antenna.

If they could adjust the wavelength over a wide range they could probe everything from potholes to road surface composition to ice and snow coverage, well ahead of the vehicle. Aka, having much better data than a human. You still wouldn't match LIDAR's angular resolution, but the usefulness of the data you could collect would be vastly superior (and not particularly weather-degraded).

Can you send a link where I can read about such radar? do you know the price?