Yep. FSD is to the automotive world what reusable rockets are to the launch industry. Who needs moats when you have the best engineers and out innovate everyone else by an order of magnitude?
(How’s that for cliches mash-up?)
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
(How’s that for cliches mash-up?)
Based on the vids I've seen, it appears that FSD has progressed extremely well since Tesla went all camera, although it was already a better driver than me at bets 1.0. I will be completely sold however when it can negotiate a French or Italian traffic circle during rush hour with no intervention required. I can't understand how people do it now!
RobDickinson
Active Member
Theres still a bunch of potential hurdles for FSD.
The one narrow FOV camera in the front is good for far distance vision, but I am not convinced the others are detailed enough or positioned well enough for pulling out onto fast multi lane roads.
And can FSD 3.0 run enough NN to cover enough edge cases?
The one narrow FOV camera in the front is good for far distance vision, but I am not convinced the others are detailed enough or positioned well enough for pulling out onto fast multi lane roads.
And can FSD 3.0 run enough NN to cover enough edge cases?
Don't forget Tesla's innovating with castings is not done. The next step is casting the entire underbody of a car in one piece, then maybe the entire upper-body. Mega-castings are a moving target.
Clearly the analogy-thinking engineers at ICEmakers never before thought of using casting machines bigger than a house, or convinced their analogy-thinking managers to try it. How long will it take them to wrap their heads around machines even bigger?
Last edited:
RobDickinson
Active Member
Elon seems to disagree.
Elon Musk: Tesla Is Trying To Make Real Cars With Casting Like Toy Cars
Tesla has not given up on its dream of manufacturing cars by creating the entire body of the vehicle with a single-piece casting.
insideevs.com
RobDickinson
Active Member
LOL, then you hit a deer and your $60k car is totaled!
RobDickinson
Active Member
Right now if more than the basic crash structure is damaged your car is totalled.
Much better to total a car thats $30k and easier to recycle?
Sounds familiar, but can't quite put my finger on where I heard this before.
engle
Member
Here is what I expect will be announced on August 19 during the 2021 Tesla Autonomy Day:
I have been carefully watching Tesla's moves in this space for a long time, ever since I leased a 2016 Model X P90D in March, 2016 with the old basic Mobileye (Intel bought them) system. I'm an early-retired software engineering consultant who has been a Realtor (2011+ full-time, now part-time managing properties and equities). Last year I taught myself Python and ML (Machine Learning), including CNN (Convolutional Neural Network) ML models, which is how Tesla and others "recognize" objects by first "training" the model and then running what is called "inference" in the vehicles. There are millions of statistical probability parameters inside these inference models that are set during training. It's very interesting. It is a brute-force method of quickly training a machine, unlike us humans who learn from neural feedback (touch, audio, vision, taste etc.) as we are exposed to our immediate environment. That's why Tesla developed a "video game" style simulator of the driving environment to train models. They cannot send a real car out on the road to crash into human-driven cars until it "learns" how to drive!
- All Teslas in production since XX/YY/2021 have the new FSD Hardware 4 installed. HW4 has #.#x improved performance over HW3.0 with 1.#* higher power and lower cost. It is a Tesla-designed SOC (System On a Chip) fabricated using a 10 nm process by TSMC (instead of 14 nm by Samsung). It offers #+ improvement in fps (Frames Per Second). Each chip is capable of over 80 trillion operations per second (3.0 is 36+), and there still are two for redundancy. HW4 also supports LIDAR sensors for Tesla Network operation with no driver in poor weather. Note that I captured video of LIDAR being tested on a Model 3 near Palo Alto back on July 4 weekend - I have been too busy to upload to my YT channel will do so soon.
- $TSLA will charge to upgrade from HW3 to HW4, since HW4 is not required for the "FSD" feature set we paid for. However, HW4 will enable features that go beyond "FSD". They will be introduced with a new marketing name. That will be eventual full autonomy. We will need HW4 for that.
- The earlier announcement of $199/mo "FSD" subscription with a now-$1,000 FSD 3.0 board swap was partly designed to clear-out extra inventory to make way for the FSD 4.0 hardware that will be announced on August 19. To not be already installing it that week in new vehicles will Osbourne Tesla sales so I'm almost certain that will be the case.
I have been carefully watching Tesla's moves in this space for a long time, ever since I leased a 2016 Model X P90D in March, 2016 with the old basic Mobileye (Intel bought them) system. I'm an early-retired software engineering consultant who has been a Realtor (2011+ full-time, now part-time managing properties and equities). Last year I taught myself Python and ML (Machine Learning), including CNN (Convolutional Neural Network) ML models, which is how Tesla and others "recognize" objects by first "training" the model and then running what is called "inference" in the vehicles. There are millions of statistical probability parameters inside these inference models that are set during training. It's very interesting. It is a brute-force method of quickly training a machine, unlike us humans who learn from neural feedback (touch, audio, vision, taste etc.) as we are exposed to our immediate environment. That's why Tesla developed a "video game" style simulator of the driving environment to train models. They cannot send a real car out on the road to crash into human-driven cars until it "learns" how to drive!
Knightshade
Well-Known Member
That they announce HW4 is in production is.... possible-- the timing would roughly line up with what was said on autonomy day.
The LIDAR bit is 1000% wrong.
Tesla has used LIDAR on test cars to verify how well their non-lidar systems results compare with what LIDAR readings are.
They've done that for years.
Tesla with LiDAR and mfg plate spotted testing in Palo Alto near Model 3 sighting
On the same day a black Model 3 release candidate was spotted driving around Silicon Valley, another Teslarati reader caught a Model S with a roof mounted LiDAR and additional self-driving hardware testing around the grounds of Tesla’s headquarters in Palo Alto, California. After further...
www.teslarati.com
That's from 2017 showing Tesla using LIDAR on a Model S for exactly that reason. What you filmed is the same thing.
They're not adding LIDAR to their production vehicles.
This is also wrong.
Everyone who bought FSD prior to roughly March 2019 were promised, at minimum, L4 self driving (it's arguably if L5 is promised or not).
So, at minimum, those folks would get free upgrades.
Indeed when HW2.5 first became public knowledge Tesla outright stated whomever owned FSD and needed upgraded HW for it would get it, free.
I think it might, very very technically, be legal for Tesla to NOT upgrade the post 3/19 buyers if they deliver city streets at L2-- but it'd be a nightmare from a customer satisfaction perspective... and the economics wouldn't be so onerous to do the upgrades for all FSD owners that the bad optics would be worth it.
NOTE- I expect that as with HW2, they'd only be doing free upgrades if you OWN FSD...not if you just want or have the subscription.
....what?
You think Tesla somehow didn't know WHEN they themselves were going to switch the computers in new production and ended up with a surplus of HW3 somehow and THAT is why they are offering the $1000 board swap?
Last edited:
Charging for a HW4 upgrade would greatly piss me off, and I'm on a TSLA investor forum and want them to succeed and have given them the benefit of the doubt on a ton of things. This would definitely be the wrong move for them.
Knightshade
Well-Known Member
Gasoline.
Since, by basic math, there has to be hundreds of millions of gas vehicles still on the roads in 2040, there should still be plenty of supply.
The industry won't look the same, or be as large, but it'll still exist.
Again- you need to replace 1.4 billion ICE vehicles. In 19 years from now.
The math flat out does not work.
Even if -100 percent- of new cars sales were EVs today you wouldn't quite make it.
And we're MANY YEARS away still from 100% EVs.
We're not even going to hit 10% EV market share of new car sales, worldwide, this year.
Why do you assume they'll lose money doing it?
The majority of oil produced is not used for gasoline even today. The fact it also won't be tomorrow isn't a change.
And as demand does drop off, some of the more expensive techniques used to obtain more of it won't be needed anymore either.
And still will.
There might not be 10 different stations off every major interstate exit, but it won't be hard to find gasoline.
And the ones left will likely have some DC fast chargers too. Right now many of the superchargers around here are located at a Sheetz for example.[
Another way to think of things is Tesla tends to deploy resources to solve the biggest problems at any point in time.
- Castings - Body shop
- FSD computer - FSD
- Dojo - FSD training
- Battery Day - Cell production
- 12V Battery - reliability
- Wiring harness - weight/copper
- Lighter weight/ glass / batteries
- Copper mining - copper
- Some computer chips - reduce supplier dependency
IMO no government will seriously attempt to slow Tesla down, not many countries produce their own oil and cars or are immune for Climate Change impacts.
Governments will almost certainly give legacy auto some assistance to transition to EV production, and most of them will need it.
During the Olympics coverage in Australia, there are Toyota ads spruiking the advantages of Hydrogen in a vague way, the effectiveness of that in term of selling Toyota Hydrogen cars in Australia will be close to zero..
Tesla mostly doesn't need to worry about governments slowing them down, competition, or the pace of competitor innovation. Competitors are too busy falling over their own feet, massive government support will be needed just to keep them in the game..
Some Chinese competitors are smart enough to copy Tesla, and Chinese car makers are improving rapidly. legacy auto may find it easier to convince governments to give them trade protection against the Chinese. Ultimately a Tesla factory in Japan or Korea might make sense... I don't expect government to objec tto Tesla building factories in their countries, Berlin was probably as hard as it is going to get. Japan and Korea have local battery suppliers, and a good pool of talented engineers. Ultimately I can see Tesla producing more than one compact model.
It is wrong to simply count the number of vehicles, we need to look at how many miles per day they are driving on average.
Robotaxis and EVs are very likely to dominate all high mileage situations...
By 2030 My hunch is around the following EV production numbers:-
- Tesla - 20 Million
- China - 30-40 Million
- VW - 5-10 Million
- Other European - 5 Million
- Japan - 5 Million
- Korea - 5 Million
That should at least double by 2040 to 140 Million per year.
If we take 70 million by 10 years - that is 700 Million half of the ICE fleet gone by 2040 on fairly conservative numbers.
By that time I bet the majority of the rest of the ICE fleet spends most of the day sitting around in garages...
So ICE fuel consumption probably drops something like 80% by 2040 that means 4 out of 5 gas stations close on average.
ICE driving will be considered similar to smoking and banned in many built up urban areas. Most commuters will prefer Robotaxis, answering emails, browsing the internet, or posting on TMC on their way to work.
Last edited:
Knightshade
Well-Known Member
I don't include imaginary things in my financial modeling.
Right now RTs are largely imaginary.
Where you get enough battery production to go into those EVs from?
Why did Elon think 30 million for the entire industry by then was more likely (with indeed 20 million of em being Teslas).
What do you know about where the other 40 millions cars worth of EV batteries will come from that he doesn't?
...what?
You think Robotaxis will come and replace a lot of new car sales- but somehow total new car sales will also double annually by 2040?
Why?
But you're not gonna have 70 million over those 10 years...and even if you did- that leaves you with the other 700 million on the road being ICE vehicles.
This seems as unlikely as all your other assumptions.
The two places that'll have the highest concentration of ICE vehicles at that point are:
Poor people who can't afford new cars.
They aren't going to have their 12 year old Civic "sitting around in the garage" because it's the only thing they have. Also they probably don't have a garage either.
and-
People in 3rd world nations where EVs will continue to have the lowest, and slowest, penetration rates- and people tend to keep anything running they can get their hands on, for as long as they can keep it running.
All those folks will still be driving those gasoline cars because they have nothing else to drive.
If you wanna tell me Norway will be 100% EVs and all gas cars off the road in 2040- sure, probably will be.
If you wanna tell me Nigeria- where they don't even have electricity roughly half the day on average- will be I'd suggest you check you do some that first principles thinking folks are always talking about
Get to the Monday!
Am hoping for a sundae.
Haz TSLA fun day?
Am hoping for a sundae.
Haz TSLA fun day?
We know where Tesla's batteries are coming from... for the record I think Tesla's lithium clay extraction works,,, and that is the tip of the iceberg in relation to a rethink on mining and materials processing..
Lots of companies are scaling up plans for battery production.. this is the only "land grab" that is happening,.
Typically 5 years estimated for a new mine, 5 years estimate for a new battery factory, I bet both can be done faster by highly motivated companies... More so if they take note of what competitors like Tesla are doing and develop similar processes.
Yes to both, that is called the valley on unmet demand - JPR007 posts a graph on Twitter that illustrates the point...
Customers will simply stop buying new ICE way earlier than many predict, they will run the old car and/or use a Robotaxi...
There is a one-time build on the Robotaxi fleet possibly 60 million worldwide..... yes many people who currently don't have transport, will have transport.
Yes, some compact EVs will be very affordable, cheaper than any new car today.
EDIT:: IMO Robotaxis are a matter of when not if, and when could be longer than expected, But in many locations the wages of a human driver are not a deal breaker.
Last edited:
Similar threads
- Locked
- Locked
- Locked
