EM: We made the investment in the GigaFactory and other companies didn't.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
- Status
One of the main take-aways for me is that HW3 is on track for end of Q1 – *this is crazy soon*. Andrej Karpathy talked about how the 10x increase allows for the deployment of a much larger (and more accurate) neural network. @neroden, I remember you posted this in the luvb2b-thread about your pessimism for FSD, and have been brewing on a response, so I hope it's cool with y'all than I clutter the celebration up with some attempted substance:
I think you might be missing some knowledge *collective gasp* about deep learning and the techniques Tesla and specifically Andrej Karpathy bring to the table. If one had to specify the problem specifically (as one normally does in CS), you would be right, that would take forever – but one of the major things deep learning accomplishes is *problem specification*. It figures out how to best approximate the problem itself.
One of the key factors in deep learning though – besides the specific architecture of the network – is figuring out how you can measure *when the network makes a mistake*. This is represented by the so-called "loss-function", which tells the network how wrong it was according to the desired output. It can then update the weights on each neuron to better approximate the desired result. The difficulty now though, is figuring out *when you are wrong*.
This, however, is quite a bit more simple than specifying self-driving - and shadow mode is exactly this. It enables training of a network against baseline "perfect" human behaviour. They do other stuff, of course, but shadow mode has the potential to be immensely helpful.
In case you guys haven't read Andrej Karpathys thoughts on software 2.0 (you might have seen his Autopilot lecture, which covers some of the same ground), this is a good primer – albeit a bit old:
Software 2.0 – Andrej Karpathy – Medium
Basically the FSD network will do something along these things:
Labeling each image > estimating 3D position of objects > figure out how to drive (okay simplifying here)
Each of these components are either independant software 2.0 processes, as described in the Medium post, or may even be merged into one network. I recall Andrej mentioning before that he believed "a single network to rule them all" was superior, but this was back when he was a phd and he was very coy about the techical details. He might have been working on some paper that never saw the light because he started working for Tesla. He might still believe this though – and the recent architecture improvements to the NN actually seem to suggest that indeed he still does.
Regardless, I think Tesla is (and has been for a long time) on the absolute forefront on applied deep learning and the sudden exponential increase that has happened to other problems attempted by deep learning, seem to be on the cusp of happening with FSD. Not that we'll be there in Q1 next year, but things seem to really be picking up steam, so I doubt it'll be a decade.
I think you might be missing some knowledge *collective gasp* about deep learning and the techniques Tesla and specifically Andrej Karpathy bring to the table. If one had to specify the problem specifically (as one normally does in CS), you would be right, that would take forever – but one of the major things deep learning accomplishes is *problem specification*. It figures out how to best approximate the problem itself.
One of the key factors in deep learning though – besides the specific architecture of the network – is figuring out how you can measure *when the network makes a mistake*. This is represented by the so-called "loss-function", which tells the network how wrong it was according to the desired output. It can then update the weights on each neuron to better approximate the desired result. The difficulty now though, is figuring out *when you are wrong*.
This, however, is quite a bit more simple than specifying self-driving - and shadow mode is exactly this. It enables training of a network against baseline "perfect" human behaviour. They do other stuff, of course, but shadow mode has the potential to be immensely helpful.
In case you guys haven't read Andrej Karpathys thoughts on software 2.0 (you might have seen his Autopilot lecture, which covers some of the same ground), this is a good primer – albeit a bit old:
Software 2.0 – Andrej Karpathy – Medium
Basically the FSD network will do something along these things:
Labeling each image > estimating 3D position of objects > figure out how to drive (okay simplifying here)
Each of these components are either independant software 2.0 processes, as described in the Medium post, or may even be merged into one network. I recall Andrej mentioning before that he believed "a single network to rule them all" was superior, but this was back when he was a phd and he was very coy about the techical details. He might have been working on some paper that never saw the light because he started working for Tesla. He might still believe this though – and the recent architecture improvements to the NN actually seem to suggest that indeed he still does.
Regardless, I think Tesla is (and has been for a long time) on the absolute forefront on applied deep learning and the sudden exponential increase that has happened to other problems attempted by deep learning, seem to be on the cusp of happening with FSD. Not that we'll be there in Q1 next year, but things seem to really be picking up steam, so I doubt it'll be a decade.
Seems to happen every time.AH fallingTo what is a stellar CC.
Wrong! He stated that service, not repair after a crash, is neutral.
mspohr
Well-Known Member
Tesla has the lowest cost per kWh and the lowest operating cost of any electric car.
Frankly, I can't even see that as possible. The VOLUME of vehicles is soon going to be Model 3, all under warranty so there is basically no service revenue. They have a lot of legacy S and X in the market, but it will soon be DWARFED by the Model 3 in the marketplace, USA at least. So, how do they see Service revenue going up significantly? I mean, are they expecting to launch a model 3 service contract, that will add 500-900$ per car in revenue, for maybe 40% of buyers and with less than that amount in spend?
Oh sweet Jesus I thought Deepak was going to leave..... then it turned out beautifully at the end.
i freakin love this quote
Are either expected to be doing autonomous taxi service in a widespread fashion any sooner? Clearly Tesla is not competing with human drivers today.
ev-enthusiast
Active Member
Kind words by Deepak. Thanking all the Tesla employees for their hard work
Brian has it over in the "estimates" thread
I expect the same old early open, mandatory morning dip. Because I think it's program trading at this point, and they won't all be reprogrammed by morning.
Stock hasn't reacted to the conf. call any more than to the report
If they can get margins in 15% range on Model 3 SR, I think they'll start selling them, even if MR/LR were experiencing high orders. Just build first come first serve (batched for efficiency, of course).
Remember that Tesla isn't purely capitalistic.
It's often (if not always) a sell on the news story. and in this market, it's take whatever you can get and get clear. We MAY get under 300$ tomorrow if the Asia/Europe selling catches up to us and we start lower.
mmd
Active Member
Summary:
1. 5k a week Model 3 is the new 10k a week. Kiss that 10k/week good bye. 5k/week enough to support Europe, China and US demand in Q1. Gotcha!
2. All talks about 10k a week vanished.
3. Was the $35k Model 3 contingent upon 10k a week Model 3 on a sustained basis? I think it was. So bye bye that too.
4. FSD bye bye.
5. $100 a KWh battery has been hyped up for so long in the past. Gone. Won't comment anymore.
6. Solar roof tiles that was ready to sell 2 years ago is now in R&D and testing!
The only sad part is I can't find shares to short.
1. 5k a week Model 3 is the new 10k a week. Kiss that 10k/week good bye. 5k/week enough to support Europe, China and US demand in Q1. Gotcha!
2. All talks about 10k a week vanished.
3. Was the $35k Model 3 contingent upon 10k a week Model 3 on a sustained basis? I think it was. So bye bye that too.
4. FSD bye bye.
5. $100 a KWh battery has been hyped up for so long in the past. Gone. Won't comment anymore.
6. Solar roof tiles that was ready to sell 2 years ago is now in R&D and testing!
The only sad part is I can't find shares to short.
Maybe, but at this point they are claiming March. My guess is production scales pretty slowly here and that the SR is delayed beyond March which I'm generally fine with since they secured plenty of room with those op.ex numbers. I don't think valuation turns on profitability with the 3, it turns on the roadmap and they just need to have a decent cushion against headwinds in the meantime.
In terms of applying rational valuations I think the market cap scales probably quadratically (or more) with how low they can get the cost on that car because it is a (nonlinear higher demand at lower price points) * (higher margin per unit) as you drop the cost so it really is the most important variable. Probably Toyota agrees with that.
This from Google... contrast that with Tesla!
As a general rule, new vehicle auto dealers have a net profit margin of 1-2% on new vehicle sales. It's pretty pitiful. Gross margins, however, run between 8 and 10% for most full-line automakers, and luxury cars often earn 10-15% margins. Depends on the vehicle, market conditions, etc.
As a general rule, new vehicle auto dealers have a net profit margin of 1-2% on new vehicle sales. It's pretty pitiful. Gross margins, however, run between 8 and 10% for most full-line automakers, and luxury cars often earn 10-15% margins. Depends on the vehicle, market conditions, etc.
Did anyone on that call ask what to expect from ZEV going forward? I might have missed it..
- Status
Similar threads
- Locked
- Locked
- Locked
- Poll
