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.
he would have to quarantine if he took an experimental shot. You can only avoid quarantine if you have one of the approved vaccines.


Anecdotal and totally Weekend OT: The son of a friend of mine who works in Cambodia came back to visit his parents this summer. He got vaccinated with Sinovac in Cambodia. This is not an approved vaccine in Germany, so when he arrived he was treated as if he was untested and unvaccinated. 2 weeks quarantine, after 6 days he could get out of quarantine with a negative PCR test.
He still does not register as vaccinated in Germany, so he's been trying to get a Pfizer/Moderna shot to "get on the books here" before he heads back to Cambodia... however, when he's been truthful about the "have you been vaccinated recently" question, he's getting turned away as they don't know the effect of Sinovac + mRNA vaccine. It's a bit of a rock/hardplace situation for him.
 
Tesla needs to figure out how to upgrade the super chargers to allow towed items or bicycles. The network just doesnt make it easy to do that without screwing other people right now.
Tesla Bot?
1629645955633.png
 
Saw some seriously long wait times posted earlier. I'm sure at least a portion of these wait times, particularly with long range S and X, is due to supply constraints. I'm also confident that demand for the entire Tesla lineup is strong.

I'm curious to hear opinions/evidence on whether 3 and Y wait times are more a function of chip (or some other part) shortages or if 3 and Y demand is just that staggering. Long-term, it doesn't really matter. Short-term, if deliveries are unable to keep skyrocketing, we might have a FUD-induced buying opportunity on our hands. If it's primarily just overwhelming demand despite Tesla continuing to grow deliveries, then we are going to lift off sooner rather than later.
 
Forward Observing

Maintenance ~ way back in the beginning of time; okay computers becoming a part of the job (for me 1977).

I’ll bet, since I own an X and a Tacoma; that all our 3 & Y owners are boasting at the bar that they have not sat thru a 5K service since owning their car. The current X had one minor glitch in the beginning; while the Tacoma requires oil changes or some fictitious maintenance every 5K. And, all that can be said is, “our chip supply“ is short handed. We bought the truck in February and the X in June 2019. The maintenance schedule is a real pain in the bottom with the truck. Cost is a joke too.

Biggest issue is my wife (better half, smarter half, partner, significant other, better looking of us ~ etc) hardly ever let’s me drive the X! God I need that Cybertruck with all the bells and whistles of X. All I can hope is that the resale on the Tacoma is still high. What will Toyota’s excuse be then?

Keep you eyes (camera) on the ball:)
I sold my Tacoma double cab 2WD recently for more than I bought it new MSRP from the dealer in 2016. I did oil changes and that's it.
 
Elon said - Rate of change is everything.

But here Ford Chief Engineer Donna Dickson said
“Thing will grow: This year we have actions and next year we have actions to target that 300 level. “

When asked for confirmation:
“so that’ll be every model year there will some updates coming out? “
“Yes. “
“Instead of a traditional mid-cycle action for these big changes?”
“No. “

Corporate structure and engineering economics of working with suppliers will continue to hold back Ford and many other OEMs pace of innovation, and that is something much harder to fix than to move from ICE to BEV manufacturing.

Ref: 30m12s mark
 
Another good deep dive on the Dojo computer. Near the end they speculate how they can make Dojo 10x faster in which the system on wafer design could be key. This is more like Cerebras vs the currently innervation of D1 tiles. There are 4 links here, each to each part of the articule.


 
Hi @jhm I don't have time to create a spreadsheet, but here are the raw Google TPU v2 and v3 numbers so someone else can do this:

Pricing | Cloud TPU | Google Cloud

View attachment 699872
I'm not showing Iowa (us-central1) because v3-32 is not available there. Either sold out to customers, or $GOOGL is using all the v3's they have in the USA to support their services. This explains what a Google TPU Pod is. Google also seems to be using their TPU v4's themselves - Sundar said they would not be available in Google Cloud to later this year -- ramping up the supply, I presume?
Is this approach as simple as Tesla leasing out spare compute (which they likely have very little for awhile), but once gen2 Dojo arrives, instead of Dojo1 going unused, they make huge $$$ and recoup all costs on Dojo1 AND Dojo2 with their "old" hardware? Rinse and repeat?
 
  • Like
Reactions: Snerruc and CLK350
Okay (upon reflection), maybe 10% of production line jobs could be done by robots in the 1st year, increasing to perhaps 50% after 5 years.

But the overall principle of adding robots to the workforce IS NOT to replace human workers, it is to enhance their value. The number of humans can remain constant, but their productivity is multiplied when freed from dangerous, repetitive, and boring jobs: just train a NN for a robot, turn over the task, and supervise the work. Eventually, just check the results.

Humans will have the greatest value as teachers, trainers, and problem-solvers. New and better techniques to do a job (or even to eliminate the need for the task), is what we humans do best.

Cheers!
Even if the goal was to keep the number of humans the same, adding robots is going to make worker's job security uneasy. Will require serious PR effort IMO
 
Last night, somehow I got into a thread from 2017 that had some of the posters projections for what this company’s stock will be worth in 10 years. Ultimately, we have already reached some of the highest projections.
That makes me wonder what, in 4 years by 2025, it will be worth and someone reading back on all this will think. 🤔
The next 4 years will be Interesting and exciting times for us, I know that much.
The only thing keeping me grounded is market cap. I'd be thinking 100x if not for market cap. Tesla has everything they need to become the largest company with the largest market cap in history, I'm just struggling with how high I can realistically see that going....
 
Personally, I get how a robot with legs and sensors isn't that big of a leap from a robot with wheels and the same sensors. My big question about the robot is how it fits in with the mission of accelerating the transition to sustainable energy. Is Tesla morphing into a broader tech company beyond sustainable energy? That's kind of where my head is at with the robot. I don't have a problem with it, I'm just puzzled about how it fits into the mission or if the mission is changing.
I don't think Robots directly fit in with the company's mission. But if it is not too much of a distraction and if it adds to the bottom line, then it only helps move faster in the primary mission for the company. For all we know they can (and to me it makes a lot of sense) to spin it off as a separate company, but share talent and knowhow.
 
Last edited:
I don't think Robots directly fit in with the company's mission. But if it is not too much of a distraction and if it adds to the bottom line, then it only helps move faster in the primary goal for the company. For all we know they can (and to me it makes a lot of sense) to spin it off as a separate company, but share talent and knowhow.
I see Tesla's mission evolving into what Elon recently tweeted...
 
Can someone please explain in simple words, what is the function of Dojo? How does Dojo help solve FSD? I keep hearing 'labelling', but I don't quite understand what that means. How are others solving this problem?

This is how I understood :

............. video ...................... labels
Cars ----------> Dojo -----------> Cars <cars can drive better now>
 
Last edited:
Elon said - Rate of change is everything.

But here Ford Chief Engineer Donna Dickson said
“Thing will grow: This year we have actions and next year we have actions to target that 300 level. “

When asked for confirmation:
“so that’ll be every model year there will some updates coming out? “
“Yes. “
“Instead of a traditional mid-cycle action for these big changes?”
“No. “

Corporate structure and engineering economics of working with suppliers will continue to hold back Ford and many other OEMs pace of innovation, and that is something much harder to fix than to move from ICE to BEV manufacturing.

Ref: 30m12s mark

Watching Ford's engineer Dickson answer questions made me realize how stuck in the past Ford is. Her job was to come on Munro Live and convince people Ford is changing, improving and moving with the times. But every answer dripped with corporate whitewash, framing every answer in a cookie-cutter style designed to make the company look good, all the things I hate about corporatism. Instead of confronting the difficult truth she kept pivoting to how Ford was changing, constantly improving. Not once did she admit how slowly a corporation like Ford actually moves or how overly dependent they are on their extensive supplier network.

Young people don't want to buy a Ford or GM product because they associate them with corporate double-speak and being slow to change. To my eye it appears Ford is not tackling the real problem here because it would change their Ford culture which is what provides for the individuals who make up their bloated corporate/engineering structure - it's who they are and why everyone has a good paying job. You would need to fire 90% of them and destroy the thing they hold dear (Ford culture) to fix it. Trying to get to where they want to go with incremental change is impossible. The only way to do it is to start from scratch.
 
The only thing keeping me grounded is market cap. I'd be thinking 100x if not for market cap. Tesla has everything they need to become the largest company with the largest market cap in history, I'm just struggling with how high I can realistically see that going....
Yes I understand. I do often feel the same way. It’s good to not set expectations too high, especially in these times.
 
Can someone please explain in simple words, what is the function of Dojo? How does Dojo help solve FSD? I keep hearing 'labelling', but I don't quite understand what that means. How are others solving this problem?

This is how I understood :

video labels
Cars ----------> Dojo -----------> Cars <cars can drive better now>
Solving neural networks essentially boils down to solving huge systems of equations…with millions of variables. While the math itself is fairly routine (solving large sets of equations), there are so many variables that you need a tremendous amount of compute power.

And you need a way to solve these networks quickly, so you can iterate by incorporating new data. PLUS, a lot of source data (labelled videos for example) is needed to feed the equations, so you need high data bandwith.

So the compute power helps solve networks quickly, and the bandwidth gets data to the equations so the variables can be solved.

Machine learning is basically about saying “here are the right answers. Fill out all these variables until you give me a gigantic system that, given these inputs, gives me the right outputs.” Networks have TONS of variables to solve for, and the more data you have, the better they get. So labelling is the process of annotating your data to tell the computer what the right answers are, which helps it find an ideal solution for this big system of equations.
 
Can someone please explain in simple words, what is the function of Dojo? How does Dojo help solve FSD?

Just as graphics cards (aka graphics processing unit or GPU) accelerate the computations needed to display 3D graphics, Dojo accelerates the computations needed to convert data from the cameras into a model that can be used to drive the car.

As mentioned in the AI Day presentation, they are basically turning the data from the cameras into a 3D graphics representation of reality, and the car is driving around that reality like a player in a video game. That takes a ton of computation, and Dojo makes that a lot faster than current systems can do.

Not sure if that is 100% accurate or too high level to be useful, but that is the gist of it as I understand it and not going too far into the technical weeds.
 
Last edited:
  • Helpful
  • Like
Reactions: CLK350 and -=buzz=-