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

[Mo City]

There's a chance that if the public sees an intervention every 50 miles, they'll start pricing in robotaxies.

Devil's advocate: Decreasing interventions an order of magnitude is a huge effort. I wouldn't be surprised if there were many rewrites to get the extra 4 or 5 orders of magnitude, with the effort for each comparable to the recent rewrite.

According to Elon, "This time, it’s right" and Dojo is the solution for the March of 9s. Dojo v1.0 will arrive in "maybe a year or so".

 

[Artful Dodger]

After-action Report: Thu, Oct 15, 2020: (Full-Day's Trading)

Headline: "Split from Macros Halts 7-Day TSLA Winstreak"

Traded: $16,018,081,672.27 ($16.09B)
Volume: 35,688,824
VWAP: $448.83

Close: $448.88 / VWAP: 100.01%
TSLA closed ABOVE today's Avg SP
Mkt Cap: TSLA / TM $418.27B / $183.32B = 228.16%​

TSLA 1-mth Moving Avg Market Cap: $399.43
TSLA 6-mth Moving Avg Market Cap: $265.86
Nota Bene: 4th tranche of CEO comp. likely unlocked last week

'Short' Report:

FINRA Volume / Total NASDAQ Vol = 51.9% (51st Percentile rank FINRA Reporting)
FINRA Short/Total Volume = 48.2% (50th Percentile rank Shorting)
FINRA Short Exempt Volume was 0.33% of Short Volume (43rd Percentile Rank)​

Comment: "Wasserrechnung zum €4.3B"

View all Lodger's After-Action Reports

Cheers!

 

[MP3Mike]

2021 Model 3 range updates: (They have updated the page to include pictures of the new console, etc.)

• SR+: 263 miles (estimated)
• LR: 353 miles (estimated)
• Performance: 315 miles (estimated)

2021 Model Y range updates:

• LR: 326 miles (estimated)
• Performance: 303 miles (estimated)

2021 Model X range updates:

• LR+: 371 miles (estimated)
• Performance: 341 miles (estimated)

2021 Model S range updates:

• LR+: 402 miles (estimated) (no change?)
• Performance: 387 miles (estimated)
• Plaid: 520+ miles (estimated)

They really should add the ~5% to the Model S LR+ so that the estimated range would be 420 miles.

 

[Tyler34]

Model 3 is now 353 miles of range and model y is 326 miles of range on the Tesla website

 

[MC3OZ]

According to Elon, "This time, it’s right" and Dojo is the solution for the March of 9s. Dojo v1.0 will arrive in "maybe a year or so".

I'm optimistic enough to believe this...

If the platform is right, there is no significant need for more expensive software development, it is mostly just more real world examples, better training and smarter Neural Nets.

If Dojo partially generates training data with limited human intervention, we have a feedback loop where the system continually improves and learns from it's mistakes, those mistakes source better training data and that new training data improves the next round of training..

I think "operation vacation" is an aspiration to statement of goals, rather than an expected outcome. The real aim is maximum improvement in the NN with minimal human effort and expense.

People who doubt this are thinking it can't be automated, or it can't be done in the NN layer, or driving requires human logic.
We drive with an element of logic and memory a NN has similar properties...

Tesla a vague about the mapping aspect, but I think mapping exists in FSD and it is similar to our memory, FSD knows to expect a Stop sign at a particular location, and where the road lanes were last time, and vision confirms or denies that expectation.

So we need to identify something that FSD can't solve, something Elon and the smart people working for him haven't thought of.
IMO things like snow storms , obscured cameras, broken cameras or radar interference are more likely to be issues than a simple lack of training... I tend to think Elon and co have thought of my list, and just about everyone's list.

 

[Artful Dodger]

Ever heard the story about crying wolf too many times?

Ever hear the story about the scorpion and the frog?

 

[Bet TSLA]

The new magical FSD rewrite is going to be exactly like what we saw on battery day. Engineers will look at it and say:
[Battery Day] Fossil fuels are dead; legacy auto is now so far behind they can't even see us; ramping up in only three years!!!
[FSD rewrite] Rough but it will clearly solve most current problems over time; will take only about a year to get smooth and by then they'll be using Dojo to work on the march of 9s. Wow!!! End in sight!

Wall St and everyday users;
Ho hum. Doesn't drive as well as I do today. Just more useless Elon hype. When will we get Lidar? At least it's a little better than before.

 

[printf42]

2021 Model 3 range updates: (They have updated the page to include pictures of the new console, etc.)

• SR+: 263 miles (estimated)
• LR: 353 miles (estimated)
• Performance: 315 miles (estimated)

2021 Model Y range updates:

• LR: 326 miles (estimated)
• Performance: 303 miles (estimated)

2021 Model X range updates:

• LR+: 371 miles (estimated)
• Performance: 341 miles (estimated)

2021 Model S range updates:

• LR+: 402 miles (estimated) (no change?)
• Performance: 387 miles (estimated)
• Plaid: 520+ miles (estimated)

They really should add the ~5% to the Model S LR+ so that the estimated range would be 420 miles.

Seems they already cleared out previous inventory.

I am in Bay Area and only see ~30 new inventory Model 3 on the website.

Very smooth transition and no Osborne effect.(almost 10% range increases is not insignificant, I was worrying about the Osborne effect since the Electrek heat pump leaks, now a splendid Q4 is secured.)

 

[Green Pete]

The new magical FSD rewrite is going to be exactly like what we saw on battery day. Engineers will look at it and say:
[Battery Day] Fossil fuels are dead; legacy auto is now so far behind they can't even see us; ramping up in only three years!!!
[FSD rewrite] Rough but it will clearly solve most current problems over time; will take only about a year to get smooth and by then they'll be using Dojo to work on the march of 9s. Wow!!! End in sight!

Wall St and everyday users;
Ho hum. Doesn't drive as well as I do today. Just more useless Elon hype. When will we get Lidar? At least it's a little better than before.

My understanding was the the the rewrite was a fundamental redesign of how the software handles the multi camera input. And that Dojo was an automation of the labeling process. These two things are both very important, but are also functionally independent. DOJO could exceed all expectations anyone has and autopilot could still need to be fundamentally rewritten every few years (could be great now, but it will never be "perfect" so it will need improving until an AI starts rewriting the code itself). I may have missed something but that was my understanding of both items of interest.

edit: the automated labeling will accelerate the improvement curve. The design of the autopilot software itself will determine whether a robotaxi is functionally possible. DOJO will just help tackle edge cases faster

 

[Criscmt]

To me it seemed that Elon was just countering the implication by Lucid that Peter designed the Model S. Obviously Lucid (and himself) are going to play that up as much as possible, and presumably Elon would have a tendency to downplay that. We all lie on our resumes after all.

“Having completed conceptual and design engineering work on Model S, Peter has decided to step away to tend to personal matters in the U.K.,”. This was the official statement from Tesla when Peter left.

And, Peter at least seems to be comfortable stating that he didn't design Model S "Model S was actually styled before I joined Tesla. My task was to retrospectively fit all the bits into it. "

One other instance I wish Elon spent time on something useful, which I believe is his default mode, rather than trying to take a dig at Peter.

 

[Mo City]

Model 3 is now 353 miles of range and model y is 326 miles of range on the Tesla website

Get used to this. Based on Battery Day, I predict Model 3 will offer 400+ miles of range in the next 3-4 years and per vehicle battery cost will be lower than it is now.

One can only imagine what the price of a Plaid Model S will fall to by 2025 (unless they jack the specs up even more than currently advertised).

Tesla: The target that not only keeps moving, it's actually accelerating.

 

[heltok]

Tesla
Pros: best hardware, best vertical integration, most data
Cons: camera sensor fusion algorithms lagging behind Lidar, but quickly catching up, lots of promises and long delays

Waymo
Pros: best sensor fusion, best control
Cons: bad vertical integration, lot to lose, slow execution, drives like grandma

Cruise
Pros: decent sensor fusion, decent control, decent integration
Cons: not best at anything, not doing anything that is hard to replicate, in few years there will be open software projects doing what they are doing

My understanding was the the the rewrite was a fundamental redesign of how the software handles the multi camera input. And that Dojo was an automation of the labeling process. These two things are both very important, but are also functionally independent. DOJO could exceed all expectations anyone has and autopilot could still need to be fundamentally rewritten every few years (could be great now, but it will never be "perfect" so it will need improving until an AI starts rewriting the code itself). I may have missed something but that was my understanding of both items of interest.

Dojo is the hardware that will make training cheaper and faster.

4D is mainly a difference in labelling, instead of labeling 2D images one at a time, all cameras with multiple time frames(video) will create a 4D point cloud that will be labelled mostly automatically(using the future as a ground truth) and when needed(to initialize the training set and to fix errors) by humans. This will enable more accurate labelling, higher performance prediction(label in 4D to predict 4D) and mostly a 3 orders of magnitude increase labeler efficiency(distance of driving labeled per human per time unit).

 

[Right_Said_Fred]

We’re half way through the first month of Q4 but no ships seem to have left the west coast yet with destination Europe. I know of one ship that left for south east Asia. Is Tesla prioritizing Asian deliveries in the first weeks of the quarter or have they shifted to east coast shipping for European deliveries?

 

[mickificki]

Model 3 with 353 mi range, black trim, all model 3s now come with power trunk lift gate. Damn we gonna sell buku cars! Let’s go Tesla!

 

[Dreadnought]

So we need to identify something that FSD can't solve, something Elon and the smart people working for him haven't thought of.
IMO things like snow storms , obscured cameras, broken cameras or radar interference are more likely to be issues than a simple lack of training... I tend to think Elon and co have thought of my list, and just about everyone's list.

Bad weather is something that humans are also struggling with. It may require hardware changes to keep the cameras clean but this is not a hard problem. For the march of 9s, they need to be really good at capturing the rare freak event.

There are plenty opportunities to identify and eliminate the reasons that cause an intervention every 50 miles. The one intervention per 500k miles like the odd armchair in the middle of a highway that just fell off a truck is, by definition, the rare event that happens once in the lifetime of the car. Driver intervention (preferred over a crash) will be the signal to identify the corner cases. But once the car can do 5000 miles by itself, it becomes likely that the driver won't be paying attention. The car will need to become a lot better at planning ahead and strategies to identify situations where it will need driver input, leaving sufficient time for the human to digest the situation.

 

[UkNorthampton]

Listen, Tesla deliveries were crap in the Czech Republic compared to all the competition. So obviously, they'll be reporting a massive loss for Q3. You fools just can't see it coming yet.

Vatican City sales...

 

[WarpedOne]

The one intervention per 500k miles like the odd armchair in the middle of a highway that just fell off a truck is, by definition, the rare event that happens once in the lifetime of the car

That is a wrong question, the right question is much simpler "Is the road ahead free?"
Is it blocked by black leather or red fabric chair with 3 or 4 legs does not change anything.

People are not capable to design an IF-THEN-ELSE set of rules precisely because they tend to ask such wrong questions.

 

[Dreadnought]

That is a wrong question, the right question is much simpler "Is the road ahead free?"
Is it blocked by black leather or red fabric chair with 3 or 4 legs does not change anything.

People are not capable to design an IF-THEN-ELSE set of rules precisely because they tend to ask such wrong questions.

While this was an arbitrary example, the point is that the car has to discern between something that requires to slam on the brakes and something that can be ignored. There is a thin line between an overly cautious car that stops for every unrecognised object and news headlines of yet another child run over by a Tesla because it wore a Halloween costume and wasn't recognised as a human.

 

[WarpedOne]

While this was an arbitrary example, the point is that the car has to discern between something that requires to slam on the brakes and something that can be ignored. There is a thin line between an overly cautious car that stops for every unrecognised object and news headlines of yet another child run over by a Tesla because it wore a Halloween costume and wasn't recognised as a human.

Do you drive over a paper bag on a HW?
What if there is a brick inside?
Exactly.

You present a problem and act like it is fully solvable and us human drivers always decide correctly.
But we don't.
Computer won't either but will still drive safer than me or you.

 

[mrdoubleb]

Any news on the heat pump for the Model 3? Can't seem to find anything on this on the official site.