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

[The Accountant]

Don't be silly now.. have you learned nothing from previous Tesla days? Here's how it's gone down up until now: Tesla shows off and tells us about a bunch of things that are extremely impressive and promising to mega-nerds and Tesla fanatics. I promise that whatever is shown on AI day will not be viewed very positively by Wall Street, unless Tesla changes their intent for these days completely. What will be shown will not be immediately and plainly profitable, and thus (in the eyes of WS) is worth nothing for the foreseeable future.
From their point of view they'd also be historically correct. Looking at battery day and AI day, how has the company profited from that since their associated days? If you can't look past a 12 month horizon, I'm sure AI day #2 will also be a display of far fetched and unlikely to become succesful (profitable) pet projects from Elon, and they wish he'd just get some more models out the factory door. They didn't find battery day worthwhile either, even considering all of the nigh impossible achievements accomplished by tesla up to that point. No matter if Tesla sets a realistic timeframe and actually meets it, it just doesnt register until the money comes pouring in.

To anyone expecting Optimus doing anything so useful that a typical factory worker might feel threatened - forget about it. You wont see Optimus installing a typical wire harness, but moving boxes is a given I'd say.

My own expectations:
- Dojo is growing, teasers about V2
- Detailed info and projected release for HW4, expect that HW3 will remain capable of FSD for the foreseeable future.
- Lots of impossible to understand (by me) but cool to watch and promising FSD stuff.
- Optimus time. Prototype shows up and does some things that wow people who know stuff about robots, but don't really impress anyone else. Possibly the other way around
- Next trading day, TSLA moves with the markets and ignores AI day completely, even after the info has been digested and everyone here understands what an enormously big deal it is.

Anything better than the above would surprise the hell out of me, and I would gladly accept such a thing, but I'm sure as hell not counting on it.

Agree. AI Day 2 is a recruiting event. What will impress AI/ML talent will likely fly over my head.
In the weeks that follow, Tesla will get flooded with resumes from the top talent in AI/ML and TSLAQ and pseudo-media will mock it. That my guess.

 

[Favguy]

Dude, these are wildly unrealistic expectations. It's expectations like this which result in disappointment after events like Battery Day and AI Day 1. Elon pushed this event a month just so they could even have ANY form of prototype whatsoever.

Well, if the bot doesn't look like Dolores in full Blue 1880's garb I'm selling up on Monday...

But seriously, I expect we'll all swoon over whatever it is we see, which will be impressive. It will do its job for recruitment and everyone else will scratch their heads saying "eh? what?" and the stock will fall on Monday.

 

[Mike Smith]

Here's why I think Tesla will show the Tesla Bot doing work in the factory:

1) It can walk. If it couldn't walk they wouldn't be showing it. It might not be super smooth but I'm sure it can get around.

2) It can navigate. Elon said the AI/Autopilot team had an end of month deadline, so it will be able to move around within some basic environment.

3) Tesla already has a lot of experience in programming robotic arms to do useful things. Remember Tesla, unlike other manufacturers, has been writing their own software for their Kuka arms for years.

Edit: My prediction is they'll show it doing three things: moving a part from one station to another, moving boxes around, and doing some very basic part installation.

 

[DragonWatch]

Forward Observing

The difference between AI day 2.0 and any other special Tesla day which falls on Friday 30 September; Sunday the 2d of October In time for my late am coffee is 3Q deliveries.

Wall street might be in for a T-bone accident. Is that a stock thing? Is there room for legacy auto manufacturers shrinkage?

Cheers

 

[Cobbler]

...

3) Tesla already has a lot of experience in programming robotic arms to do useful things. Remember Tesla, unlike other manufacturers, has been writing their own software for their Kuka arms for years.
...

As a former certified Fanuc system integrator, I can tell you that other manufacturers, also have custom firmware/hardware for their robots (Toyota, VW,...) to have a better integration into their factory automation. I don't think Tesla buys the hardware and completely rewrote the software from scratch. Maybe communication protocols, tool integration, loading models for path planning,...etc could all be Tesla's own.
Industrial robots are programmed to execute motion paths with their TCP and do not have any neural nets on board for their motion. 4D points are executed, translated into movements for each axis to reach a certain destination within a time, position tolerance. (Some machine learning is already used in industrical automation: for eg. vision applications for pick & place scenario's, use neural nets to find stuff, send the coordinates to the robot to pick it up)

The Honda Asimo bot could walk because of many algorithms, set of rules and logic, defining every scenario based on sensor data, to control motors to show a walking robot. Everything defined. Similar to industrial robots.

Tesla bot movements, interactions, ... will rely heavily on a framework using neural nets. Based on a lot of training data, a model could be achieved that exponentially increases the Teslabot possibilities for walking, jogging, overcome obstacles, grabbing stuff, ... very rapidly.

 

[petit_bateau]

As a former certified Fanuc system integrator, I can tell you that other manufacturers, also have custom firmware/hardware for their robots (Toyota, VW,...) to have a better integration into their factory automation. I don't think Tesla buys the hardware and completely rewrote the software from scratch. Maybe communication protocols, tool integration, loading models for path planning,...etc could all be Tesla's own.
Industrial robots are programmed to execute motion paths with their TCP and do not have any neural nets on board for their motion. 4D points are executed, translated into movements for each axis to reach a certain destination within a time, position tolerance. (Some machine learning is already used in industrical automation: for eg. vision applications for pick & place scenario's, use neural nets to find stuff, send the coordinates to the robot to pick it up)

The Honda Asimo bot could walk because of many algorithms, set of rules and logic, defining every scenario based on sensor data, to control motors to show a walking robot. Everything defined. Similar to industrial robots.

Tesla bot movements, interactions, ... will rely heavily on a framework using neural nets. Based on a lot of training data, a model could be achieved that exponentially increases the Teslabot possibilities for walking, jogging, overcome obstacles, grabbing stuff, ... very rapidly.

In setting out what we hope is shown at the reveal we have to think about the likely state of project progress, and the different levels of understanding in the audience/viewers.

One reasonable thing that Tesla bot may be able to do at the reveal is to walk across a space with some obstructions/objects in it. THEN someone moves a few of the objects and the bot walks a DIFFERENT path back across the space. If it can do this to voice command, and pick up/carry something (solid/uncrushable) that would be even better.

 

[Mike Smith]

In setting out what we hope is shown at the reveal we have to think about the likely state of project progress, and the different levels of understanding in the audience/viewers.

One reasonable thing that Tesla bot may be able to do at the reveal is to walk across a space with some obstructions/objects in it. THEN someone moves a few of the objects and the bot walks a DIFFERENT path back across the space. If it can do this to voice command, and pick up/carry something (solid/uncrushable) that would be even better.

Everyone is assuming that Tesla just started working on this a year ago, but I suspect they've been working on it in secret for a lot longer than that. Elon didn't really give any indication of how far along they already were at the first AI day other than to say the one on stage wasn't real.

 

[traxila]

Reports of a fire at GigaBerlin seem to be hitting SP this AM.

Fact that fire has to do with cardboard recyclables and is out with no one hurt means nothing to the algos.

Let's see how many headlines they can come up with that imply batteries burning or the entire factory having been smoked.

 

[jeewee3000]

To anyone expecting Optimus doing anything so useful that a typical factory worker might feel threatened - forget about it. You wont see Optimus installing a typical wire harness, but moving boxes is a given I'd say.

Even moving boxes sets a very high bar of expectations.

As many here have said: the hype for these events is always (too) high. What will be shown will show great promise in the longer term but it'll be just that: plans and expectations.

AI Day is a recruiting event, which they hold because the goals they want to reach are not reached yet. They need more talented people to create the vision set forth by Elon/Tesla.

So I don't expect anything from the Optimus demo besides maybe some movement through the room. (Basically FSD Beta but with (very slow) bipedal movement).

If Optimus picks up any object I'll be very impressed.

On another note, let's hope we hear a clear updated roadmap for FSD Beta (version 11, anyone?), since Elon has repeated many times he expects FSD to be on par with human driver statistics this year. (i.e. less interventions and/or crashes per mile than the crashes/mile average of humans, which is a pretty low bar).

Either way I don't expect the SP to increase due to AI Day. (Due to Q3 numbers, that's a different matter).

 

[UncaNed]

IMHO "don't worry Tesla will handle it" would also be an excellent way to consider whatever Optimus might do at AI day, isn't it? Perhaps even highlighted to a thread all it's own just for extra emphasis and utility. If Tesla hired a few folks from Boston Dynamics, then expectations for Optimus should be high.

Also, I took my very first FSD Beta drive yesterday afternoon. I'm impressed! A bit put off by the wild steering wheel gyrations at intersections while it makes up it's mind, and as expected it has a way to go, but I am happy.

What sort of camera/phone mount arrangement works best to video FSD in action over the driver's shoulder showing the touchscreen and forward vehicle view so I can show my friends and spread the Tesla word??

 

[JusRelax]

Reports of a fire at GigaBerlin seem to be hitting SP this AM.

Fact that fire has to do with cardboard recyclables and is out with no one hurt means nothing to the algos.

Let's see how many headlines they can come up with that imply batteries burning or the entire factory having been smoked.

Our very own @avoigt reported on the likelihood of this fake news reporting, and apparently got himself blocked by Fredlectrek... unsurprising..

https://twitter.com/x/status/1574344405141671937

Thanks for warning the twitters about this fake news, @avoigt!

 

[MikeAtkinson]

What I want to see at AI day

Optimus hardware: individual elements, hands, skin, haptics, motors, face (eyes+hearing) and how they are going to improve them over time.
Optimus software: the neural nets, general training for walking, etc.
Optimus training: how it is trained for specific tasks
Optimus networking & feedback: how they learn
Optimus manufacturing: development bots, large scale manufacturing

DOJO update, v2?, scaling, performance, selling systems or time on systems.

FSD update, training, HW4
AI for robotaxi service

AI use in Factory OS.

AI usage elsewhere in Tesla (+SpaceX?)

If we get an update on all these, with progress and development path I will be more than happy.

Any progress towards AGI would be a bonus.

 

[canoemore]

Our very own @avoigt reported on the likelihood of this fake news reporting, and apparently got himself blocked by Fredlectrek... unsurprising..

https://twitter.com/x/status/1574341934272974849

Thanks for warning the twitters about this fake news, @avoigt!

At first glance I read this as "extinguished by fireworks". Was disappointed upon more careful inspection.

 

[scaesare]

It needs to beat this

A pair or group of droids doing a task would look great.

That seems like a pretty low bar.

Also, natural biped walking must be awfully hard...I've yet to see a robot do it well.

 

[elasalle]

Chinese EV maker Li Auto falls after it cuts delivery outlook; Beijing extends tax breaks for electric cars

Li Auto said the cut to its forecast is a "direct consequence of the supply chain constraint." Meanwhile, China extended a tax break for electric car purchases.

www.cnbc.com

To help maintain growth for electric cars, China’s Ministry of Industry and Information Technology and Ministry of Finance extended the period that new energy vehicles will be exempt from a purchase tax until Dec. 31, 2023. New energy vehicles include fully electric as well as plug-in hybrid cars.

 

[Maarten]

Trading strategy that worked in past years may not work this year.

View attachment 856867

Average S&P500 performance for 2022 so far is that during the week following a one day 1% decline is another 1.2% decline.

Hope not to see more one day 1% declines this year. Too many already.



View attachment 856868
link

By..... WSJ.
Inventing another way to scare owners out of their TSLA shares.
Well, not me, thank you.
Let's see where we stand in Jan 2023 when the results because of the production of Q4 have been published!

 

[willow_hiller]

By..... WSJ.
Inventing another way to scare owners out of their TSLA shares.
Well, not me, thank you.
Let's see where we stand in Jan 2023 when the results because of the production of Q4 have been published!

Also, that sort of causal inference doesn't really work with daily changes in stock prices. If every 1% decline was followed by a 1.2% decline, then the 1.2% decline would be followed by a 1.214% decline, which would be followed by a 1.230% decline, etc etc. It just doesn't make logical sense in the context of a market that has no definitive short term trend. But we know there is a long term trend, and the best way to capitalize on that is holding through the 1.2% short term dips.

 

[elasalle]

That seems like a pretty low bar.

Also, natural biped walking must be awfully hard...I've yet to see a robot do it well.

the men in suits was low bar

Immediately after Elon announced Optimus, these guys put together/demoed an actual robot - (but we don't know if there was a man behind the curtain controlling his every move or he was full AI - if full AI, I would be impressed as they have some software as well as hardware (actuators etc ..). cheers!!

 

[Singuy]

Interesting data point. My kid's daycare owner just got a Lucid which he ordered 6 months ago, said it would be delivered in 60 days but took 6 months. He also had a Model X refresh on order and just got the notice that it's ready after 1.5 years. He of course is cancelling the model X because he got his Lucid.

I was blown away that Tesla made like 20x more Model X in the last 1.5 years than Lucid but still lost out on delivering it in time vs a Luvid. I was also surprised that the initial promise time for a Lucid was lower than any car from Tesla 6 months ago (when electric cars were at peak demand).

 

[Featsbeyond50]

Wall street might be in for a T-bone accident. Is that a stock thing?

I think you may have invented the analogy. I like it.
I'm also afraid you may be right. Either by design or a panic driven over reaction, the Fed is going to break the market. Housing market for sure.