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

[Usain]

I wonder if the input date to the FSD trainer includes the location of the driving being analyzed,

The data definitely includes the location. Remember when Tesla was having trouble with streetcars in (I think it was) Montreal? Tesla disabled FSD for that area until they had gathered lots of data from that small area to train the system.

 

[Buckminster]

https://twitter.com/x/status/1754259298123657725

 

[GhostSkater]

Just something random that I just noticed, Tesla does have different charging times for different regions depending on which cells are being used there, which is obvious they would but might be a good way to find out which battery pack goes where

For example, in the US Model 3 charges 51.3% in 15 minutes, in Europe it's 44.8%

Someone asked a while ago where from the vehicles for Chile will come, the percentage in 15 minutes there is 44.8%, so it the information is correct, they will be Shanghai made with LG cells

Weirdly enough, for Model Y is 39.7% in 15 minutes, unless now there is one more battery pack with LG cells into the mix that charges slower

 

[2daMoon]

For those who are into TA (or T&A)

It is always interesting to hear someone's professional perspective, but that interest has to be balanced with the aspect of TA being a kind of Voodoo.

Some of what he said seemed to be good guidelines, such as buying into an upward trend, rather than presuming the SP has bottomed at some point before actually demonstrating a pivot, only to buy in and see further downward movement. The mere fact that he thinks this way offers insight into his perspective. For a technical guy he came across fairly mild-mannered.

His angle on volume and momentum influences seemed okay, and these are always difficult to determine whether the SP followed the Volume or vice-versa. There is so much of the "hindsight is 20/20" feel that doesn't give me confidence in using this to make a prediction.

I'll hold onto his short-term guess that we might go into the 160s, then turn around by mid year. That's just confirmation bias, I know, but I like how it sounds. As for the potential of an ATH this year, well, that would be nice and all, but I'm not going to get blue in the face holding my breath. There is always the potential for an ATH in any year, isn't there?

Granted, with others prognosticating rate cuts in that time frame as well there could be some levers that get moved. With that in mind, Tesla is poised to accept a lot of volume buying due to where the SP has fallen and volume has as well, while company fundamentals remain strong. (even if Tesla isn't extending the spoon to the analysts who never learned how to math)

Thanks for sharing the link.

 

[Thekiwi]

Thumbs down received can be misleading. I rated this post as Informative as I wanted to see what the WSJ millennials were claiming about Tesla, its BOD, etc. Did I buy the FUD, No! But it got me behind the paywall. I read this Forum for information. I do not have to like it and maybe do not give it credence but it is important to know what's being broadcast to segments of the investing and buying public. This Forum can subsequently do a great job evaluating any such posts.

Yes I was merely attempting to summarize the article that I was providing the free link to, and I myself added the “alleged” to what the WSJ reported, and myself noted that even if Tesla was hypothetically in breach of NASDAQ listing rules they would still have plenty of time to rectify it (12 months).

Think some other users of this thread get their knickers in a twist about rather innocuous things sometimes.

 

[UncaNed]

Looking over today's early morning posts, I truly feel for all the poor, lonely FSD posts over in the FSD thread, and think they should all be moved over there where they can play and dance and live and be happy together with each other, and hopefully entice future FSD posts to appear there instead of in this main thread until Robotaxxis are a working thing.

Autonomous driving will happen when it happens, almost a decade away yet, and it's getting really tedious to hear the same old boilerplate posts saying how the new version has the problem licked, over and over and over again. Yes, I get that V12 is very, very different, but it still has a long, long way to go to drive with adequate nines, and enough added value without killing anyone, making wrong turns, and obliterating traffic patterns.

oI'm not trying to be negative; indeed, compared to the likes of practical nuclear fusion, "clean coal", and finding a cure for paralysis, FSD is, indeed, right around the corner.

 

[Thekiwi]

Ohhh "alledged" performance issues...


Seriously? Combining "fact" and "may"? Are they or aren't they? Here's a hint, the majority board has been in place over 12 months other than with JB replacing Hiro last year.

Want to research yourself (something WSJ seems to have not bothered to do since they can't tell if Tesla is compliant):
The Board: Corporate Governance | Tesla Investor Relations
The Rules: Nasdaq Listing Center

It’s the WSJ alleging that Tesla may be breaching the NASDAQ rules, not I.

That WSJ allegation is based on the whole articles contents attempting to imply that not enough of the tesla independent directors are actually independent (echoing the Delaware judge), and therefore do not meet NASDAQ listing requirements.

And as I said, even if NASDAQ decides this is the case, Tesla would still have ample time following notice from NASDAQ (12 months). I cant see how it is in anyway a threat to being delisted as WSJ notes is an outcome of that hypothetical.

The only way that would happen in this hypothetical situation is if Tesla refused to make any changes to its board, after failing to convince NASDAQ with a counter argument that it’s current board was compliant to listing rules.

 

[Thekiwi]

I guess they forgot to mention that he got tested for illegal substances by NASA/US govt once again…

As has been pointed out many times, one of the main drugs that Elon is often alleged to be abusing (ketamine) is a prescription medication, so wouldn’t be covered by a drug test result for illegal substances.

 

[mongo]

It’s the WSJ alleging that Tesla may be breaching the NASDAQ rules, not I.

That WSJ allegation is based on the whole articles contents attempting to imply that not enough of the tesla independent directors are actually independent (echoing the Delaware judge), and therefore do not meet NASDAQ listing requirements.

And as I said, even if NASDAQ decides this is the case, Tesla would still have ample time following notice from NASDAQ (12 months). I cant see how it is in anyway a threat to being delisted as WSJ notes is an outcome of that hypothetical.

The only way that would happen in this hypothetical situation is if Tesla refused to make any changes to its board, after failing to convince NASDAQ with a counter argument that it’s current board was compliant to listing rules.

Your post was the conduit through which the bull arrived to me, thus your post become the recipient of the bull I shoveled back.

 

[heltok]

That's not the way it works. Learning is cumulative. Just because FSD is better in San Francisco does not make it worse somewhere else.

Driving differently in different places is an interesting engineering problem. Tesla likely have thought about it a lot. After a few seconds of thinking about it I see at least 4 different ways of doing this.

1. Do nothing. The neural network will just have to learn to be an expert geoguessr to figure out which rules apply.

2. Input the GPS location to the neural network and let the neural network figure out by itself which rules apply where based on video data.
3. Figure out which rules are relevant based on GPS with software 1.0 and input a variable for which category of rules ie "rules category: [0 0 0 1]" if there are 4 regions.
4. Train a specific neural network for each region and have the car use the neural network and potentially switch between these during the drive if crossing borders.

I think Tesla might be doing 2 but might add 4 for China due to data transfer rules.

 

[Krugerrand]

Yes I was merely attempting to summarize the article that I was providing the free link to, and I myself added the “alleged” to what the WSJ reported, and myself noted that even if Tesla was hypothetically in breach of NASDAQ listing rules they would still have plenty of time to rectify it (12 months).

Think some other users of this thread get their knickers in a twist about rather innocuous things sometimes.

If you preface the link: ‘Just sharing what I read’ or some such then nobody has to guess where you stand and you don’t have to guess if their knickers are in a twist.

 

[lascavarian]

2. Input the GPS location to the neural network and let the neural network figure out by itself which rules apply where based on video data.

I would essentially agree but add that NN training should not rely on GPS.

Vision is the key training input. Even before GPS existed, a driver knew which rules likely applied simply by looking at the most common license plates recently observed or by the shape and format of traffic signs. Lots of simple visual clues available. NNs can derive amazing inferences.

 

[heltok]

I would essentially agree but add that NN training should not rely on GPS.

Vision is the key training input. Even before GPS existed, a driver knew which rules likely applied simply by looking at the most common license plates recently observed or by the shape and format of traffic signs. Lots of simple visual clues available. NNs can derive amazing inferences.

Basically think of the NN as an expert driver that just has access to 30s of video data. For most situations this will be okay, but sometimes knowing approximately where you are will help you greatly. Often you can figure out where you are, but there will be times when guessing if you are in turn on red or not area will be tricky.

Adding another sensor adds failure modes. But without the GPS available, navigation will be very tricky anyway, so you might aswell assume it will be available but add some data where you remove it to teach it to drive acceptable without it.

Compared to the vision the amount of data is very low, two doubles vs millions of ints. It cost very little to input it to the neural network in extra compute/bandwidth.

 

[Featsbeyond50]

Looking over today's early morning posts, I truly feel for all the poor, lonely FSD posts over in the FSD thread, and think they should all be moved over there where they can play and dance and live and be happy together with each other, and hopefully entice future FSD posts to appear there instead of in this main thread until Robotaxxis are a working thing.

Autonomous driving will happen when it happens, almost a decade away yet, and it's getting really tedious to hear the same old boilerplate posts saying how the new version has the problem licked, over and over and over again. Yes, I get that V12 is very, very different, but it still has a long, long way to go to drive with adequate nines, and enough added value without killing anyone, making wrong turns, and obliterating traffic patterns.

oI'm not trying to be negative; indeed, compared to the likes of practical nuclear fusion, "clean coal", and finding a cure for paralysis, FSD is, indeed, right around the corner.

If FSD is a decade off, we've got our bets on the wrong horse. Waymo will have the market by then.

 

[MC3OZ]

Basically think of the NN as an expert driver that just has access to 30s of video data. For most situations this will be okay, but sometimes knowing approximately where you are will help you greatly. Often you can figure out where you are, but there will be times when guessing if you are in turn on red or not area will be tricky.

Adding another sensor adds failure modes. But without the GPS available, navigation will be very tricky anyway, so you might aswell assume it will be available but add some data where you remove it to teach it to drive acceptable without it.

Compared to the vision the amount of data is very low, two doubles vs millions of ints. It cost very little to input it to the neural network in extra compute/bandwidth.

I think FSD has a high level map and compares what it sees to the map, sometimes updating the map.

What it sees always take priority over the map... but if the majority of high rated drivers also turn left on red in a particular location, in the absence of any other visual clue, I think the car will perhaps assume that is what it should be doing... When there is no map data, what it sees is the only consideration...

Map data is most useful when vision is reduced...

So driving in a snow storm, with no lead car, and no map data, because no other Tesla has ever driven that route, may be problematic...

There is a difference between map data as a guide, and map data as an essential component,

Often what trusted safe drivers do in response to a particular set of visual inputs is the right thing to do, regardless of the technicalities of road rules. When safe drivers don't obey the letter of the law, the law might be sub-optimal.

In relation to judging V12 progress, I'll reserve judgement until Chuck Cook puts it though a series of tests on unprotected left turns, narrow roads etc, or I get it, and can test it myself. My hunch is Chuck will have it long before I do, and he will quickly find any problems,

If Chuck finds problems (and he probably will) the next questions are how fast FSD improves and if it regresses in some areas. What I expect is Chuck will find problems, but we should see a steady rate of improvement with no regressions... Any timeline predictions I may do, will be after I see Chuck test the next 3 versions....

 

[evermore]

Learning why Mizuno left the board was worth reading the WSJ article.

 

[theschnell]

Guys, I hope the mods will stand for one weekend off topic post.

The quoted post below is from the “off topic galore” thread. If interested, please DO NOT reply here. Please go to that thread to reply. I appreciate advice from anyone interested.

Guys I need your advice.

I assembled a 50hp electric pontoon boat with the idea to start a nascent market for electric boats. I also added solar panels on the upper deck, although if I were doing it again, I would skip this, but live and learn. Anyway, my challenge it’s one thing to assemble a boat like this, and a whole other thing to know how to market it, assuming there is a market for it. Any advice?

Here is a link to my YouTube video where I was trying to beat the Guinness record for the longest solar powered distance by a boat in 24 hrs. That attempt failed miserably primarily because it was attempted in September and the days were already too short, although it would be tough to win even on the longest day of the year, but there would at least be a chance then.

View attachment 1015356

Anyway, any advice or ideas are appreciated.

 

[Artful Dodger]

in Brazil we're anxious and slightly dismayed to see our nice little neighbor get preference

I suspect Tesla needs to balance large exports (ie: copper) purchased in Chilean peso (CLP) with imported finished goods in USD.

 

[Knightshade]

I need to be more clear as well. I am assuming that Tesla engineers know what they are doing. In the case of Tesla FSD, overfitting will make it better in a geofenced area without making it worse somewhere else.

That's.... not really how that works.

The less hard code there is the less able they are to tell if training in one area will cause a regression in another without extensive testing in every area.

What it sees always take priority over the map...

Absolutely untrue.

For over a year FSDb would try and make a left at an intersection it was not physically possible to make a left at becuause construction had changed the shape of the intersection-- because the map hadn't been updated. It was visually obvious this manouver couldn't be done but it'd keep trying until a map update a month or two ago.

The current map FSD always stops at a stop sign on the map in my neighborhood. Except there is no stop sign there, and never has been.

And as you said, the problem might be hard, but it is solvable with end-to-end. For your turn on red example, the training data includes the location of the clip it is learning from. The system will learn not to turn right on red in places where that is against the law.

It might also need to learn what day it is because the law can change based on that. Or what time it is because the law can change based on that. Or if it's a holiday because the law can change based on that.

Further in/out of city limits might change the law-- so you'd need a LOT of very specific data in every possible location before it gets it right consistently.

Now you can avoid SOME of these issues by just telling it, for example, NEVER make a right (or left) on red, anywhere ever. But then people will complain about how it's driving in places you're expected to do so and it'll annoy other drivers.

And it won't fit other unique-by-time/date/location issues.... Parking for example.... sometimes it's legal to park certain places, sometimes it's not, and that can vary by not just location, but time/date/etc.... you can't fix that by just saying "never park"

SOME places will have signs for this (but then you get the issue that HW3 cameras aren't really good enough to read small print signs- which means RT is never coming to like 90% of the current fleet if you're relying on signs to fix this) but some places still won't.



Anyway, market opens in the morning, prob want to take further on this here:

AI, Autopilot, & Autonomous/FSD

Discussion about AI, Tesla Bot, Tesla Autopilot (AP), the promise of Full Self Driving (FSD), as well as other Autonomous Vehicles.

teslamotorsclub.com

 

[Usain]

Yes, I get that V12 is very, very different, but it still has a long, long way to go to drive with adequate nines, and enough added value without killing anyone, making wrong turns, and obliterating traffic patterns.

I am presenting an investment thesis. I'm arguing that V12 really doesn't have a long, long way to go. That's why I've been talking about it here in this thread.

My thesis is that Tesla will, by necessity, start its robotaxi business in a geofenced area. The area they pick will be the city where FSD is already performing the best. By oversampling specifically for that area, FSD will quickly improve for the chosen city and they can have a version of V12 that is ready to start testing a robotaxi service within a year or so.

And that will move the stock higher. Possibly MUCH higher.