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

[Stretch2727]

I agree. A longer battery and drivetrain warranty would help the next wave of adopters make the switch. Playing devil’s advocate, “If electric cars are so great, why isn’t the drivetrain warranty better than what I can get on an ICE vehicle?”

I agree longer warranty would be beneficial but most ICE powertrain warranties are 5 years/60K miles and they don't cover a lot of ancillary stuff on an ICE that is likely to go bad (sensors, connectors, hoses, cooling).

Tesla is already above this at a minimum of 8 years/100K miles minimum for powertrain and battery. Some models are 8 years/150K miles.

 

[jerry33]

What vintage vehicle?

Thae pack on my early 2013 S went at about then too, and I remember hearing a few such cases. Got a refurb pack under warranty, and it still plugging away at 180K miles...

(Although it is telling me I need a fuse replacement, but that's only a couple hundred bucks I understand...)

On my early 2013 S, the battery was still fine at 130,000 miles seven years later. I have no doubt it is still going strong. Note: B battery

 

[Linkmeup]

I just learned about the progress of StarFactory on Youtube...

...its incredible to even be associated, to one of the companies of Elon Musk and only as a public market investor, to the teams building up the future like this. I wish I was an investor in more Elon Musk companies outside of Tesla!

 

[Stretch2727]

This is one of the most important posts on this forum given the high importance FSD robotaxis has suddenly become in supporting any future apprecation of share price, so it's surprising there's not much discussion.

You have given some good evidence for why it's quite likely the models are going to grow much bigger to support lower error rates / lower disengagements. There will probably be advances in ability to distill these big models into something smaller that works mostly as well but still... this a serious issue with inference compute. If Tesla eventually trains a model with high enough fidelity for robotaxi, it may be orders of magnitude too big to fit on HW3, HW4, or who knows, HW5.

This is a very serious risk.

I'm not an expert here but James Douma seems to think they can squeeze efficiency out of the NN with software engineering optimization as they learn and gives examples of this. It's a long video but the core of his argument starts 20:24.

 

[Blue horseshoe]

I’ve been trying to think of a company that locks critical public safety features behind a software toggle and monthly subscription but am not having any luck, can you think of something comparable?

To be honest I think the public has a stake in all this, because what’s being done with FSD today wouldn’t be possible without all the data coming in from drivers, from the infrastructure, even from non-Tesla drivers on the road who are interacting with the vehicles. Data collection in vehicles will probably become a much bigger topic in coming years.

But I don’t think arguing about this really matters because I genuinely believe the company’s leadership wouldn’t lock safety critical stuff out like this, and I don’t believe Elon would do that. And if/when the leadership did just include it by default, people will flip and herald it as a great move anyways lol

I think the appropriate analogy to this is the pharmaceutical industry. They sell life saving drugs and even with covid when the initial belief was that it was a potentially catastrophic pandemic, there was no imminent domain type situation where companies were forces to provide products for free due to public safety

 

[insaneoctane]

Free OTA oil changes for life...

FIFY

 

[Andy O]

Um - where do you think governments get their money?

Answer: The taxpayers

So, instead of all tax payers being mad at Tesla owners, why not just let Tesla owners pay for the FSD and they’ve made the roads safer for themselves and everyone else anyway.

I know where you’re going next - and so I say to you, Tesla has already said they were willing to license their tech to other OEMs.

With this whole line of thought, it’s people with a solution looking for a problem. Not to mention, literally overnight the narrative went from ‘FSD won’t happen/won’t happen for years’ to ‘everyone should get FSD for free’.

Another way that governments get their money is by reducing costs. At the high cost of a motor accident, governments might find they actually save money by buying FSD safety features for everyone. As for all taxpayers being angry at Tesla owners, well, they might be angry at the government but many would say good, for this increase in taxes, I am safer on the road as a pedestrian, bicyclist, child, or driver.

 

[wtlloyd]

To whom though?

Business people who do work on their phones and laptops probably, but certainly not most people like nurses, electricians, machine operators, plumbers, etc. Most people would simply watch Netflix or play games while their car drives them around, or even sleep, and for those people I doubt having a virtual chauffer is worth $12K or more.

Now, would it be worth $99/mo? Or even $49/mo? Probably,

Now don't get me wrong, FSD is going to make Tesla lots of money. I just don't think its going to be anywhere near as super lucrative as most TSLA bulls seem to want.

I expect that Insurance companies will have the final say.

 

[SPadival]

Not convinced by this. SAAS companies deserve a higher multiple because they may scale very quickly without any incremental costs. If we are talking FSD for Teslas alone this is not the case, since the max number of FSD subscription is still the same as the number of Teslas with the right hardware. So the long term growth potential should be the same as sold cars (until and unless they license FSD to others). Also, I am not sure about the take rate (even at 99 USD). It seems a lot of people don't even want to buy a car if the monthly cost is slightly higher due to interests rates, which makes me wonder how many are prepared to pay over a thousand USD per year just for added convenience. It seems to me a "nice to have" but not a "need to have". Lastly, I cant see how his excel is set up and I didn't double check it but just from how he describes it, it seems he mixes revenues with profits.

This is just additional income until robotaxis are launched at scale (not just unveiled). Also, he has a margin of 90%. So whether it is revenue or profit is marginal.

Aaand..

https://twitter.com/x/status/1778881361249800203

 

[capster]

FSD 12.3 4 downloaded on 2021MXLR. It seems to have solved the driveway test. I am now quite gruntled again.

Be less wrong, be more gruntled. Seems like a pretty good recipe for life.

 

[Stormy24]

https://twitter.com/x/status/1778877155944099931

 

[unk45]

Um - where do you think governments get their money?

Answer: The taxpayers

So, instead of all tax payers being mad at Tesla owners, why not just let Tesla owners pay for the FSD and they’ve made the roads safer for themselves and everyone else anyway.

I know where you’re going next - and so I say to you, Tesla has already said they were willing to license their tech to other OEMs.

With this whole line of thought, it’s people with a solution looking for a problem. Not to mention, literally overnight the narrative went from ‘FSD won’t happen/won’t happen for years’ to ‘everyone should get FSD for free’.

An ever so sarcastic and unjustified by any logic but still irresistible correction:
No government actually "gets their money for taxpayers". Those governments themselves creat the money coin it, print it, multiply it as they choose. Taxpayers don't do that, it's just a way of making them have fewer goods and services so the governments can have more fo them. Today, there is zero backing for the US Government other than the government itself. Within living memory of some, like my 99 year-old mother-in-law, it was not that way. The US$ then was backed by silver and a US dollar was a 'silver certificate'. Still valid g'gold certificates' continue to exist. Richard Nixon finally nixed backing...and nobody much noticed!
(I'm sorry about this. It's Friday night so my mind wanders. (FWIW I was in graduate school the day he did that, in a seminar with an exceptional Gentleman named Scott E Pardee of FOMC fame, one of the most memorable experiences in my life.)

 

[capster]

https://twitter.com/x/status/1778877155944099931

Game, set, match.

 

[unk45]

This is one of the most important posts on this forum given the high importance FSD robotaxis has suddenly become in supporting any future apprecation of share price, so it's surprising there's not much discussion.

You have given some good evidence for why it's quite likely the models are going to grow much bigger to support lower error rates / lower disengagements. There will probably be advances in ability to distill these big models into something smaller that works mostly as well but still... this a serious issue with inference compute. If Tesla eventually trains a model with high enough fidelity for robotaxi, it may be orders of magnitude too big to fit on HW3, HW4, or who knows, HW5.

This is a very serious risk.

There are other risks wrt to training: How hard will it really be to improve performance 100x?

There are papers and studies around "neural scaling laws". They look at how much increase in compute, data, and model size affects the abiliy to lower error rates.

Here's one on vision transformers (on images) which could relate reasonably well with FSD.

https://openaccess.thecvf.com/content/CVPR2022/papers/Zhai_Scaling_Vision_Transformers_CVPR_2022_paper.pdf

Here's the thing, even on a log scale, as you increase the compute, data, and model size, you get sublinear returns in error rate improvement.

I.e. it becomes harder and harder to reduce the error rate further. The model size, data size, and compute needed to get crit disengagements (CD) to 3000 miles / CD may be a 10x increase, but then to get from 3000/CD to 30000/CD may take more than another 10x increase, it might be another 30x increase in data / compute / model size.

Tesla has the ability to scale up data and compute some, but not these orders of magnitudes in a "short" time frame of a few years (let alone by August).

Point being, there's a lot of empirical data hinting that training FSD to a robotaxi level could take a lot longer than people think.

View attachment 1038036


View attachment 1038037


View attachment 1038040

Will geofencing reduce the size escalation rate? With everything I ever did, too long ago to be relevant, reducing the scope of any model, by defining and much as can be defined, reduces the model complexity. Although in AI when I was doing it, that was very primitive as was LISP, our only usable language at the time.

So, is that still true. Could Tesla simply use pre defined elements, and by geofencing reduce the unknowns in many cases. Of course those compromises might not matter were Robotaxi operate, as logic suggest it will, in specified geographical limits for each trip. That would make distinct version substitutions in every major market. They already do that with maps, although now in quite huge continental-sized chunks.

 

[GrandEnigma]

Welp, time to modify my order I am picking up Monday to remove FSD... no point in paying $12k for FSD that now has a 10yr ROI. Interesting change as you would think upfront cash would be better than sub revenue, but I guess they are hoping for a greater uptake, which well could be the case.

 

[Linkmeup]

Welp, time to modify my order I am picking up Monday to remove FSD... no point in paying $12k for FSD that now has a 10yr ROI. Interesting change as you would think upfront cash would be better than sub revenue, but I guess they are hoping for a greater uptake, which well could be the case.

I just thought the exact same thing lol

 

[Musskiah]

This is the hardest traffic circle in NE Ohio, perhaps all of Ohio.

Lander Circle is 3 lanes, unmarked with a bus stop IN the circle (N.E.), 5 roads from the N.,S.,E.,W.&S.E. A gas station entrance(N.W.), a church entrance (S.W.), a bank entrance(S.E.), and a grocery store entrance (N.E.). It's hectic in the day time.

V11 (every version) was not fit to handle it more than every once in awhile, and it wasn't chill. Just imagine what happened when I took V12.3 through it FIVE consecutive times. Ladies & Gentlemen, I am thoroughly convinced V12.3 did better than most any human. I would describe it as confident, effortless, and if it continues to act as it did in the circle today...dare I saw...flawless!View attachment 1031214

This is bad, this is really bad...

FSD V12.3.3 on MYP has been nailing this traffic circle day after day. I have done it dozens of times in the last couple weeks from all different approaches to all different exits.

So I figured, I should take MXLR with V12.3.4 to give it a whirl. First time around, it was even smoother than V12.3.3 and I was quite excited.

Then, on my next approach into the circle (coming from the left on the image I posted) the unthinkable happened; the LEFT TURN SIGNAL came on as the MX awaited a clear entry. A few cars passed and when the coast was clear from the left, it began to enter the circle in a CLOCKWISE fashion - the WRONG WAY!!! I intervened with much Chagrin! Get it? I was on Chagrin Blvd...nevermind...anyhoo...

Now, my route planner had me going to that "bp" entrance off the circle in the upper left corner of the image, so undoubtedly, it thought it could just take a quick shortcut in and out of the circle instead of traversing 330 degrees of the entire circle. In my younger days, before my prefrontal cortex was fully developed, I may have impatiently attempted the same if the coast was clear.

I have no doubt Tesla will relentlessly feed Dojo traffic circle tokens in order to avoid this type of behavior in future versions.

The question is: How long might it take to ensure this behavior occurs very very veeeery infrequently? As we all know, FSD just has to be much better than humans, and many-a-human has gone the wrong way down one ways, roadways, highways, and traffic circles.

Sadly, I am no longer gruntled (that didn't take very long). I anxiously await V12.4...

 

[2daMoon]

An ever so sarcastic and unjustified by any logic but still irresistible correction:
No government actually "gets their money for taxpayers". Those governments themselves creat the money coin it, print it, multiply it as they choose. Taxpayers don't do that, it's just a way of making them have fewer goods and services so the governments can have more fo them. Today, there is zero backing for the US Government other than the government itself. Within living memory of some, like my 99 year-old mother-in-law, it was not that way. The US$ then was backed by silver and a US dollar was a 'silver certificate'. Still valid g'gold certificates' continue to exist. Richard Nixon finally nixed backing...and nobody much noticed!
(I'm sorry about this. It's Friday night so my mind wanders. (FWIW I was in graduate school the day he did that, in a seminar with an exceptional Gentleman named Scott E Pardee of FOMC fame, one of the most memorable experiences in my life.)

I concur, and firmly believe that the only purpose for Income Tax is for use as a vacuum cleaner to suck vast quantities of inflated currency out of the system in order to offset the negative effect of the money creation side of this system.

The Federal Reserve and the Internal Revenue Service are private entities, both of which were created in 1913. They were designed to work in concert with one another to manage the money supply in the hope that they can keep the balls being juggled in the air. No fiat currency has ever stood the test of time, but some have lasted a few centuries.

Hopefully, the abundance on the horizon from drastically reduced energy production will bring a positive effect toward making a flawed monetary system last a little longer than it might otherwise have.

Tesla specifically, and Elon generally, are riding the wave of this convergence of technologies that with any luck will lead to change that will save our bacon.

 

[Linkmeup]

TSLA jumped ~1% AH

 

[Krugerrand]

Another way that governments get their money is by reducing costs.

Ok the only way I can respond to this is

At the high cost of a motor accident, governments might find they actually save money by buying FSD safety features for everyone. As for all taxpayers being angry at Tesla owners, well, they might be angry at the government but many would say good, for this increase in taxes, I am safer on the road as a pedestrian, bicyclist, child, or driver.

Again,

I’m not laughing at you or your ‘ideal’. I’m laughing that these are ever possible scenarios of today’s governments.