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

[StealthP3D]

No legacy OEM would put a 1% failure rate L5 system into service. Their tolerance for risk is FAR lower than Teslas.

It's a misconception that Tesla has a higher tolerance for risking the lives or safety of their customers compared to the legacy auto companies. Actions speak louder than words.

Tesla has a higher tolerance for business risk, meaning they are more likely to take bold moves that risk failure, ie. the Cybertruck would never be released by a legacy manufacturer. But you are just plain wrong when you claim Tesla has a higher tolerance for risking the safety or well-being of their customers. If anything, Tesla, as a company, takes this more seriously than all the legacy manufacturers combined. This is demonstrated, not by words but, by actions:

1) Tesla vehicles are designed to protect the occupants to a higher standard than just about any legacy manufacture spends the research time and effort to achieve. This goes beyond the fact that they don't have an engine as evidenced by the crash test results and injury/death rates of the products of GM, Nissan, Toyota, etc., all who have produced EV's that protect the passengers to lesser standards compared to Tesla products. In some cases a lot less. This is no accident. Pun intended. Accident and injury statistics support this objectively.

2) Tesla's core mission is to protect the one planet, the only planet that provides a safe living environment for all humans alive today. If that's not caring about the safety of people, I don't know what is.

3) Tesla has put huge effort into having the safest and most advanced driver aids available. These aids make the car more safe than the competition and reams of data support this in an objective manner.

4) Not to crap on legacy auto but they have a long and sordid history of putting their customers lives at risk for the sake of a few bucks. Ford, GM, Toyota, etc. I won't go into the specifics but suffice to say, it has been repeatedly well documented how these companies knew defects that risked their customers lives were present but they decided it would cost less to ignore the problem or even cover it up rather than spend the money to eliminate the danger. This has cost the lives of at least hundreds of thousands of innocent people.

The narrative that Tesla plays fast and loose with road safety is a false narrative propagated by those set on destroying or slowing Tesla down. It appears that either you are gullible enough to have swallowed this fish tale hook, line and sinker or you are knowingly participating in the movement to bring Tesla down or, at the very least, slow them down. I don't claim to know which it is but simple logic says it must be one or the other. If the latter case, you are the one playing fast and loose with human safety. Because hundreds of thousands (at a minimum) die every year from from the negative impacts of fossil fuel consumption and poorly designed automobiles of legacy manufacture, people who would have lived if they were in a Tesla. Again, this is not empty rhetoric or fanboy irrationality, it is fact supported by reams of crash data, thousands of well-documented court cases, testimony from senior engineers at legacy auto plants, and the list goes on.

I know these are strong words but it pains me to see this forum, a forum that aspires to assist the transition to safer, more sustainable energy and transport, be used, either inadvertently or on purpose, to slow the transition down. It's not empty corporate rhetoric that Tesla cares deeply how their products affect the safety and well-being of their customers, it's supported by their actions and deeds and backed up by actual data.

In other words, no, Tesla would not put a L5 system into general usage that had a 1% failure rate. That's absurd!

 

[LYTRIDR]

Part 3 of James Douma valuation.

I really don’t get it. Tesla would basically be handing over most of the profits to fleet operators. I cannot see Tesla doing that. Much more likely would be a high upfront cost (e.g. every Tesla sold has a mandatory > $30k FSD package) PLUS Tesla gets 30% of the revenue, or Tesla operates their own fleet and keeps ALL the profit. Tesla has never ceded that kind of profit to any supplier or partner.

in James’ model a new Tesla buyer could purchase a $40k Tesla and earn $77k revenue which is mostly profit, the 1st year alone. Those money making opportunities just DO NOT exist in the real world.

Would love other people’s take. One other small nit is that he showed Global production, but then the rest of the calculations were on U.S. alone.

I see the model similar to you, my guess is $40-50,000 for FSD fleet access plus 30-40% of per ride revenue to Tesla. On the other side, fleet operators paying for cleaning, maintenance, charging, parking, financing, and everything else will still be nicely profitable, just not $77k per year...

I do wonder if FSD in the future will have a separate line item for personal use vs fleet access. Seems that the value proposition is wildly different and the take rates will be quite low for personal use at $30,000 or more.

 

[ZeApelido]

OT:

What does this do to your brain?

Spoiler

 

[JRP3]

What does this do to your brain?

Elaine?

 

[lafrisbee]

Ok...Ant Time
As an ant I can see more smell more touch more...and
I see a better move for Elon with FSD. Someone mentioned Bluetooth type domination and licensing. And I agree with that.
Elon agrees with me too. He wants to save the world. And by licensing FSD to all the other BEV makers he would accelerate his goal of a cleaner world...and he could crush ICE and PHEVs production by never licensing FSD to that sector. This would clearly accelerate the adoption of BEVs and the abandonment of ICE vehicles. Winner Winner Chicken Dinner...Or CHicken Picnic to us ants....
Cuz man oh man (ant oh ant), TSLA would become unpurchasable. Think. TSLA would sell FSD Software on a microsoft level but for $2000 a copy and a subscription attach to it as well....and then there is the Robo-taxi money generated from it as well.
Even Elon would have to start giving huge dividend checks. it would be elon's world. Fellow Ants Rejoice! Everyday would be a picnic!

 

[bpjod]

Haven't seen the comical FUD posted here yet from Road and Track. Won't link, but this is the level of hard hitting journalism we expect ;"their report on a 13-minute video posted to YouTube".

Tesla's 'Full Self-Driving' Beta Called 'Laughably Bad and Potentially Dangerous' ​

Hey R&T, let me FTFY:

Tesla's 'Full Self-Driving' Beta Called The Average Human Driver is 'Laughably Bad and Potentially Dangerous'. Half are worse. ​

 

[Rarity]

Yes, Karpathy has said that the best vision NN algorithms are available in research papers, but I'm not sure if others can catch up quickly on data curation. It seems Tesla AP team spends much of their energy on data curation and the in-house labeling software workflow. Labeling itself requires a lot of discipline and standardization.

Perhaps others will copy the approach, but they need to first copy the sensor suite, the software workflow, and then train disciplined labelers, etc. Then they'd have to collect and programmatically address all the edge cases again. This sounds like a job for a startup, not lazy slow moving OEMs that still fail at OTA updates.

And most importantly, they have to copy the scale, as Karpathy emphasized in his recent podcast interview. Amazingly, Tesla just keeps building its lead there.

From Tesla's perspective, it's safest to assume that somebody will successfully copy them pretty quickly (say, three years). But from an investor's perspective, we should look at the range of possibilities.

 

[printf42]

Imagine how fast they'd have done it if they already knew the right solution at the start and had a fleet 10-20 times bigger when data gathering began.

That’s under a big assumption that they can be Andrej working for them, that’s just impossible with their compensation structure and company mission.
Maybe some start ups could, but I couldn’t see it happening for any of the OEM.

 

[Mike Smith]

Elon's tweet was no mistake. Consider:

1) There was nothing ambiguous about the tweet. The conversation was about Tesla becoming the biggest company and he said quite clearly that it would probably happen in a few months.

2) The 'mis-tweet' theory makes no sense. You'd have to believe that first Elon accidentally replied to the wrong tweet, which is very unlikely to begin with, but practically impossible considering the response to the wrong tweet actually made sense. He then quickly deleted the tweet because he knew it would be a serious problem if people thought he was saying Tesla's stock price would probably quadruple in the next few months. He then saw that the tweet was spreading widely on twitter and in the media anyway, but for some reason didn't clarify that he had mis-tweeted?? That doesn't make any sense.

 

[Oveeus]

Part 3 of James Douma valuation.

I really don’t get it. Tesla would basically be handing over most of the profits to fleet operators. I cannot see Tesla doing that. Much more likely would be a high upfront cost (e.g. every Tesla sold has a mandatory > $30k FSD package) PLUS Tesla gets 30% of the revenue, or Tesla operates their own fleet and keeps ALL the profit. Tesla has never ceded that kind of profit to any supplier or partner.

in James’ model a new Tesla buyer could purchase a $40k Tesla and earn $77k revenue which is mostly profit, the 1st year alone. Those money making opportunities just DO NOT exist in the real world.

Would love other people’s take. One other small nit is that he showed Global production, but then the rest of the calculations were on U.S. alone.

The idea for robo-taxi is simple. People and even funds would simply buy fleets worth of car. I'd think that the "snake" charger prototype that Tesla has shown a few years back would make into production once robo taxi is a reality. So the car can do everything automatically like a Roomba in your house where it'd go back to the charging station by itself once done.

And the incredible part of AI anything (be it robo-taxi, or quasi-human level AI) is that as long as it's powered, it can roam by itself with minimum intervention. I could just send my Model 3 out to be robo taxi after riding from home to work and it'd be right there waiting for me by the time I'm off work. And once getting home, it would go out to work again and be home the next morning to drive me to work.

And Tesla can and will have their own fleet too, however, it might not be in their best interest to own ALL of them because there's still liability issue. Say if a car breaks down or simply how much capital needs to be locked up in order to have those out. Instead, Tesla can simply own the platform. Make 30% cut off every trip and still sell as many cars it manages to produce. No liability, no fund to lock and pure $$$.

 

[CorneliusXX]

Elon's tweet was no mistake. Consider:

1) There was nothing ambiguous about the tweet. The conversation was about Tesla becoming the biggest company and he said quite clearly that it would probably happen in a few months.

2) The 'mis-tweet' theory makes no sense. You'd have to believe that first Elon accidentally replied to the wrong tweet, which is very unlikely to begin with, but practically impossible considering the response to the wrong tweet actually made sense. He then quickly deleted the tweet because he knew it would be a serious problem if people thought he was saying Tesla's stock price would probably quadruple in the next few months. He then saw that the tweet was spreading widely on twitter and in the media anyway, but for some reason didn't clarify that he had mis-tweeted?? That doesn't make any sense.

Also, if he did mis-tweet, why didn't he reply to the correct tweet with the same reply.

 

[lklundin]

4 years for Soviets to steal and copy A bomb.

OK, let's compare the Manhattan Project with Tesla's FSD project, from the point of view of copying:
1) the main principle behind the fission bomb was available in the open literature (Leo Szilard patented it before the war - for a reactor), much like Neural Network processing of image data,
2) The actual details of the two very different designs developed in Los Alamos very actually not that complicated, there we challenges in proving that they would indeed work[*], but the actual specifics of the designs were quite simple, enough for a nuclear scientist to keep in their head. The software needed for FSD is significantly more complicated, but stealing software today can be done quickly too by a single, trusted person.
3) Getting the uranium ore and building the machines purifying the required isotopes and the production of plutonium (which is not found in nature) was actually the really hard and time consuming part of getting to the fission bomb. Once the USSR had acquired the designs, getting the fissile material is what actually took time for them. This can be compared to the configuration of the neural network that runs the Tesla FSD. These weights - which are updated over time - are in principle stored in every Tesla with FSD - but we know that Tesla went to efforts to cryptographically ensure that no one else can modify a Tesla's configuration of its FSD - so probably they also made it equally difficult to extract this data from the car - and they probably guard their in-house copies well, even though a given snapshot will be of limited use to a competitor. As for a competitor to generate similar data, well they would need a large fleet of driving, sensor equipped cars, which is not easy to come by.

So in some sense, the comparison between the Manhatttan Project and Tesla's FSD is relevant - but it does not seem like espionage or reverse engineering will help that much in the case of Tesla's FSD, partly because it is an ever evolving product and a snapshot does not help anyone to take the next step themselves.

[*] The first Atomic bomb dropped over Japan was of the design validated empirically 3 weeks before at the Trinity site. The second Atomic Bomb dropped over Japan was a completely different design - using a different fissile material - and cynics argue to this day, that this bomb was dropped not so much to speed up Japan's surrender, but to actually validate the design (on the enemy's population)...

PS. In my "analysis", I did not mention the FSD hardware. While it is truly excellent - especially taking into account size and power usage - I think something capable of supporting the required NN would be relatively easy to make for someone else (in time also with similar small size and power consumption), e.g. Nvidia.

 

[MABMAB]

There is a >0% chance I will have a board seat on TSLA as the 2nd largest shareholder in the next few months.

One of the great things I've observed with Tesla as a company is the near continuous positive expectancy of something more and greater and life changing in the future, over and over again. If this intentional on the part of the leadership they are doing a good job of it. Early on it felt great to be drinking that Kool Aid, but by gawd it does get old. The air of mystery has gotten to be smokey. The delays of everything are like some youthfully aware tease playing the lads for trinkets, keeping them along the desired path.

If my manservant social media editor (come on, you all saw him in the Gene Wilder house interviews) does not approve this message I will delete the above comments after a bit, adding all the more to the intrigue!

I now find it much easier (psychically) to ignore all the near term expectations, take a step back, live a little life, and follow the big picture changes.

 

[Knightshade]

In other words, no, Tesla would not put a L5 system into general usage that had a 1% failure rate. That's absurd!

It's so absurd in fact that literally nobody here including in the post to which you are replying, claimed they would.

It was in fact about someone else suggesting a legacy OEM would do so in a rush to compete with a 99.9% safe Tesla RT system.

So while I appreciate the many paragraphs you wrote- it was about an argument nobody was having.

 

[Knightshade]

My numbers are arbitrary of course but assuming that the end result is the same. Either they release it and Tesla is the far safer product (initially), or they wait even longer to release.

What is their hurry that would prompt them to put a significantly less safe commercial product on the road?

Does VW have some massive GET ROBOTAXIS TO MARKET RIGHT AWAY initiative I haven't read about?

Unless Tesla can make infinite cars, which we know they can't, legacy OEMs will continue to have customers for a good while to come even if they can't immediately serve as RTs (and there'll certainly be plenty of folks who have no desire, or ability, to run their cars at RTs anyway).

How long did it take Barnes and Nobles to realize that Amazon's model was superior? Or for Nokia/Blackberry to realize that about Apple? Or for legacy cable providers, content creators (Hulu and Disney+), and video rental companies to see that about Netflix? Legacy companies tend to discount the threat at first, then dither about and try to keep a foot in each space, and then try to move aggressively after it's way too late.

Running a rideshare service, and being a vehicle OEM, have been entirely different businesses up to this point though.... so that comparison doesn't make a ton of sense.


A better one is some Legacy OEMs are realizing building your own battery factories is a crucial piece of transitioning to EVs... GM and VW for example appear to be taking this lesson from Tesla to heart.... (Ford continues to "think it over"....)

We still haven't seen the other players in autonomy ditch their models and try to copy Tesla. When that happens it's a minimum of several years before they can hope to catch up.

Yes, it is... it's just not the 10+ years some folks seem to think (or at least there's no reason for it to be... if Tesla gets there this year, as I mentioned, it means they went from NOTHING to DONE in ~5 years...including a number of dead ends along the way.... if you only count the time since they realized they needed to re-write everything for 4D/video/BEV, it's more like 2 years.


There's no reason an OEM with a large fleet shouldn't be able to duplicate that in another 5 or less once the "final" version is a proven solution.

Licensing is possible but it's more complicated than something like bluetooth. FSD will depend on the specific hardware stack that Tesla uses so those items will need to be incorporated, chips, cameras, etc.

It doesn't seem all that much more complicated... it's not like the sensor stack is some SUPER ADVANCED UNOBTAINIUM hardware... for the most part it's 5 year old off the shelf cameras, radar, and ultrasonics.

The only really "special" part is the FSD computer, and there's no reason they can't license that and let Samsung just make more in the same factory.

The software is the real secret sauce.

I guess Tesla could offer a company like VW training services to train the core NN using VW's hardware suite. Tesla would need to be very production limited to spend that effort though.

This might actually be an interesting way for Tesla to essentially get free next-gen-HW testing en masse.


For example some have suggesting better located parking cameras low on the car would be an improvement- or side-looking front bumper/fender cams for corner peeking when the B-pillar cams can't see well... let's say VW is going to include one or more of those on their licensed FSD.

Tesla runs the extra training stuff on Dojo for VW.... but ALSO gets to use it themselves as part of the license agreement so if they add such cameras to their own cars later, they're already trained.

If it turns out it's not worth the HW cost, they used VW $ to find that out. If it does turn out worth it- they still used VW $ to find that out

 

[EVMeister]

It's so absurd in fact that literally nobody here including in the post to which you are replying, claimed they would.

It was in fact about someone else suggesting a legacy OEM would do so in a rush to compete with a 99.9% safe Tesla RT system.

So while I appreciate the many paragraphs you wrote- it was about an argument nobody was having.

Actually, I read your comment "No legacy OEM would put a 1% failure rate L5 system into service. Their tolerance for risk is FAR lower than Teslas." as clearly implying that Tesla were reckless in a way legacy OEMs are not. And I'm sure the dozens of other people who agreed with @StealthP3D's post thought the same. You might not have meant to imply that, but you certainly did.

 

[StealthP3D]

It's so absurd in fact that literally nobody here including in the post to which you are replying, claimed they would.

It was in fact about someone else suggesting a legacy OEM would do so in a rush to compete with a 99.9% safe Tesla RT system.

So while I appreciate the many paragraphs you wrote- it was about an argument nobody was having.

My main point was it's false that legacy manufacturers care more about customer safety than Tesla. And even someone who did not have English as their first language can parse the logical implication of your claim:

No legacy OEM would put a 1% failure rate L5 system into service. Their tolerance for risk is FAR lower than Teslas.

 

[Knightshade]

My main point was it's false that legacy manufacturers care more about customer safety than Tesla.

Then you again appear to be arguing with something nobody said.

Corporate risk liability- which would be massive for a commercial robotaxi service that failed 1% of the time- is something no legacy OEM would consider putting into service. (neither, of course, would Tesla).

And even someone who did not have English as their first language can parse the logical implication of your claim:

Apparently you can't though, as you again seem to be trying to make up an argument nobody was having just to disagree with me, specifically.

 

[Knightshade]

Actually, I read your comment "No legacy OEM would put a 1% failure rate L5 system into service. Their tolerance for risk is FAR lower than Teslas." as clearly implying that Tesla were reckless in a way legacy OEMs are not. And I'm sure the dozens of other people who agreed with @StealthP3D's post thought the same. You might not have meant to imply that, but you certainly did.

Again, in the hypothetical, Tesla was the one running a system an entire order of magnitude safer than the OEMs.

There was no risk for Tesla to be taking there (or at the least generous reading, an incredibly tiny one).

The point being there's simply no reason any legacy OEM would run something 10 times riskier, especially as a commercial offering.

In a world where Tesla has a 99.9% safe system, they'd either license Teslas system-- or they'd take some years to develop their own (or license one from someone else who did so)-- or they'd just not offer RTs at all.


The original suggestion- that car makers are in SUCH A HUGE RUSH TO PUT RTS ON THE ROAD they'd run a system that failed 1 out of every 100 times is the absurd bit.

 

[Beltsbear]

I remember Elon saying that the FSD chip has security features to ensure that only Tesla software runs on it. I'm not sure if these security features also reduce the likelihood of someone trying to reverse engineer the NN model.

Even if they're able to reverse engineer the model, we already know that Tesla will continually update it, because it's a game of trailing 9s, not 100%. So whoever is reverse engineering the system will essentially have to create a software workflow to automate the reverse engineering and have unit tests to figure out what changed with each update, lol. This seems like a lot of work in itself.

I can see some companies in China trying to do this, but if it were so easy to reverse engineer fsd, a lot of companies would already be doing it with current AP.

Apparently greentheonly implied that he can see the neural network layout somehow. Either through watching downloads or looking at memory. Code signing would protect the chip from outside code though.