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

[dc_h]

I bet not re: Moody’s upgrades.

Good point, a Moody's upgrade (for what it's worth), should be coming following the 22nd earnings.

 

[marvinat0rz]

I pulled the trigger on my move. So for the time being, only 25% of my net worth is in TSLA Hoping to see that number climb back up to wealth manager heart attack inducing levels in the months and years to come, and to have no regrets when that happens.

Thanks for your input, guys. It's nice to be able to bounce this line of reasoning off someone who's largely in the same boat. For obvious reasons, this is something I have to keep largely quiet about in my normal social circle. I value this greatly.

I emphasize that this is really just a hedge. As option-based hedges are unreasonably expensive, this is really the only reasonable choice if I was to enter a meaningful hedge at all. And maybe in a different financial situation, in the future, I wouldn't, but although we are talking about a truly exceptional company here, it's still a single company. Likely to succeed, but still exposed to risks both of the valuation and tail-risk variety. This is me choosing the timing of my hedge to a period of great optimism, epidemic nonwithstanding. I am happy to have done it today rather than two months ago, which might also have been very tempting.

This is not an exit, so you'll still see me around

 

[Chunky Jr.]

From my understanding a neural net is just one piece of the FSD puzzle. The NN will take images and figure out what is in those images (cars, bikes, people, signs, lines, cones, animals, etc). Basically a NN is used to find patterns in the video streams, then other code (that is not a NN) decides how to manipulate the car in response to those patterns

 

[Artful Dodger]

"Elon Musk at WAIC* 2020" (16:54)

*WAIC is the World Artificial Intelligence Conference

Cheers!

 

[hobbes]

Balanced report by Deutsche Welle (kind of German NPR) on GF4 in English:

https://twitter.com/dw_business/status/1280940465273974784

Not much news, but good general overview of the situation.

 

[Joe F]

Hey we already have ionity here with up to 350kw charging. And they cost only 0,79€/kWh! Cheap!
It's also much cheaper if you have a german ev, different rates. Which is rather interesting from a legal point of view, Ionity did get EU subsidies for building the network.

Right. All three locations for a sum total of 6 chargers.

I guess it's a start.

 

[ZsoZso]

I think it's important to put down Elon's exact words:
"I think I remain confident that we will have the basic functionality for level five autonomy complete this year. So, yes, I think there are no fundamental challenges remaining for level five autonomy. There are many small problems. And then there's the challenge of solving all those small problems and then putting the whole system together and just keep addressing the long tail of problems so you'll find that you're able to address the vast majority of situations."
...
"But I'm absolutely confident that this can be accomplished with the hardware that is in the Tesla today and simply by making software improvements we can achieve level five autonomy."

I think you need to draw a distinction here between having "the basic functionality for level five autonomy" and having actual level five autonomy.

Thank you for the exact quote. You are right, that is very far from saying Tesla cars will be level 5 by end of year.
I would caution anybody from making that leap of faith and making investment decisions based on that stretch (e.g. buying much higher CALLs than current SP).

In the most optimistic scenario (if all goes well and Elon time is not in play), you may see Tesla release "feature complete FSD", but it will be officially still level 2!
What does that mean ? The software (NN) will have all the code needed to recognise all road signs, traffic lights, handle all driving situations in theory. However, the driver is still responsible to take over any time the car makes a "mistake". And it will make mistakes, because there will be lots of edge cases not handled correctly. It will take many months of data gathering , learning from driver take-overs to improve the system to reach level 3, then even more time for level 4 and probably several years before it becomes level 5. Remember level 5 requires handling all kinds of whacky situations and bad weather conditions anywhere a human could drive.

 

[Bobfitz1]

When you know how the car and the HW3 computer are built and that’s simply not possible. Tesla could make another computer, probably called HW4, and replace all the cars with that. But you’re definitely gonna have to swap that entire board for any substantial upgrade.

The processing chip and HW3 board it sits in are two different pieces of hardware. If the new chip is designed to be compatible with the HW3 board (as I suspect it has been all along) then that new chip can replace the original. I thought there was some comment during Autonomous Investor day presentation that this was the plan. I'll have to go back sometime to see if that's the case.
What's clear is that they would not be pushing hard to build an even faster main processor knowing it would be incompatible with the HW3 that was only starting to go into new cars.

 

[dc_h]

I'm considering to reduce my TSLA holdings down from the current 50% to 25% of the amount I initially purchased many years ago.

My expectation is still that Tesla will become one of the world's dominant technology companies and retain that position for decades. My initial investment thesis from then still holds - namely that electric transportation will become the world's dominant transport technology, and that the established transportation industry is caught in a classic Innovator's Dilemma which makes them structurally incapable of competing, eventually resulting in their downfall.

Tesla has executed beyond what I had even dared to think in the most optimistic scenario, and the competition has floundered worse. They've done almost all the critical strategic bits right. Still not sold on the SolarCity acquisition, but I have my suspicions that this move averted an obscure liquidity problem that would have been a grave threat to Tesla. The future is looking bright, and Tesla is the only player that shows real credibility for innovating in electric transport and tech manufacturing. I wouldn't be surprised if they're making aircraft in five or ten years. So few companies and capitalists dare to put their money where their mouth is. It's so promising to support such a movement.

With that said, the current stock rally is putting me in a position where I am within arm's reach of my life financial goals, almost a decade ahead of schedule. The freedom that comes from being in such a position might be greater than potentially losing out on having a relatively ridiculous (hey, I'm middle class) ownership stake in tomorrow's greatest tech behemoth.

I am seriously considering to lock in most of those gains and put them in something diversified. It's been years of light anxiety and worry. It's scary to have the finish line in sight.

Dunno how you folks have been managing this situation. Few on this board are independently wealthy, and I notice we've lost a few members over the years of rollercoaster appreciation.

I think some base retirement funds are critical. Presumably it is not just about you, but your family. I'm approaching a similar place. I have the good luck of having a defined pension to take care of my baseline life stuff. One option is to sell in the money or near money options a year or more out as an insurance policy. It depends on how many shares you have, but if you sell near money options a year out, you can diversify and if Tesla goes down a bit (unlikely), you have other more stable investments to cover you. If it goes up, you are locked in on today's price and you have the call funds. I'm not an advisor and everyone's risk and life requirements are different. My 10% investment in Tesla become more when I dumped FB, MS and AAPL a couple years ago. If I was already retired, it would have been very stressful and I may have taken some losses. Any normal retirement portfolio is probably going to have 30% down years. An all Tesla fund could see 60-70% down and 400% up years. I think we're due for another 100%+ up year, but if I knew, I'd sell all my stock and buy leaps.

Define your risk model and then define your portfolio. If you are good on Excel, you can combine the volatility of your investments and see what the potential sine wave looks like. What does it look like when Tesla goes down 50% for a year, do you have to tighten your belt, or can you make other adjustments. No one should be switching to Mac and Cheese, peanut butter and jelly and cheap booze in their 60's.

I do think we are in for a good run, even with all the great results of the last few months. Tesla is hitting scale and their advantages are just becoming common knowledge among informed investors and will become apparent to mainstream people globally over the next 2 years. That said, it will remain a battleground stock. It is a bullet aimed at the heart of the fossil fuel industry, legacy auto, big oil and perhaps big grid utilities. It's role in the future economy will make it subject to lobbying and PR from yesterday's leaders. Being a 500 billion stock in 2021 won't stop the backstabbing from the laggards and Luddites.

I"m not sure if there is an exit strategy thread, but adding some risk management ideas for TMC'ers approaching retirement would be great. I'd like to hear other people's thoughts. I'd prefer not to sell for as long as possible, but at some point, i'll probably need to diversify.

 

[StealthP3D]

That's a 65.30 Euro drop (USD$73.90) from High to Low for NO REASON at all, and suspiciously ignored by the U.S. markets on Monday, July 06. It's Bears sending each other covert signals for cheap (low volume in Frankfurt), messages for which they can't be prosecuted for collusion.

Speaking of covert bear signals, I've noticed a suspicious drop in the share price repeatedly about 10-15 minutes before the real selling begins. But it only shows up in the daily chart with 1-minute intervals that I get from Schwab (open-high-low-close). These brief down legs really stand out visually because the drop is about 5-10 times the magnitude of surrounding lines. So, I suspect it is some kind of errant trade that is entered to send a message to others. I say "errant trade" because the drop is eventually corrected (erased) and doesn't appear in the chart after about an hour or so. But very often 10-15 minutes after this signal some persistent selling shows up and the share price drops.

 

[lafrisbee]

JULY 22nd after market closes. Earnings report>>>
Did that get posted already?
OK..so I forget stuff...

 

[MP3Mike]

The processing chip and HW3 board it sits in are two different pieces of hardware. If the new chip is designed to be compatible with the HW3 board (as I suspect it has been all along) then that new chip can replace the original.

While technically possible, it isn't going to happen. The chip is soldered to the board and would requires special equipment and skills to replace. Tesla isn't going to do that, especially not at a service center. And given the construction of the computer module in the Model 3/Y they aren't even going to replace just the FSD board, they would replace the whole unit, including the MCU, just like they did for the AP2.5 to AP3 upgrade. (Assuming they would offer an upgrade at all.)

 

[Chunky Jr.]

The stock has risen 50% since I decided to be "smart" and sell a few shares in the low 900s after that run up earlier in the year. I'm not making that mistake again.

 

[ggr]

FWIW I coded my own neural network from scratch and have been coding for a scary 39 years *cracks knuckles*.
Yes, teslas approach is not only the best, but IMHO the ONLY one that can possibly offer a truly universal solution, and the only one that can produce a really safe solution to FSD.

But hell... this stuffs hard. Neural networks become a lot more like art than science. Its a messy *fuzzy* business, and starts to feel more like voodoo than normal code. Despite all the memes, its not that hard to predict how far through a project you are when its normal object-oriented traditional C++, but the minute you have a massive neural net involved... things get way harder to predict.

FWIW my entirely subjective opinion is that they may still be another small rewrite away from FSD. I think the big gear change was karpathy. He understands that you need the code to be mostly neural net, with minimal procedural code. I think thats the only answer, and its a tough one to agree with unless you work on this stuff all day, because it feels SCARy as a coder to effectively cede control of decision making to a neural net nobody can understand.

From what I read, karpathys time at tesla has been one of progressively lobbying for the NN to take over more of the decision making. I think it will make up an overwhelming proportion of all the FSD code eventually.

I dont think we will get FSD until end of 2021 at best. But I am 100% certain nobody will beat tesla to it, or even get close. Persuading trad auto companies to let software control a car is very hard. persuading them to let software *nobody understands* to do it is almost impossible. Its only possible at tesla because elon is a coder with an interest in how brains work.

Tacking onto your post, an observation: Tesla is rewriting all the code to make it work in three dimensions, but they still have all the training data that has been gathered to date. None of that has been invalidated in any way, it's still just images, radar returns, and ultrasonic returns. So once the software appears, a week or two later it'll be ready. This is obvious to some, but inconceivable to others.

 

[UnknownSoldier]

No one seems to be buying or selling. July 22nd gonna be a big day.

 

[humbaba]

The processing chip and HW3 board it sits in are two different pieces of hardware. If the new chip is designed to be compatible with the HW3 board (as I suspect it has been all along) then that new chip can replace the original. I thought there was some comment during Autonomous Investor day presentation that this was the plan. I'll have to go back sometime to see if that's the case.
What's clear is that they would not be pushing hard to build an even faster main processor knowing it would be incompatible with the HW3 that was only starting to go into new cars.

Chip swapping makes no sense. Even if the chip were socketed (which I'm so certain it is not I won't even bother checking) it would make no sense. Why?

Producing the chip is going to be the most expensive part of the board. The rest of it will have some cost, but it won't be that much. So while there is some marginal cost savings to be had here there isn't much. And it would be reduced by using a socket (which would also introduce other problems).

Even if you wanted to have the problems of a socketed chip and for some reason care about the slight improvement to cost let's talk about swapping the chip. This is *not* your PC where you unplug it, pop the case off, pull the CPU, socket the new one and call it good. For starters, the entire board will have to be removed anyway and this is a significant operation taking quite a bit of time to perform. The board was designed to be swappable, but it was crammed in and is not as simple as saying "swap the board." If the chip were socketed it wouldn't be that much more to do (because you already pulled the board), but there is greater risk (e.g., the service tech damaging a pin). The point is that it is *more* work to do the service replacement which is going to have an incremental cost that eats into whatever savings might have been achieved by solely replacing the chip. And since the chip is surface mounted it becomes a *lot* more work.

So, doing a chip swap will cost more, take longer, and has higher risk with effectively the same cost. Why bother? It was designed for board swapping and that is how Tesla will handle them (not that I think they will ever upgrade current vehicles past HW3, but that is a different post).

[edit: as for being designed for compatibility? Nope, no way. Just look at the HW3 design -- it is absolutely optimized for what it is without any concession. Trying to cram a new design into an old introduces compromises and there's no point to do doing it in this case. Again, not PCs]

 

[Singuy]

Thank you for the exact quote. You are right, that is very far from saying Tesla cars will be level 5 by end of year.
I would caution anybody from making that leap of faith and making investment decisions based on that stretch (e.g. buying much higher CALLs than current SP).

In the most optimistic scenario (if all goes well and Elon time is not in play), you may see Tesla release "feature complete FSD", but it will be officially still level 2!
What does that mean ? The software (NN) will have all the code needed to recognise all road signs, traffic lights, handle all driving situations in theory. However, the driver is still responsible to take over any time the car makes a "mistake". And it will make mistakes, because there will be lots of edge cases not handled correctly. It will take many months of data gathering , learning from driver take-overs to improve the system to reach level 3, then even more time for level 4 and probably several years before it becomes level 5. Remember level 5 requires handling all kinds of whacky situations and bad weather conditions anywhere a human could drive.

Basic functionality of level 3 maybe. Level 5 requires the car to make the right decision Everytime or else you don't have the basics. Basic of level 5 is level 5. Any functionality that is not fully mature is designated for lower levels. So the statement Elon made imo is optimistic and has missed his goal.

 

[lafrisbee]

Totally OT
There is an old black and white movie about Lloyd's of London if I remember right. It has a scene in it where a windmill and fire light are used to send messages across the English channel. Can anyone post the name...I am feeling as though now I have spent the last month on this site it will have new meaning and evoke a deeper set of emotions.
EDIT: Never mind. I think I found it. I didn't think it would be named "Lloyds of London."

 

[ninpb]

FWIW I coded my own neural network from scratch and have been coding for a scary 39 years *cracks knuckles*.
Yes, teslas approach is not only the best, but IMHO the ONLY one that can possibly offer a truly universal solution, and the only one that can produce a really safe solution to FSD.

But hell... this stuffs hard. Neural networks become a lot more like art than science. Its a messy *fuzzy* business, and starts to feel more like voodoo than normal code. Despite all the memes, its not that hard to predict how far through a project you are when its normal object-oriented traditional C++, but the minute you have a massive neural net involved... things get way harder to predict.

FWIW my entirely subjective opinion is that they may still be another small rewrite away from FSD. I think the big gear change was karpathy. He understands that you need the code to be mostly neural net, with minimal procedural code. I think thats the only answer, and its a tough one to agree with unless you work on this stuff all day, because it feels SCARy as a coder to effectively cede control of decision making to a neural net nobody can understand.

From what I read, karpathys time at tesla has been one of progressively lobbying for the NN to take over more of the decision making. I think it will make up an overwhelming proportion of all the FSD code eventually.

I dont think we will get FSD until end of 2021 at best. But I am 100% certain nobody will beat tesla to it, or even get close. Persuading trad auto companies to let software control a car is very hard. persuading them to let software *nobody understands* to do it is almost impossible. Its only possible at tesla because elon is a coder with an interest in how brains work.

This is why I and my retirement love this forum. A lot of smart people from different spaces giving really good analysis. I might not understand the details but I can grasp the 30k foot view you guys are saying. I'm not a smart guy but I'm smart enough to listen to people smarter than me.

 

[j6Lpi429@3j]

From my understanding a neural net is just one piece of the FSD puzzle. The NN will take images and figure out what is in those images (cars, bikes, people, signs, lines, cones, animals, etc). Basically a NN is used to find patterns in the video streams, then other code (that is not a NN) decides how to manipulate the car in response to those patterns

Yes exactly, that seems to be the current model, but I think they will end up going further. Essentially the code is:

if(NeuralNetDetectsATrafficCone())
{
Do some hard coded C++ avoidance stuff here;
}

And I think thats one level too low. I think the number of edge cases and weirdness is just too massive for any normal software engineering to handle. I reckon that code will end up looking like this:

if(NeuralNetSeesSomethingWeird())
{
neural nets decides what to do here;
}

I get the impression thats what Karpathy wants, but there is likely pushback because its scary as hell. The problem is, the first approach STILL requries humans to anticipate every poissible 'issue' and wriote code to respond. You are *bound* to miss things. Thus the 75%/25% NN/C++ split may be a 'local maxima' where that approach gets you to 99% autonomous but never 100%. They may *have* to switch to pure NN systems to get to 100%. We know a pure NN can do it, because human drivers do it.

It doesnt worry me as an investor though, because the value of Tesla having autopilot at 99% FSD is still HUGE, so they can ride out that cash for however long it takes them to go 99-100%.

Disclaimer: Teslas coding team know way, way more about this than me, obviously.