Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

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

This site may earn commission on affiliate links.

Attachments

  • 8977BA19-1160-4DDA-92DD-B720CB8C9882.jpeg
    8977BA19-1160-4DDA-92DD-B720CB8C9882.jpeg
    1.5 MB · Views: 90
So with the news of the tanker attack and surging oil prices, I would expect a green day for TSLA today.

That seems rational at least, but expecting Wall Street to be rational is a lot to ask.

Also, given the news of high short interest, it would seem that now would be a good time for big institutional investors to get back in and force the price back up.
 
I'm long...like 5 years or more long, so....

Make cars...
Sell cars...
Innovate...
Motivate...
Keep getting closer to autonomy...
Keep getting closer to solar roof mass release.

Everything else...meh.

Relentlessly pushing the envelope. It's what defines Tesla besides the mission - identifying the right problems and finding solutions. Everyone else is busy with workarounds or distracted to some degree.

That's the only thing leaving me a little queasy after the latest event, all the while recognizing that scaling sans drama is a highly desirable development. Where's the daring new push? Sooner or later, something will be brewing.

Concerning advertisement, there's been a lot of excellent input here - as expected.

Besides a continuous effort from Tesla to put forward easily accessible information that cuts through the FUD, I think effective messages might be reduced to real bumper stickers echoing the big changes that have happened over just the last ten years:
- Tesla range & ease of "e-fueling". SF to LA! That's big!
- renewable energy growth
- drive train warranty

I wonder what Tesla has spent on R&D and capital goods up to now?

Just repeating these facts ad nauseam would help [notice I'm not specifying a medium] - it will still take a long time until the neurosurgeons on Hilton Head get the message.
 
Reminiscent of the LP to CD transition. People were happy with their records... until they had compact discs:
Did not need turning over half way
Were more scratch resistant
More bump resistant (you could dance without the needle jumping)
Fit on a smaller shelf
Didn’t wear out
Were equal in quality to the master

(Please no debate from audiophiles on lossiness, the point is historical - that CDs won by being more convenient).

As a former Bimmer aficionado, I think comparing Teslas to ICE vehicles is more akin to comparing DVDs to 8-Track tapes. The difference is that dramatic. And, while clunky, inefficient 8-track tapes were a spectacular fail foisted upon consumers in the lost Seventies, at least they didn't kill you if you left them running in a closed garage!
 
Last edited:
So this is from the Tesla Annual Shareholder Meeting's video stream, which ran for a couple of minutes and is included in the Youtube version as well, starting exactly at 8:00:


It definitely appears to be a vehicle's exterior from very close up, with a Tesla logo and the weird section you outlined. The 'rotation' is I believe camera movement. The whole sequence is repeated around 3 times before the live feed is cut in. I believe it's showing various details of a single car the whole time: the color scheme is similar.

Edit, here's a few detailed shots of the Roadster 2020:


I don't recognize that 'rotating' sequence as any of the details on the Roadster 2020.

I don't recognize these details from any of the existing production models or known protytypes either.

Could this be another Tesla Pickup Truck leak, hidden in plain sight? The black color scheme would support that interpretation too. :D

Yeah I also figure it could only be the Pickup, actually Tesla Pickup Truck is currently the only major vehicle whose complete face we have not seen. (Well... except for that "really exciting" Tesla Submarine Aquatic Car :eek:)

The rotating stuff also fits what Elon described as straight out of sci-fi scene.
 
I don’t believe it is possible for *all* ‘short low, cover high’ operators to be that stupid. Which leaves the only possible explanation that some large operators are sponsored, by <insert Tesla enemy>.

And I hate to say it but this leaves open the possibility that there is more dry powder where that came from. Much more. I hope I'm wrong and it should be more clear by approximately Monday or so.
 
  • Informative
Reactions: dqd88 and UncaNed
Luckily I don't think there have been *any* Model 3 fires. Not a single one so far.

I mean, eventually someone will have a Model 3 burn because it's inside a building which is on fire (this happened to a Model S in Canada, IIRC), so it is still an issue, but they seem to have made the pack even more fireproof than on S and X.
I'd bet a factor in this is that the Model 3 pack appears to have incorporated intumescent goo.
 
Battery-grade nickel sulfate isn't made from metallurgical nickel; it'd be far too expensive to produce the metal, purify it, and then convert it to battery-grade sulfate. It's made directly from nickel sulfide ore deposits, which are fairly rare, and usually with significant overburden. Traditionally you can't economically use even rich nickel laterites (by contrast, lower-grade laterites are extremely abundant globally, with little to no overburden). But the new tech push is price reductions in HPAL to be able to use low-grade laterites cheaper than sulfides.

Both nickel sulfate and lithium carbonate/chloride will be really feeling the crunch if Tesla scales quickly. The latter because, like you say, there's been significant underinvestment in lithium mining, and the former because we really need to move off of sulfides and on to laterites if li-ion EVs are going totally replace ICEs, since nickel sulfate is required in abundance (nickel = 80% of non-oxygen cathode mass). There's a Chinese company (can't remember the name off the top of my head) that's really been pushing the envelope in this regard. Fun side effect: HPAL recovery of nickel from low-grade laterites is also good at recovering cobalt as a side stream. :) So in case it can't be completely eliminated...

Yeah I knew all that. Just wanted to see if someone else did... :eek:
 
So this is from the Tesla Annual Shareholder Meeting's video stream, which ran for a couple of minutes and is included in the Youtube version as well, starting exactly at 8:00:


It definitely appears to be a vehicle's exterior from very close up, with a Tesla logo and the weird section you outlined. The 'rotation' is I believe camera movement. The whole sequence is repeated around 3 times before the live feed is cut in. I believe it's showing various details of a single car the whole time: the color scheme is similar.

Edit, here's a few detailed shots of the Roadster 2020:


I don't recognize that 'rotating' sequence as any of the details on the Roadster 2020.

I don't recognize these details from any of the existing production models or known protytypes either.

Could this be another Tesla Pickup Truck leak, hidden in plain sight? The black color scheme would support that interpretation too. :D

At about 8:45 there’s a right angled intersection of flat panels which doesn’t match any car in production that I can tell, and would support this being the truck. That’s my guess. I can’t really make any of it jibe with the earlier reveal shot, but maybe it doesn’t have to.
 
Luckily I don't think there have been *any* Model 3 fires. Not a single one so far.

I mean, eventually someone will have a Model 3 burn because it's inside a building which is on fire (this happened to a Model S in Canada, IIRC), so it is still an issue, but they seem to have made the pack even more fireproof than on S and X.

I think we have seen at least two: One that appeared to be a victim of the California wildfires and one that was a victim of a garage fire. I never heard what the cause of the garage fire was, but it wasn't the Model 3.
 
  • Helpful
Reactions: neroden
@skybluecgreen
remember from long ago, Betamax was a better picture than VHS, but VHS tapes were everywhere, and dominated. until superceeded by better. how to win, "firstest with the mostest"

VHS didn't win because it was first, it won because a) their tapes were long enough for a full movie (betamax tapes were originally only 60 mins long) and b) because the players were less expensive.

So to win, be practical and affordable (in that order). Blackberry was firstest with the mostest once, where are they now?
 
At about 8:45 there’s a right angled intersection of flat panels which doesn’t match any car in production that I can tell, and would support this being the truck. That’s my guess. I can’t really make any of it jibe with the earlier reveal shot, but maybe it doesn’t have to.

This frame, right?

upload_2019-6-13_15-30-36.png


That bit caught my attention as well, not because I recognized that it's a new panel intersection, but because they quickly cut to the 'rotating' sequence after just a glimpse of this panel intersection was shown - which felt weird.

Edit: another part that was weird is that the inner part of the panel gap is showing some sort of structure:

upload_2019-6-13_15-35-17.png


See those two vertical lines?

Edit #2:

I believe the "rotating" scene right after 8:45 is showing something that is opening right then. If you check the shadows and the distances it's all a bit off, as if something was lifted/opened.
 
Last edited:
Perhaps my favorite thing from the Autonomy Day was the information that they are making a large number of different specialized neural nets to do different specific things, and then passing information between them
I agree it's modular, but IMHO the different NNs mostly pass their info up to heuristic code that makes driving decisions. Passing info to each other creates lots of unnecessary complications.

IMHO these "different" NNs almost all have the same architecture, but were trained to recognize different things. This fits really well with their chip design and makes it very easy to add and or modify one NN without affecting any of the others.

Even if we say that Tesla scale data sets are necessary but not sufficient, who says that Tesla is solely using inference a la multi-multi-layer backprop?

Biological brains are a bag of tricks. You can find mis-matched based learning for categorical memory, match based learning in motor control, and opponent processing based reinforcement learning for timed behavioral selection. Some of the tricks involve learned mappings from one coordinate system to another, e.g. retinal coordinates to spatial to motor codes in movement. Some of the tricks transform input, e.g. visual inputs are spit into form and boundary channels that reconverge for further processing. And so on.

You can pick and choose from this bag of tricks. Oftimes you can leverage some of these insights in data preprocessing and input representations. Sometimes you might need to do more. You can even use a hammer: You can add 'rules,' algorithmic pre-, inter-, or postprocessing.

Just because Tesla isn't publicizing every trick they use doesn't mean they aren't using them. Don't be blinded by your own strawperson.

In point of fact, our evolution is the existence proof that a sufficiently large data set plus search/inference can find the solution to every problem in navigating in the real world and more.

Tl:dr; Silicon Valley understands nothing so well as search. Recast this problem in your mind as a search, i.e. evolutionary, problem. Recognize that the compute power to perform this search is available today (given a sufficiently large and comprehensive data set ;)). And, perhaps, you will come to a different answer.

Well, if it's really a "search" problem I'd bet on Google vs Tesla 99 times out of 100.

Tesla "could" have tricks up their sleeve but all they've said, the progress they've shown and the hacker exposes point to a straightforward approach. Heck, Karpathy even showed stuff they aren't doing and probably won't bother with (e.g. the parallax paper and 3D reconstruction from motion). Why get fancy when they're making great progress just training different Inception instances to extract more features from the same scenes?

What I see is a team in the heady early days, when capability grows by leaps and bounds. Other teams also enjoyed rapid early progress, but then hit a wall. I've tried to show the wall is not a lack of data. Otherwise Waymo would have a hundred vans out there doing nothing but unprotected left turns, amiright?

Maybe Tesla will be the one team that cleverly vaults over the wall instead of slowly chipping away at it. That just seems like wishful thinking, though. Anyway, I'll quite clogging the thread. We can revisit it next summer.