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

[mongo]

Um, no, that's not how it works.... Information in a neural network is not localized; it's distributed. There is no "car neuron" in HW3, sitting next to a "truck neuron", next to a "pedestrian neuron", etc

As for the number of connections, at least in the cases I'm familiar with, each layer involves a connection between each neuron in that layer and each one in the next adjacent layer.

Um, yes it is. For each object type you want to identify, you need a final probability value for that type. AlexNet works against the ImageNet data set which was 1,000 categories. AlexNet has, as its final output a 1,000 element softmax array.
Understanding AlexNet | Learn OpenCV

Depending on the kernel and stride (CNN), it is not a one to all mapping between layers. Even non-CNN can use subsampling between layers.

 

[KarenRei]

Um, yes it is. For each object type you want to identify, you need a final probability value for that type. AlexNet works against the ImageNet data set which was 1,000 categories. AlexNet has, as its final output a 1,000 element softmax array.

Ah, you're talking about the output layer. That's not what was being discussed on the topic of the capability of a neural net. In a deep neural net, you have an input layer (arbitrary number of neurons; can involve multiple layers for complex tasks), an output layer (different arbitrary number of neurons), and an arbitrary number of hidden layers (yet another number of neurons), with each layer generally all having the same number of neurons and one connection between each neuron in each layer to every neuron in the subsequent layer. It's the hidden layers that are responsible for generating the output, and thus the layers under discussion. The output layer is a data sink; it does not "determine itself".

ED: In your link, the hidden layers are referred to as FC layers.

 

[Nocturnal]

FSD represents a lot more than just the Tesla Network. Also, I would challenge the statement that "most people think FSD is a long way off". Neither you nor I have any way of knowing what "most people" think or don't think so that argument becomes nothing more than speculation unless you have data to back thus up?

Dan

Most people seem to think FSD is impossible, or at least impossible anytime in the near future. I think those people are seriously underestimating technology.

 

[copyhacker]

I'd love to be able to just download a CSV of energy usage to my thumbdrive so that I could get all geeky over it. And if Tesla collects it (while they certainly could, I doubt they bother with all of it given the size of the fleet) then making it available for download from your account. Either way works for me.

The Stats app does this, AFAICT: ‎Stats: For Tesla Model S/X/3

I just downloaded a CSV of my charging sessions to my phone.

 

[mongo]

You're talking about the output layer. That's not what was being discussed. In a deep neural net, you have an input layer (arbitrary number of neurons; can involve multiple layers for complex tasks), an output layer (different arbitrary number of neurons), and an arbitrary number of hidden layers (yet another number of neurons), with each layer generally all having the same number of neurons and one connection between each neuron in each layer to every neuron in the subsequent layer. It's the hidden layers that are responsible for generating the output, and thus the layers under discussion. The output layer is a data sink; it does not "determine itself".

ED: In your link, the hidden layers are referred to as FC layers.

I may have been tangential to the discussion, I was referring to:

More neurons isn’t necessarily better, it’s the weights that store information.

If you want to recognize more things, you need more neurons (and likely more neuron in the layers below).

 

[Dan Detweiler]

Ah, you're talking about the output layer. That's not what was being discussed on the topic of the capability of a neural net. In a deep neural net, you have an input layer (arbitrary number of neurons; can involve multiple layers for complex tasks), an output layer (different arbitrary number of neurons), and an arbitrary number of hidden layers (yet another number of neurons), with each layer generally all having the same number of neurons and one connection between each neuron in each layer to every neuron in the subsequent layer. It's the hidden layers that are responsible for generating the output, and thus the layers under discussion. The output layer is a data sink; it does not "determine itself".

ED: In your link, the hidden layers are referred to as FC layers.

Damn it @KarenRei, just when I am convinced you really are a live human being you go and post something like this. Now I'm back to Karen the bot. Geez!

Dan

 

[Nocturnal]

Before I forget, the "German SEC" is opening a case against the FT's alphaville section head honcho for allegedly manipulating Wirecard's stock to the benefit of short sellers. Interesting.

Anzeige erstattet: Bafin verdächtigt „FT“-Journalisten im Fall Wirecard | 16.04.2019

Das SEC?

Germany's biggest coalition party is stating in public it will not take climate actions that where Tesla wins and current German auto industry loses instead of choosing the right path:

View attachment 398491

Yet Europeans constantly attack Americans for being jingoistic and not caring about the climate etc.

 

[humbaba]

Most people seem to think FSD is impossible, or at least impossible anytime in the near future. I think those people are seriously underestimating technology.

Maybe. But I'd say you are seriously underestimating the difficulty of the problem.

That said, it all depends on how you define the problem. In the extreme you have the Waymo/GM Cruise approach where you restrict the solution to a geographical box with specific routes. That is far easier than the Tesla approach of (currently) restricting the solution highways.

The way that Tesla defines advanced summon means that I cannot use it to summon my car in the work parking lot to pick me up at the door. So when Musk says "feature complete by end of 2019" what I hear is "meets some subset of features not currently defined" rather than "full autonomy."

So why am I so happy with Tesla's autopilot? Because it doesn't need to be "full autonomy" to have real utility for drivers. A useful subset adds value and NoA is an already available example of such a subset.

My cynical side says FSD will be "feature complete" by end of 2019 in order to recognize revenue rather than from actually having all features in place. (Of course, with the proper semantics it could be feature complete while still not being complete -- and that's not even counting the accuracy/safety estimate for end of 2020.)

Regardless of semantics games I expect Tesla's FSD to progress and improve into the foreseeable future.

 

[Sancho]

I am also highly skeptical of FSD any time soon. But, I recognize that I have zero basis for an informed opinion on that.

If I had been around 100 years ago and somebody started claiming we’d all be commercially flying soon, I would have thought it crazy.

In fact, I still wear my lucky shirt every time I fly, because it still seems counterintuitive to me that a big hulk of steel stays airborne so reliably.

Elon, by contrast, does have the chops and experience to have an informed opinion on FSD, and his track record on STEM things is pretty incredible.

So perhaps FSD will happen soon. In which case, I better get my lucky shirt to the laundry.

 

[humbaba]

The Stats app does this, AFAICT: ‎Stats: For Tesla Model S/X/3

I just downloaded a CSV of my charging sessions to my phone.

Thanks! I'll take a look at that.

 

[SOULPEDL]

Happy to discuss. The post on Reddit is garbage. Yes, the SR+ will be in demand but even the more expensive models will not pay exorbitant prices either. As I said before the Model 3 is perfectly priced for Denmark. If you want to know how much fees there are for DK, go to Tesla, configure the car and go to check-out. There you can see how much Sales Taxe and how much Registration Fees are in the prices.

Edit: while I was in Tesla shop in Hamburg yesterday I noted that a few Danes were stopping by, too. So it is interesting to see that the car has enough of a pull for folks to spend time during their vacation to look at it... The sales adviser in the store said that they are looking for more staff in Hamburg as they can't handle the volume of orders coming it. He said he himself processes about 60 orders per week. To me that doesn't seem extremely high. I guess there are more orders online and of course there are many more sales folks out there, too. While I was at the store a constant flow of people came in. It was busy but not crazy busy.



Here you go: Interesting insight on the state of Autonomy from the Founder of Creative Destruction Lab and AI expert : teslamotors

I did read it. Interesting, not revolutionary. Some of it seems a bit 3rd hand knowledge of a guy who is giving his estimate about a field he generally know something about but without being in the details.

Edit: @shlokavica22 beat me to it...

What the hell is your problem? Like any long term investor, I want the company to succeed. I have 452 shares. How many do you have?

You can be a fan and a shareholder of the company while also reasonably questioning their tactics, especially when they have underperformed and straight lied so many times. You have gone off the deep end if all you see is positive.

DocX, seems I was straying off the issue. It does boil down to trusting Tesla or not. I don't see lies, and maybe your def'n of a lie is simply his being late. I have very stong faith in FSD based on progress/experience, while others have lost faith because he is late and it's a seemingly impossible task. I also shared your same frustration a year ago at time of purchase, as stated. I'm way past that now; the ball is in motion and it's massive!
Meanwhile, with slamming comments like what I read, and continue to read by nay sayers and FUD alike, it's hard to separate FUD from honest skeptisizm which you may be having. But a hard pill when you have that many shares and talk like FUD. When I see FUD, I try and call it out and it is unfortunately emotional because, since you asked, I own 1050 shares and the rest is in ARKQ. I over bought a bit. Plus the planet...

My sincere appologies if I pegged you wrongly. Calling a truce and recommend reading that reddit article. It's an opinion piece, but sometimes you can just see logic presented and totally get it.
You will get FSD, given Tesla's current trajectory IMHO. Have a little faith and know that others on the fence could sell the stock based on what you wrote and that affects both of us.
At this point, I am leaning toward Tesla owning self driving, not just leading the way. Unless someone shows me that Google search suddenly has a real competitor like AltaVista or the financial system is so corrupt that Tesla fails as a company and is sued into the dirt, stock prices should sky rocket. If you don't have EAP or FSD, I highly recommend it.
There is a big difference between a "lie" and being delayed. No lie, the stock will likely hit 1,000 this decade IMO. When, is likely later than it should, so I am patient.

 

[KarenRei]

I am also highly skeptical of FSD any time soon. But, I recognize that I have zero basis for an informed opinion on that.

If I had been around 100 years ago and somebody started claiming we’d all be commercially flying soon, I would have thought it crazy.

In fact, I still wear my lucky shirt every time I fly, because it still seems counterintuitive to me that a big hulk of steel stays airborne so reliably.

Elon, by contrast, does have the chops and experience to have an informed opinion on FSD, and his track record on STEM things is pretty incredible.

So perhaps FSD will happen soon. In which case, I better get my lucky shirt to the laundry.

If you have to wear a lucky shirt to fly, you're going to want a whole lucky outfit for when Starship starts doing Earth-to-Earth hops

 

[Reciprocity]

I'm still hoping that Tesla pulls the same stunt as they did in Germany, and makes a $44,9k CAD Tesla that's so software stripped-down and backordered that nobody will ever want it - but still making it available for sale, and actually delivering for the few people that decide to "test" Tesla and order one.

As soon as the Canadian incentives are signed into law, that is.

They can accuse Tesla of trying to rig the incentives all they want... but they clearly tried to rig the incentives to exclude Tesla, so...

Blame Trudeau. If the Canadian dollar wasn't so bad, Tesla would qualify. In 2009ish the Canadian dollar was worth more than the US dollar.

 

[humbaba]

I am also highly skeptical of FSD any time soon. But, I recognize that I have zero basis for an informed opinion on that.

If I had been around 100 years ago and somebody started claiming we’d all be commercially flying soon, I would have thought it crazy.

In fact, I still wear my lucky shirt every time I fly, because it still seems counterintuitive to me that a big hulk of steel stays airborne so reliably.

Elon, by contrast, does have the chops and experience to have an informed opinion on FSD, and his track record on STEM things is pretty incredible.

So perhaps FSD will happen soon. In which case, I better get my lucky shirt to the laundry.

While I think I have more than "zero basis for an informed opinion" I share your sentiment about betting against Elon. Back when I first heard about SpaceX and their ridiculous idea to land and re-use rockets... I was not kind. In fairness to myself, I was not alone.

Nevertheless, I think the moral of the story isn't that Elon achieved "the impossible" but that he proved the feasibility. I'm not sure how many people with knowledge of the subject thought it was impossible but rather that it was pointless because it was both difficult and unprofitable. And now SpaceX is beating the competitors because it is both achievable and profitable.

With FSD I think it is more about timelines and approach than about difficulty and utility. Tesla is not asserting the problem is easier, they are taking a different approach. And everyone agrees on the utility. I think the risky proposition here is not doubting Elon's optimistic FSD timeline, but doubting that Tesla's approach will not result in a usable FSD (for some definition of practical FSD).

 

[Unpilot]

I am also highly skeptical of FSD any time soon. But, I recognize that I have zero basis for an informed opinion on that.

If I had been around 100 years ago and somebody started claiming we’d all be commercially flying soon, I would have thought it crazy.

In fact, I still wear my lucky shirt every time I fly, because it still seems counterintuitive to me that a big hulk of steel stays airborne so reliably.

Elon, by contrast, does have the chops and experience to have an informed opinion on FSD, and his track record on STEM things is pretty incredible.

So perhaps FSD will happen soon. In which case, I better get my lucky shirt to the laundry.

Hey I fly for a living and I wear magic underwear!

Oops wait that was in a previous life when I was a Mormon.

Maybe that was tmi and OT

 

[winfield100]

While I think I have more than "zero basis for an informed opinion" I share your sentiment about betting against Elon. Back when I first heard about SpaceX and their ridiculous idea to land and re-use rockets... I was not kind. In fairness to myself, I was not alone..

@humbaba
obviously not a long term Science Fiction fan. (since late 1950's)
How are you gonna travel to other planets w/o the capability?
Chesley Bonestell had lots of paintings when humanity was more hopeful, now the pendulum has swung back
(Falcon heavy launch from Cocoa Beach, Florida, impressive, Quad sonic booms when landed, cheers from the beach)
(and technology is catching up with where the dreamers imagined, and surpassing a wee bit)

(we need to be multiplanetary for the species to survive it seems)
Chesley Bonestell: A Brush with the Future - Home

 

[Lucky_Man]

The mainstream financial media will LIE about the Q1 report no matter what it contains. Remember how in early October they falsely reported 2018Q3 production numbers as a 'miss' when in fact it was a 'beat', dropping the SP? Corrected hours or days later? But not before being reprinted in syndication by hundreds of media outlets. Those outlets do the damage, and they never print the retraction.

Then on Oct 26 at 2:00 pm (one day after the big beat on the 2018Q3 ER), and with TSLA up $31 intraday, the WSJ posted a fraudulent 'hit piece' that dropped the SP $20 in a few minutes. And the SP only slowly recovered, closing up only $16 for the day. That is two billion seven hundred sixty-eight million dollars ($2.768b) in Shareholder value destroyed in 2 hours by the fraudulent reporting by the WSJ.

[edited for brevity]

2018Q3 ER was a surprise 'beat' so WSJ knowingly published a false story to 'engineer a dip' to help unnamed shortz exit their positions. This is fraud. The SEC remained silent.

You've described perfectly where the true fraud is. I find this detrimental effect of WSJ's ongoing effort to destroy #TSLA's SP is sickening. Still, it goes under the SEC radar.

All the time SEC spends hanging on his every tweet is wasted. The current SEC leadership is but one small example of the corruption and greed rampant today. I can't wait to get back in the voting booth!

 

[humbaba]

The Stats app does this, AFAICT: ‎Stats: For Tesla Model S/X/3

I just downloaded a CSV of my charging sessions to my phone.

Thanks again for pointing the app out. Sadly, I think I'll have to pass. Reading the description it appears that it would necessarily have to have access to the account/API which at present I'm not comfortable with, though that could change. Chalk it up to paranoia instilled from a career in security.

It looks like an excellent app and does what I was asking for and more. Its not the developer's fault that Tesla's API is the way it is. I'd like Tesla to refine their API so that it allowed finer grained control than all-or-nothing.

 

[gene]

You've described perfectly where the true fraud is. I find this detrimental effect of WSJ's ongoing effort to destroy #TSLA's SP is sickening. Still, it goes under the SEC radar.

All the time SEC spends hanging on his every tweet is wasted. The current SEC leadership is but one small example of the corruption and greed rampant today. I can't wait to get back in the voting booth!

And while you are voting, consider this: Bernie Sanders Calls For An End To Fossil Fuels | CleanTechnica

 

[SOULPEDL]

Blame Trudeau. If the Canadian dollar wasn't so bad, Tesla would qualify. In 2009ish the Canadian dollar was worth more than the US dollar.

Pierre or Justin?