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

[kyzhk]

Looks like elon's twitter is going dead. I know, I dread fud based on his expressions. But does anyone else also regret the loss of communication from this great man? Never forget, he fought to communicate with us.

he is still very active talking about quantum physics now:

 

[Cobbler]

Volkswagen I.D3 opening preorder soon, but will be sold at a 3000€ loss minimim

 

[Fact Checking]

Looks like elon's twitter is going dead.

He will probably be able to tweet more freely with the new settlement: there's now a well defined list of Tesla news he has to seek Tesla approval for first before disclosing, instead of the dangerously vague "reasonably could be material" clause of the previous settlement that the SEC interpreted extremely broadly.

The new settlement is by all means less restrictive.

 

[ZeApelido]

Holy crap, people. I had a crazy work day and didn't check this thread for 16 hours...

...and you provided 350+ posts in that span, more than a post every 3 minutes.

It's Sunday. Nothing of merit happened today relating to Tesla's stock. The closest is the referral program boost, and that's not exactly market-moving.

Chill out. The poster who earlier today said it's impossible to keep up with reading--much less contributing to--this thread while having a life is correct. Try to self-moderate a bit. If you find yourself coming in for the 6th post of the day on a newsless Sunday, you are [part of] the problem.

 

[Artful Dodger]

Yes, but note that the biggest part of FSD not handled by NNs isn't even just the high level policy decisions/heuristics, but physics:

• Newtonian mechanics: how fast other vehicles are moving (speed), how large they are (probable inertial mass),
• probable near-future position of other vehicles,
• how the Tesla is moving: current position and speed (can be integrated based on inertial sensors such as accelerometers and GPS), attitude (angle of the car),
• map position and expected road conditions based on the GIS data, selected route,
• road conditions and slippage based on wheel rotation sensors,
• path evaluation and collision detection, safe free road section ahead and on parallel lanes,
• state of battery charge versus distance to destination, battery temperature optimization for the next charging stop,
• path optimization and targeting: within the safe boundaries established by the above physics calculations, what speed and position should the car target, which lanes to navigate intersections - with safety, efficiency and comfort optimized at once,

This is a LOT of physics and very little heuristics, with a lot of physics calculations, fed by the 3D object detection of the vision NNs, and these are well understood algorithms that make little sense to offload to the NNs.

FSD is architecturally pretty close to a game engine: a 3D object engine, physics engine, high level game logic and a user interface - except that in FSD the 3D engine detects 3D objects via an NN - while in a game the 3D engine generates a 3D scene.

Good overview! But the NN for route planning (likely with 2nd Gen FSD Computer) will also bring us closer to human-like skill in decision making. For example:

"I think somebody is shooting at me. I see muzzle flashes and bullet holes. Other cars are stopping. There's 2 in the ditch, 1 one fire. Hmm, I think I'll slow down."
​

Yes, this happened to me in Utah on I-15 North of SLC. But I was minding my own business, I promise.

In the mean time, I'll keep a steering wheel in my car (just in case *sugar*).

Cheers!

 

[Cobbler]

Yes, but note that the biggest part of FSD not handled by NNs isn't even just the high ...

Watch this

From the link posted yesterday by another user here. Merging and responding to a blockage using ML and not written code.
Very impressive.

Tesla has made the right choices.

My impression is that Tesla has now almost every piece of the puzzle ready (featureset complete), but MobileEye started already assembling the puzzle, but at slower rate?
Just an impression, not factual.

 

[schonelucht]

EC rent, a Dutch company specialized in EV rentals is ceasing operations because half of their Tesla fleet is out of commission since December 2018 due to parts shortages and no improvements in sight. Owner was a member here. Then there is Shenma which is a large ride sharing operator who claims to be the biggest Asian fleet buyer but took out billboards on Time Square to complain about Tesla service leaving 20% of their fleet inoperable.

This is how you destroy an indistructable brand. And, given the mounting losses in service and maintenance, there is not a lot of financial room to improve on this. There was a time that Tesla had few cars on the road, and perpetuallly flush with cash to provide top notch service. That’s when it was easy to do better than a dealership. Under strain, it’s doing worse in a lot of cases. Time to unbundle and let third parties engage in service and more, especially in countries without silly dealer/manufacturing regulatory fire walls. If it isn’t already too late.

 

[K3jSe7Llp7]

He will probably be able to tweet more freely with the new settlement: there's now a well defined list of Tesla news he has to seek Tesla approval for first before disclosing, instead of the dangerously vague "reasonably could be material" clause of the previous settlement that the SEC interpreted extremely broadly.

The new settlement is by all means less restrictive.

Thank you my friend. I deleted my unfounded post.

 

[Carl Raymond]

he is still very active talking about quantum physics now:

View attachment 401866

The “As sure as night follows day” reply from Elon, stemmed from this:
On The Future of Tesla and Full Self Driving Cars

Worth a read. Supports the thesis that vision obsoletes lidar for FSD. Cameras do the same job, and more, for a tiny fraction of the cost. (Edit: when paired with a supercomputer and suitable software)

 

[Fact Checking]

Watch this

From the link posted yesterday by another user here. Merging and responding to a blockage using ML and not written code.
Very impressive.

There's also the "path prediction" video posted by Tesla last week:

Tesla on Twitter​

Which might also be NN based.

Note how these are still "3D object attributes" based on vision input: the color coding of the object shows whether it's considered "blockage" or "vehicle part of traffic not moving". I assume this attribute is used by the "driving logic" - which is traditional code.

I agree that we'll see more and more if this: the NNs feeding the traditional code with richer and richer data.

Eventually NNs will also be able to learn to "drive" based on deep learning, with very little traditional code left, but that's probably still years away, IMO.

 

[AcesDealt]

As much as I have enjoyed the drama with Elon's tweeting including getting a couple responses myself over various mischievous provocations, I think Tesla would be far better off if he quit. He needs to become the unflappable and high status center with a serious responsibility not a circus monkey.

 

[Remus]

Please don't feel I'm picking on you, but this is exactly what I'm talking about. Amir Efrati (the anti-Waymo version of Anton Wahlman) once said Waymo struggled with unprotected left hand turns. This somehow morphed into "Waymo can't make left turns" here on TMC. Then a clip shows up with Tesla following a lane line on a protected left turn, and suddenly Tesla is "ahead" of Waymo.

It's pure wishful thinking. Tesla does not even attempt unprotected left turns yet.

Meanwhile, Waymo vans perform unprotected lefts every day. They are cautious when doing so. Perhaps too cautious. Last year Andrew J. Hawkins of The Verge rode in a Waymo van. It made several unprotected left turns, but he talks about one in particular that took a bit longer than he wanted:
Waymo Unprotected Left Turn

Is excess caution a problem? It is if it causes angry drivers behind you to riot. But insufficient caution is an even worse problem. You have to strike a balance, while also provide a smooth, non-jerky ride that doesn't scare your robotaxi customers. Waymo has worked on this and similar problems for years. They have better sensors than Tesla. They have more smart people than Tesla. Can someone tell me why it won't also take Telsa years to solve problems like unprotected left turns?

Thanks for setting the record straight about Waymo's capabilities. And we should all use caution. As I never believed robotaxi next year.

However the rest of your post contains flawed reasoning that contradicts historical facts, as I have pointed out repeatedly.

Futher more your conclusion is based on speculations not facts. Tesla "production" build doesn't attempt unprotected left turn. You don't know the capabilities of their engineering build.

Remember when they introduced the new chip the first time, they said the reason they need it is because the autopilot team had an NN that worked much better but unfortunately the on board computer can't handle it? How do they know that? They have engineering cars with big ass computers in the trunk too, just like the waymo ones. They even had engineering cars with lidar. Till today no Tesla customers have the more advanced NN on their cars yet.


Does this mean Tesla is definitely ahead? We probably need more evidence. But, Tesla doesn't need the left turn to make money, there is no unprotected left turn on highways.

On the other hand we know waymo will face a difficult decision soon. Lidar gives you a 3d model of the environment, a low resolution one. If both Tesla and Cornell's team demonstrated that you can get pretty much the same with cameras, why would you still use lidar? On the other hand changing the sensor suites means throwing a large portion of your work out.

 

[anthonyj]

EC rent, a Dutch company specialized in EV rentals is ceasing operations because half of their Tesla fleet is out of commission since December 2018 due to parts shortages and no improvements in sight. Owner was a member here. Then there is Shenma which is a large ride sharing operator who claims to be the biggest Asian fleet buyer but took out billboards on Time Square to complain about Tesla service leaving 20% of their fleet inoperable.

This is how you destroy an indistructable brand. And, given the mounting losses in service and maintenance, there is not a lot of financial room to improve on this. There was a time that Tesla had few cars on the road, and perpetuallly flush with cash to provide top notch service. That’s when it was easy to do better than a dealership. Under strain, it’s doing worse in a lot of cases. Time to unbundle and let third parties engage in service and more, especially in countries without silly dealer/manufacturing regulatory fire walls. If it isn’t already too late.

Great. Can’t wait to see that on TV tomorrow >.> maybe TSLA deserves the FUD that comes from it. I’ve been waiting for rear seat replacements for 4 months since the Service Center destroyed mine. They received a replacement last week but it was Gen 1 when I needed Gen 2... I don’t mind waiting another couple of months but it’s a major turn-off to the average consumer

 

[ClimateOptimst]

The “As sure as night follows day” reply from Elon, stemmed from this:
On The Future of Tesla and Full Self Driving Cars

Worth a read. Supports the thesis that vision obsoletes lidar for FSD. Cameras do the same job, and more, for a tiny fraction of the cost. (Edit: when paired with a supercomputer and suitable software)

Superb. also worth a read is Steve's blog from 8/23 after the Investor Autonomy day regarding Tesla's platform advantage & the foresight it took to start down their vision-based & FSD design path several years ago. the potential moat grows ever wider it seems On Tesla’s Incredible Platform Advantage

 

[Mike Smith]

Looks like around 240 open in Germany

Produits de Bourse

 

[Doggydogworld]

There are a number of areas in which you can again claim Google has had more time & smarter folks etc. But they haven't always won in the marketplace ...

If you replace Waymo with Microsoft and Tesla with Apple, didn't we hear similar things about this before?

I didn't ask about marketplace, I asked a technical question. Why advantage does Tesla have over Waymo solving unprotected left turns?

With lots of real world training data examples, which tesla has, it should be fairly trivial to train the neural net to handle any common driving case....

Thanks for at least addressing the question, but if it's trivial why did it take Waymo so long to come up with a semi-satisfactory solution? They have millions of instances of left turns - this isn't a corner case with airborne cars that you need a huge fleet to capture. Tesla has no data advantage here. The neural network doesn't make these driving decisions, so their chip doesn't help. Waymo knows the location, speed and (almost certainly) classification of every object, so why hadn't it been "trivial" for them?

Maybe all of Google's smart people are solving Go and StarCraft and winning ImageNet every other year. All that's left to work on the trillion dollar self-driving opportunity are interns?​

 

[Causalien]

Yes, computer vision is the Mobileye/Intel approach. But their business model is to act as vendor to the larger auto industry, like DELCO or BOSCH or LG.

Telsa's vertical integration means they will ALWAYS have a lower cost basis in volume production, after R&D is amortized. Tesla's long view means they plan to eat the auto industry's breakfast, lunch, and dinner.

Sounds like the legacy auto industry would benefit from some computer eyes in the back of their heads right now...

Cheers!

I don't know man.
When the competition can put a 10 000+ FSD hardware on a 15 000 ICE car. A tesla at 35 000 still cost more to produce.

What I think most are not thinking about is the willingness of rivals to burn cash in order to kill off the competition. It was quite an eye opening experience to watch Uber and DIDI fight over the past few years.

Hindsight, everyone probably agrees that it is stupid to go up against a gov owned corp with near infinite resources. But before such a duel took place, nobody knew whether attractive tech from Silicon valley will win against gov backed corp.

Probably a good idea that TSLA is such a Chinese Gov darling. One rule of thumb for Elon's twitter about China. If it is not a praise, do not tweet it. Chinese "Face" can make or break your sales.

 

[Cobbler]

I don't know man.
When the competition can put a 10 000+ FSD hardware on a 15 000 ICE car. A tesla at 35 000 still cost more to produce.

That's not gonna happen.
Don't expect to see an expensive FSD option in an entry ICE car anytime soon. Those will be offered on the high end class vehicles (Audi A8, Mercedes S, BMW 7...) only and will slowly be offered on lower cost vehicles in the coming years.

 

[Fact Checking]

Thanks for at least addressing the question, but if it's trivial why did it take Waymo so long to come up with a semi-satisfactory solution? They have millions of instances of left turns - this isn't a corner case with airborne cars that you need a huge fleet to capture. Tesla has no data advantage here. The neural network doesn't make these driving decisions, so their chip doesn't help.

You are taking certain assumptions for granted, and your conclusions are not necessarily true: the confidence in the neural network data feed depends on the chip's speed and on the quality of training the NNs got: i.e. it depends on the "chip" and on the "data advantage" in essence.

If on the other hand you are using primarily LIDAR, which has a number of limitations that might result in false negatives (objects not detected or detected too late), and don't have ~140 TFLOPs NN processing speed onboard (because LIDAR solves all the hard vision problems for you) then it might be prudent for Waymo to avoid unprotected left turns, to not get into at-fault accidents.

Based on the "data advantage" and the ability to test their NNs on a very large fleet Tesla can probably also determine the millions of miles it can drive doing unprotected left turns before the NN makes a serious mistake. Waymo possibly cannot.

You could be right in the end, but they way you are presenting this argument is way too over-confident:

The neural network doesn't make these driving decisions, so their chip doesn't help.

Again, that's factually wrong, the neural network very much makes a decision about the characteristics of detected objects, and it will also determine the amount of free driving space it can see and be confident about, and can be confident about not colliding with moving objects or obstacles.

I.e. the NN's very much feed into the driving logic that performs unprotected left turns and determines the probability of the FSD car getting into at-fault accidents.

Waymo knows the location, speed and (almost certainly) classification of every object, so why hadn't it been "trivial" for them?

While I agree that it's non-trivial, we don't know why Waymo is avoiding unprotected left turns:

• Too little processing power of their vision NNs?
• Too little confidence in LIDAR input?
• Limited field of view of LIDAR systems, which makes 90 degree turns problematic?
• Legal liabilities?
• Condition of their insurance policy?
• Is it a regulatory requirement/constraint?
• Or just (justified) caution?

Your assumption that Waymo not performing unprotected left turns cannot possibly be due to a technological limitation where Tesla's chip and data advantage might provide a break-through is unsupported.

Maybe all of Google's smart people are solving Go and StarCraft and winning ImageNet every other year. All that's left to work on the trillion dollar self-driving opportunity are interns?

Given the flaws of your own argument your sarcasm and ridicule is IMO unwarranted.

 

[Artful Dodger]

I don't know man.
When the competition can put a 10 000+ FSD hardware on a 15 000 ICE car. A tesla at 35 000 still cost more to produce.

What I think most are not thinking about is the willingness of rivals to burn cash in order to kill off the competition. It was quite an eye opening experience to watch Uber and DIDI fight over the past few years.

Hindsight, everyone probably agrees that it is stupid to go up against a gov owned corp with near infinite resources. But before such a duel took place, nobody knew whether attractive tech from Silicon valley will win against gov backed corp.

Probably a good idea that TSLA is such a Chinese Gov darling. One rule of thumb for Elon's twitter about China. If it is not a praise, do not tweet it. Chinese "Face" can make or break your sales.

Lol, well I know I won't be getting in any $25K "best of breed" Chevy Spark with lasers on the roof when instead I could be hiring a ride in one of the 3 SAFEST cars ever tested by the NHTSA.

Cheers!