General Discussion: 2018 Investor Roundtable

[Drumheller]

Am I the only owner in here who has not yet received an invitation? Order placed on the evening of day 1.

When was the last time you logged into your My Tesla account?

 

[dc_h]

It is a copy

View attachment 280291

Post the original please.

MarketWatch - 17 minutes ago: FCC chairman urges commission to approve Elon Musk's SpaceX application for broadband satellite

How might this affect Tesla?

nothing for a few years. Longer term could provide very high speed low cost bandwidth to the fleet. If successful SpaceX could be much larger than Tesla. A global telecommunications company with ubiquitous high speed low latency and cost effective networks.

 

[sub]

I'm assuming there are some current owners that bought their model S/X after they made their reservation that were moved to the front of the line and have received an invite?

I'm in this situation, waited in line to reserve but bought a model X two months ago, my reservation has not changed. However, and this may be the reason, I used a different, easier to communicate email address for my model 3 reservation. Maybe they are not connecting my two accounts by name and no bump in the line?

Im not going to buy a model 3 right now so I'm not worried about it, just curious. Kept one of my two reservations after buying the X just in case, I'll only cancel once the line is gone.

 

[ValueAnalyst]

I’d like to put an end to the nonsense of Tesla being valued similarly to companies that produce several times more units.

Not only the “competition” produces vehicles that live in the past, Tesla’s ENTERPRISE value is a fraction of those of functionally antiqued road transporter (FART) makers.

More than half of the ENTERPRISE VALUE of many traditional FART makers is held bondholders, so market cap alone is not an informative comparison.

Be informed against FUD.

 

[Zaxxon]

Maybe they are not connecting my two accounts by name and no bump in the line?

From what I've read elsewhere, this is definitely the case. If your 3 reservation is not attached to your existing car's account/email, it is not going to get owner priority.

 

[Starno]

Not sure what to think of this Tweet from Andrej Karpathy (head of AI at Tesla). Is it also including Autopilot development (as it is using reinforcement learning) ??? :

Here's the link of the article he's tweeting about :

Deep Reinforcement Learning Doesn't Work Yet

 

[avoigt]

This is a video of two Germans receiving the M3 and testing it in various aspects. Its Video No5.

Most of you won't understand what the say as they speak German all the time but I want to post out for you following.

The guy with the beard said about himself that he is an extra critical character and did not think very positive about the Model 3 until he reviewed it. So he was more biased to the negative when he arrived. He sold cars in the past and is very knowledgeable about every aspect to assess a vehicle.

Two key takeaways that I want to bring to the attention of the forum:

1. Interior, he found it very well manufactured without any issue, gaps or other aspects that can be criticized. "An interior that you expect for a car of that price range".

2. Exterior: They did measure every gap and concluded its just perfect and there is nothing that needs to be readjusted or ist not as it supposed to be. Nothing standing out all gaps are perfectly aligned.

He also mentioned that he has no idea how Autonline/ Munro could possibly compare the M3 with the Kia in the 90. He sold them at the time and its just no true what they said he concluded. Model 3 is just in a very different league.

We Germans are grown up with an extra "Quality Gene" in our DNA, so having seen how this two guys are inspection the Model 3 and coming up with almost only positives makes me feel very well about my reservation!

 

[bdy0627]

I’d like to put an end to the nonsense of Tesla being valued similarly to companies that produce several times more units.

Not only the “competition” produces vehicles that live in the past, Tesla’s ENTERPRISE value is a fraction of those of functionally antiqued road transporter (FART) makers.

More than half of the ENTERPRISE VALUE of many traditional FART makers is held bondholders, so market cap alone is not an informative comparison.

Be informed against FUD.

It's Car Wars: SMART (SeMi-Autonomous Ranged Transporters) cars vs FART cars

 

[ValueAnalyst]

Not sure what to think of this Tweet from Andrej Karpathy (head of AI at Tesla). Is it also including Autopilot development (as it is using reinforcement learning) ??? :

View attachment 280503

Here's the link of the article he's tweeting about :

Deep Reinforcement Learning Doesn't Work Yet

It means "3 months maybe; 6 months on the outside" is really 3 years... who could've thunk it?

 

[mongo]

It means "3 months maybe; 6 months on the outside" is really 3 years... who could've thunk it?

@Starno
Well, that is one interpretation.
My thoughts are that they are not using DRL for the entire driving problem. Rather, they are using image recognition/ detection + physics model (path prediction ) + rules of the road.
If you already have a codified set of laws/rules for driving, it seems highly inefficient to then train a DNN to follow them via DRL. Especially since to do so would require implementing the entire law/rule set to produce the reward function.
Example: stop on red light
Examine scene, is there red light in our lane? If so stop behind stop line."Here is what you do"
V.s. negative reward for going through red light with no additional data regarding what a red light is, or a stop line. "Oh, you know what you did wrong/ figure it out yourself"

 

[22522]

Im not going to buy a model 3 right now so I'm not worried about it, just curious. Kept one of my two reservations after buying the X just in case, I'll only cancel once the line is gone.

Even though it is no cause for worry, you want to fix this, as it influences the information that is used to feed your curiosity.

 

[22522]

Yeah, the paper states clearly: "The rule-of-thumb is that except in rare cases, domain-specific algorithms work faster and better than reinforcement learning."

Tesla will have it working in no time (very soon) as they will copy the algorithms that people use and add "fuzzy logic" (may be using this term wrong) to make a good decision when the algorithms break down.

The task resolves to:

1) Create a set of appropriate situational algorithms.
2) Create a set of plan B, C, D, and Es for when each algorithm fails to work.
3) Use deep learning to shade the plan B decisions.

I expect Tesla:
1) is still growing their object identification library, and
2) has not acknowledged that human drivers track the center line by sighting off the left front corner of the hood. They don't register the vehicle in the center of the lane. The right side line is secondary information most of the time.

Once they do this:
a) Lane wandering goes away. The car tracks the left side of the roadway.
b) Collisions with concrete barriers on the driver's side do not happen.

 

[jhm]

Then you will have a very small market size. Tesla is aiming for what 500,000 a year? BMW M3 sales are well under 100,000 units a year. 500,000 is more than Camry, Accord and Malibu the three largest selling mid-sized sedans. To crack that market in the volumes Tesla needs price is very important. Sure some people will stretch and some have plenty of money, but not 500,000/year.

View attachment 280362

I believe the aspiration it to sell 500k/yr globally, but you are comparing with US only sales.

 

[mongo]

2) has not acknowledged that human drivers track the center line by sighting off the left front corner of the hood. They don't register the vehicle in the center of the lane. The right side line is secondary information most of the time.

Once they do this:
a) Lane wandering goes away. The car tracks the left side of the roadway.
b) Collisions with concrete barriers on the driver's side do not happen.

Very true about the hood line alignment. I took a rental SUV to the FH launch, so I did a lot of lane centering recalibration in my brain. However, I do not think it is quite that simple.

• If the left lane disappears, the vehicle needs a secondary alignment mechanism it can cleanly and smoothly transition to.
• The left hood/line offset needs to be adjusted based on lane size (less so based on vehicle width for these use cases) otherwise you will be centered in a normal lane, but potentially scrape the right side of the vehicle in a construction zone.
• Similarly, the vehicle needs to avoid encroaching vehicles on either side.
  
• It also needs steering damping to handle poorly painted (not straight) left lines.
• The AP camera is centered vs the left of center driver's view, so either lane line is similar for an algorithm point of view.

My guess is that the lane recognition is pretty good, but the mapping of that to steering input needs tuned. Maybe they are using the wrong coordinate system transform for car rotation vs camera position? Or need to allow for more error, as long as the lanes are not crossed.

 

[Dennisj00]

The Electrek article about ConEd embracing Powerpacks indicates that the US market is ready for loft-off!

 

[jhm]

I’d like to put an end to the nonsense of Tesla being valued similarly to companies that produce several times more units.

Not only the “competition” produces vehicles that live in the past, Tesla’s ENTERPRISE value is a fraction of those of functionally antiqued road transporter (FART) makers.

More than half of the ENTERPRISE VALUE of many traditional FART makers is held bondholders, so market cap alone is not an informative comparison.

Be informed against FUD.

Let me pivot this distinction between enterprise value and market cap. A capital intensive company with low growth will use a lot of debt so that the return on equity is competitive with other stocks. This, the high debt to equity is just compensating for low growth. But Tesla suffers from no low growth problems. It's able to exceed 50% annual organic growth in revenue. So it really does not need to lever up to have offer an exciting return on equity to shareholders.

So I do think that market cap is the better way to size up auto companies in value to shareholders because this metric does a better job of valuing growth than what debt does. Clearly if old FART makers where to swap out a bunch of debt for equity, they would dilute their shares and have a very hard time boosting their market cap. Put another way, it is much easier to increase enterprise value by issuing more debt than it is to increase market cap by issuing more debt. So growing market cap is a much more impressive feat.

Edit. Of course if you are simply trying to counter the FUD that uses a ratio of market cap per production units, then enterprise value would be a much smarter numerator. All capital is ultimately used for production. Also the unit of production ought to be something more like dollars of revenue rather than number of vehicles. Then we are talking about how efficiently capital is deployed to generate sales.

 

[Drax7]

Yeah, the paper states clearly: "The rule-of-thumb is that except in rare cases, domain-specific algorithms work faster and better than reinforcement learning."

Tesla will have it working in no time (very soon) as they will copy the algorithms that people use and add "fuzzy logic" (may be using this term wrong) to make a good decision when the algorithms break down.

The task resolves to:

1) Create a set of appropriate situational algorithms.
2) Create a set of plan B, C, D, and Es for when each algorithm fails to work.
3) Use deep learning to shade the plan B decisions.

I expect Tesla:
1) is still growing their object identification library, and
2) has not acknowledged that human drivers track the center line by sighting off the left front corner of the hood. They don't register the vehicle in the center of the lane. The right side line is secondary information most of the time.

Once they do this:
a) Lane wandering goes away. The car tracks the left side of the roadway.
b) Collisions with concrete barriers on the driver's side do not happen.

Basically More mobileye

 

[Curt Renz]

The Electrek article about ConEd embracing Powerpacks indicates that the US market is ready for loft-off!

Indeed: Tesla to deploy new Powerpack battery systems with NY’s largest electric utility

And this too: Tesla Model X all-electric SUV converted into police cruiser by Ontario Police

 

[TheDoktor]

It means "3 months maybe; 6 months on the outside" is really 3 years... who could've thunk it?

Call it 10 and I’m on board.

Highway driving? Sure, on the interstate. FSD in all environments? 10-20 years.

 

[mongo]

Pulling from market

Current owners have another vehicle should the early production M3s have an issue. They require less education. Its an incentive for people to buy the higher margin and more expensive vehicle, etc. Also, Tesla said from the start that current owners have priority and that you can cut the line if you purchase an S or X. The only people that object are non owners that have become impatient. Its a sentiment that I understand, but that doesn't make them right.

They already have a charger connection/ system for charging and less learning curve.

Anyone have a wrecked Tesla I can buy for a week?