TSLA Market Action: 2018 Investor Roundtable

[AudubonB]

...Therefore, I kindly request our gracious moderators to ban, for now, any talk of model 3 demand from the investors general and market action threads....

Anyone else with me? I guess you can like if you agree and disagree if it should be left here.

MY guess is, of course, as good as anyone else’s as to how Moderators think and what their decision-making process is,, but I’m rather suspecting that to the extent this thread gallops off in all directions at once like a bee-stung jackrabbit, which it does, then they are somewhat grateful that one at least can make a cogent argument that Model 3 demand numbers are somewhat relevant to Market Action discussions.

But what do I know?

 

[FirebirdAlpha]

That Barron's article is awfully written and poorly explained, but there is a nugget of truth in there.

The article misses the brand value. I'm short, but even I will admit that driving a Model S is an awesome experience. There's real, but intangible value there.

Even if Tesla were valued like a traditional automaker, its brand would still justify a best-in-class valuation.

The trouble is valuing the future of the automaker industry. The industry will have a huge shift in the value chain with the creation of Level 4 autonomous vehicles. And I really believe Google/Waymo is going to dominate the autonomous vehicles war.

The reason is Google's talent and access to capital. They are really ****ing good at bleeding edge AI and Machine Learning. Tesla might have a long-term data advantage (and I stress might), but they don't have the talent advantage, and Google is getting close. They're doing amazing stuff with Google Assistant, Google News, medical imaging, etc.

Waymo cars are racking up serious miles on real roads and highways with no accidents. You see them here in San Francisco, and you see them in Phoenix.

Credit to GM/Cruise though. They're promising a car without a steering wheel in 2019.

Google does match Tesla's data collection. Part of how Google is overcoming their lack of data is by outright purchasing car feeds. They then use Recaptcha to train their object recognition models (ever been asked "which of these contains a stop sign?"). Android phones and Google maps users give it road and traffic data.

All of that does introduce one hilarious risk unique to Google: monopoly risk.

Regardless, whoever wins the race to scalable Level 4 autonomous driving will be on the brink of an incredible, world-changing opportunity. I don't think it will be Tesla. They will have to pull off a come from behind, and they're behind Google.

Without that, the question of how you value "traditional automakers" goes away, because the value chain is massively disrupted.

 

[oneday]

That Barron's article is awfully written and poorly explained, but there is a nugget of truth in there.

The article misses the brand value. I'm short, but even I will admit that driving a Model S is an awesome experience. There's real, but intangible value there.

Even if Tesla were valued like a traditional automaker, its brand would still justify a best-in-class valuation.

The trouble is valuing the future of the automaker industry. The industry will have a huge shift in the value chain with the creation of Level 4 autonomous vehicles. And I really believe Google/Waymo is going to dominate the autonomous vehicles war.

The reason is Google's talent and access to capital. They are really ****ing good at bleeding edge AI and Machine Learning. Tesla might have a long-term data advantage (and I stress might), but they don't have the talent advantage, and Google is getting close. They're doing amazing stuff with Google Assistant, Google News, medical imaging, etc.

Credit to GM/Cruise though. They're promising a car without a steering wheel in 2019.

Google does match Tesla's data collection. Part of how Google is overcoming their lack of data is by outright purchasing car feeds. They then use Recaptcha to train their object recognition models (ever been asked "which of these contains a stop sign?"). Android phones and Google maps users give it road and traffic data.

All of that does introduce one hilarious risk unique to Google: monopoly risk.

Regardless, whoever wins the race to scalable Level 4 autonomous driving will be on the brink of an incredible, world-changing opportunity. I don't think it will be Tesla.

Without that, the question of how you value "traditional automakers" goes away, because the value chain is massively disrupted.

Certainly a cogent argument. This is indeed a long term risk. The necessity of the electrification of transportation including the fact that Google tends to be perfectly ok with selling their technology derisks these possibilities to some degree. The auto market is very large, autonomy is very likely to take a lot of time. But I will concede this is the best bear argument I’ve heard from you.

 

[LargeHamCollider]

What matters is not who is first to level 4.

What matters is who is first to level 4 AT scale WITH a reasonable price tag. Google leads in autonomy but will take 3-5 years to scale fleet. Tesla's fleet is scaling at 1000 cars/day but will prob be 1-2 years behind google to level 4.

 

[ggr]

That Barron's article is awfully written and poorly explained, but there is a nugget of truth in there.

The article misses the brand value. I'm short, but even I will admit that driving a Model S is an awesome experience. There's real, but intangible value there.

Even if Tesla were valued like a traditional automaker, its brand would still justify a best-in-class valuation.

The trouble is valuing the future of the automaker industry. The industry will have a huge shift in the value chain with the creation of Level 4 autonomous vehicles. And I really believe Google/Waymo is going to dominate the autonomous vehicles war.

The reason is Google's talent and access to capital. They are really ****ing good at bleeding edge AI and Machine Learning. Tesla might have a long-term data advantage (and I stress might), but they don't have the talent advantage, and Google is getting close. They're doing amazing stuff with Google Assistant, Google News, medical imaging, etc.

Credit to GM/Cruise though. They're promising a car without a steering wheel in 2019.

Google does match Tesla's data collection. Part of how Google is overcoming their lack of data is by outright purchasing car feeds. They then use Recaptcha to train their object recognition models (ever been asked "which of these contains a stop sign?"). Android phones and Google maps users give it road and traffic data.

All of that does introduce one hilarious risk unique to Google: monopoly risk.

Regardless, whoever wins the race to scalable Level 4 autonomous driving will be on the brink of an incredible, world-changing opportunity. I don't think it will be Tesla.

Without that, the question of how you value "traditional automakers" goes away, because the value chain is massively disrupted.

OMG, I'm replying!

Re: recaptcha; I hadn't realized that google owned recaptcha. Thanks for that nugget of information. Anyway, they show a few (6-9) images that either do, or don't, contain a something (stop sign, store front...). That's very specialized. Now, if they showed pictures and asked "Which of these is an imminent collision", that might be different. They don't have a billion miles of driving data to solve some of the other problems. I think it's a bit like the supercharger network versus electric city cars... sure, they'll have their cute little flivvers, driving surface streets at 35mph, but Tesla will be taking people cross country on freeways.

Secondly, no one company ever completely wins the race. Whether Waymo or Tesla wins the race, Tesla or Waymo will be there too. (Who's on third?) Remember, Ford won the race with the Model T... and yet a decade later GM was the largest manufacturer.

 

[FirebirdAlpha]

What matters is not who is first to level 4.

What matters is who is first to level 4 AT scale WITH a reasonable price tag.

I agree, but they are charging ahead.

Google ordered 62,000 Chrysler Pacificas and 20,000 Jaguar SUVs. For delivery this year. All of them somewhere between level 3 and level 4.

I think you're underestimating Google's ability to scale here and just buy the market. These fleet purchases don't make a dent for them.

I'll also admit that I'm skeptical of Tesla's approach. I think you're going to need the LIDAR. I'll admit I could be wrong, but I think there are enough edge cases that make LIDAR necessary.

 

[FirebirdAlpha]

They don't have a billion miles of driving data to solve some of the other problems

Yes, but Google is doing it right now.

It doesn't matter how they solved the data problem. They DID.

They have cars on the road close to level 4 autonomy right now. And they're going from 10k --> 100k of them this year.

Tesla is still at zero. Still no cross-country trip. Still no journalist test drives. Still no test cars on real roads.

Tesla's approach can't just magically work overnight. They need to start testing on real roadways soon. They need to start marketing its release (put big "self driving Tesla" branding on the test cars on the road). They need to start building trust.

That's why I said they will need to "come from behind."

But they're behind Google. They're behind the second most valuable company in the world.

Investors are betting big on Google winning here, too. And if Google needs a 25B equity raise to buy a massive fleet, investors won't blink.

 

[YasB]

Re: recaptcha; I hadn't realized that google owned recaptcha. Thanks for that nugget of information. Anyway, they show a few (6-9) images that either do, or don't, contain a something (stop sign, store front...). That's very specialized.

It's called labeling and something Tesla still seems to have a lot of trouble with, according to those slides from Andrej Karparthy that were linked some weeks ago. Google has been doing that for years and seems to have a pretty good working system.

Now, if they showed pictures and asked "Which of these is an imminent collision", that might be different. They don't have a billion miles of driving data to solve some of the other problems. I think it's a bit like the supercharger network versus electric city cars... sure, they'll have their cute little flivvers, driving surface streets at 35mph, but Tesla will be taking people cross country on freeways.

It's almost impossible to judge, if there may be a collision from a single picture, since things aren't moving and you can't make assumptions about trajectories in a safe way. To say anything about trajectories you have to identify objects first, follow them over time and assign them a meaning. Only when you have done that, you can start to make assumptions about what they are going to do. It's improbable that this thing you identified as a bicycle is suddenly going at 50 mph or that billboard will cross the road. Labeling is one of the first steps you have to solve to achieve that. From what i've seen, which of course is only the stuff they published or that has leaked, Tesla isn't very far with it right now.

Also: I have been trying for a long time to find out what those "billion miles of data" actually contain, since that says nothing about the quality or amount of data. It's obvious they don't send much raw data like videos or an image every second because of the volume that would need. So what do they send back to the mothership?

 

[FirebirdAlpha]

Also, re:miles driven. Just remember, Google can acquire GM for a less than 10% dillution hit. That gets them 12 million vehicles per year.

Google investors know the stakes of this market. In 2016, Google went from 100 to 1k cars. Last year, they went from 1k to 10k. This year, they're going from 10k to 100k. Are they going to be at 1 million next year?

Uber investors have also realized this is a huge problem. Uber lost 4.5B last year, and a key part of their valuation has been participating in autonomous vehicles. Google pays for their cars in cash while Uber has been burning money.

It sure looks like Google is crushing Uber right now, especially when Google was willing to put $1 billiom into Lyft.

 

[gora321]

Yes, but Google is doing it right now.

It doesn't matter how they solved the data problem. They DID.

They have cars on the road close to level 4 autonomy right now. And they're going from 10k --> 100k of them this year.

Tesla is still at zero. Still no cross-country trip. Still no journalist test drives. Still no test cars on real roads.

Tesla's approach can't just magically work overnight. They need to start testing on real roadways soon. They need to start marketing its release (put big "self driving Tesla" branding on the test cars on the road). They need to start building trust.

That's why I said they will need to "come from behind."

But they're behind Google. They're behind the second most valuable company in the world.

Investors are betting big on Google winning here, too. And if Google needs a 25B equity raise to buy a massive fleet, investors won't blink.

Please don not underestimate the guy who wanted to buy ballistic missile and by the end of the flight figured out how to do it by himself.

 

[FirebirdAlpha]

Please don not underestimate the guy who wanted to buy ballistic missile and by the end of the flight figured out how to do it by himself.

You know what?

Even Elon Musk would still lose at a game of Go to Google's AI. I'm sure AlphaGo would be dying to take on Elon's AI any time.

My point is Google is really good at this as an organization. They are on the cutting edge of AI. They are years ahead of projections.

To say that Elon could just catch up undermines the research and achievements they've put forward. They are well versed in this area; this is Elon's first go.

 

[Remus]

I saw two people with no knowledge of how deep neutral network works pretending they're the experts, congratulating each other on successfully picking up scrums on the internet.

 

[YasB]

I saw two people with no knowledge of how deep neutral network works pretending they're the experts, congratulating each other on successfully picking up scrums on the internet.

Wow! Show me where, so i can make fun of them.

 

[gora321]

I saw two people with no knowledge of how deep neutral network works pretending they're the experts, congratulating each other on successfully picking up scrums on the internet.

True. To judge about the knowledge you should be the expert. So teach us.

 

[WarpedOne]

So, Google will be first to mars then... good to know.

 

[wnorris]

Investors are betting big on Google winning here, too. And if Google needs a 25B equity raise to buy a massive fleet, investors won't blink.

Good discussion on this topic and nice tidbit on reCaptcha. Now I know why they use the rather annoying street sign puzzles.

Any insight into how Google will service their fleet? If I recall correctly they have an agreement of some sort with Avis, but not sure how deep that goes. The ability to service the vehicles is a seldom discussed aspect of autonomy.

 

[Remus]

True. To judge about the knowledge you should be the expert. So teach us.

I like one of the Taylor Swift song that said you don't know what you don't know. In this topic at least I know what I don't know.

 

[BioSehnsucht]

recaptcha is more about validating their models than actually training, I suspect (as you're not clicking on the specific part of the image that should be tagged). IIRC it's been around long before any driving efforts, it was birthed by various other projects that predate their driving efforts by a fair amount I believe. That's not to say they can't use it for some pretty basic untargeted validation, but it's not really a direct path to training neural nets for autonomous cars.

 

[gora321]

I like one of the Taylor Swift song that said you don't know what you don't know. In this topic at least I know what I don't know.

You just repeated my previous thought.

I'm sorry guys to went out of topics. Quitting.

 

[Tslynk67]

There were about 100 Model 3s at the Charlotte, NC service center this morning with the entire front row lined up by color for all of the traffic to see.

Photo, or it didn't happen!