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

[Reciprocity]

Many credible posters here are doubting what FSD progress the Tesla AI team could possibly have achieved in the 3 years since they split with MobilEye - an automotive vision pioneer with 30 years of history.

Here's a comparison of AP1 (the last MobilEye iteration) and AP2:

Tesla Autopilot 1 vs. AP2: Incredible Improvements On Difficult Road

The Tesla testing takes place on a two-lane road with plenty of sharp turns, blind curves, dips, and elevation changes. Most of the time, you can see by the lane markings that passing isn't allowed on this road, which generally indicates that it's a challenge. Later in the video, there's a complete lack of lane lines.

"As you can see, Autopilot 1 struggles constantly on this drive, and I can assure you, it used to be WORSE on this road. I have previous tests with AP1."

"On this test, AP2.x absolutely NAILS IT! Successfully navigated the test roads"
​

I believe that's what a grounds-up, first-principles approach and the "Tesla data advantage" of a fleet of hundreds of thousands of vehicles gives you in just 3 short years.

And this AP1 to AP2 comparison is still on a HW2.5 basis and not the very latest firmware - HW3 gives ~21x more processing power.

I believe we've seen nothing yet in terms of FSD magic ...

People forget, the mobileye in Tesla's on AP1 are highly leveraged using Tesla code. Meaning off the shelf mobileye won't do what AP1 does today and AP1 keeps improving as well. But yes, Tesla did need to completely replace the core functionally of Mobileye which is the vision recognition, arguably the the most complex part. Tesla also started at almost the exact time of the breakup to develop the fsd chip and board, which surpasses where eyeq4 which isn't even out yet. They essentially replaced a $15B company with a 30 years head start in 3 years and over the next 3 years, Intel and Google will be so far in the rearview mirror that you wont be able to see them. Google doesn't have the data and is overly reliant on lidar. Mobileye doesn't have the data either because their partners won't give them that info, tough some map data is shared with mobileye which has a great deal of value but still only 1/1000th the value Tesla has with it's data collections capabilities.

To most the lead is not apparent and the meaning us even less understood. The head is insurmountable never AI and machine learning thrive with data. Yes why Google is so good at search and advertising. But it's also why Google lags on driving and has to the simulators to train their NNs. Their simulator can only be as good as they can make it which yes as complex a problem as creating self driving cars. They are equally difficult. Simulation still has a place and driving policy is one of them. But without the real world examples in large volumes, you can't train the NN to deal with it and you can't even know it needs to be dealt with until it's too late. If you look at the goal, 5 nines or 6 or 7 (99.99999%). You need a billion corner cases solved. So you need billions of miles driven. Only Tesla has that data for the foreseeable future. This cannot be understated. You cannot solve fsd without solving a billion corner cases and you can't even find them without a billion miles of data. Gathering and identifying the corner cases are almost as complex as creating a perfect simulator. I say almost because what Tesla is doing is creating a system that automatically identifies corner cases and then gathers examples then humans get involved to annotate the images. Then the process happens again to identify more occurrences based on annotations. Then those are fed into the ML NN to train. This then goes into shadow mode to validate and the whole process starts over until it's completely validated. The way Tesla does it, they can leverage massive amounts data with little human annotation. No one else is doing this, everyone else has an army of people annotating images. They are at ridiculously far behind it's not even funny.

 

[Todd Burch]

I think in US everybody knows Tesla.

You'd be AMAZED at how many Americans haven't even heard of Tesla, STILL. Lots of people are clueless.

I still get people pulling up next to me at stoplights:

Dude: "Hey, what kind of car is that?"
Me: "It's a Tesla."
Dude: "Nice, man."
Me: "Thanks"
Dude: "Who makes it?"
Me: "Tesla"
Dude: "Is that Italian?"
Me: "No, they're made in California."
Dude: "Cool, are they fast?"
Me: "Yes, very."
Dude: "What's the gas mileage?"

Yes, there are lots of clueless people out there.

 

[RobStark]

I think in US everybody knows Tesla.

Most Americans don't know Tesla is a company that makes electric cars.

Most people under 35 that live in cities on West Coast,Northeast,and Florida know Tesla is a company that makes electric cars. In these areas Tesla also has good awareness among people under 60.

Most new cars are purchased by people over 35.

From Idaho to Georgia and from Minnesota to Louisiana Tesla could use a brand awareness boost.

Tesla doesn't need more demand for $40k Model 3s but it could use more demand for $60k Model 3s in the USA. More demand for long range and performance Model S would also be helpful.

Tesla doesn't need to boost brand awareness from people that live on Youtube or listen to NPR.

Tesla could use a boost to brand awareness among people who live the NFL and participate in NFL Fantasy Leagues. And their high Holy Day is Super Bowl Sunday, watching the Super Bowl on TV is a religious obligation.

 

[Singuy]

People forget, the mobileye in Tesla's on AP1 are highly leveraged using Tesla code. Meaning off the shelf mobileye won't do what AP1 does today and AP1 keeps improving as well. But yes, Tesla did need to completely replace the core functionally of Mobileye which is the vision recognition, arguably the the most complex part. Tesla also started at almost the exact time of the breakup to develop the fsd chip and board, which surpasses where eyeq4 which isn't even out yet. They essentially replaced a $15B company with a 30 years head start in 3 years and over the next 3 years, Intel and Google will be so far in the rearview mirror that you wont be able to see them. Google doesn't have the data and is overly reliant on lidar. Mobileye doesn't have the data either because their partners won't give them that info, tough some map data is shared with mobileye which has a great deal of value but still only 1/1000th the value Tesla has with it's data collections capabilities.

To most the lead is not apparent and the meaning us even less understood. The head is insurmountable never AI and machine learning thrive with data. Yes why Google is so good at search and advertising. But it's also why Google lags on driving and has to the simulators to train their NNs. Their simulator can only be as good as they can make it which yes as complex a problem as creating self driving cars. They are equally difficult. Simulation still has a place and driving policy is one of them. But without the real world examples in large volumes, you can't train the NN to deal with it and you can't even know it needs to be dealt with until it's too late. If you look at the goal, 5 nines or 6 or 7 (99.99999%). You need a billion corner cases solved. So you need billions of miles driven. Only Tesla has that data for the foreseeable future. This cannot be understated. You cannot solve fsd without solving a billion corner cases and you can't even find them without a billion miles of data. Gathering and identifying the corner cases are almost as complex as creating a perfect simulator. I say almost because what Tesla is doing is creating a system that automatically identifies corner cases and then gathers examples then humans get involved to annotate the images. Then the process happens again to identify more occurrences based on annotations. Then those are fed into the ML NN to train. This then goes into shadow mode to validate and the whole process starts over until it's completely validated. The way Tesla does it, they can leverage massive amounts data with little human annotation. No one else is doing this, everyone else has an army of people annotating images. They are at ridiculously far behind it's not even funny.

So how is Tesla exactly solving corner cases today with city driving when the feature is not available therefore no disengagements are logged from let's say an incorrect turn from an intersection? Are they just comparing what the ai would do vs what actually happened in shadow mode?

 

[anthonyj]

What to do with these options... sitting on massive losses right now

View attachment 407545

If it was me, hold and sell on strength. Don’t give into the FOMO. As soon as you are in the green or close to it, sell. Don’t be me. I have similar losses but in stock

 

[anthonyj]

You'd be AMAZED at how many Americans haven't even heard of Tesla, STILL. Lots of people are clueless.

I still get people pulling up next to me at stoplights:

Dude: "Hey, what kind of car is that?"
Me: "It's a Tesla."
Dude: "Nice, man."
Me: "Thanks"
Dude: "Who makes it?"
Me: "Tesla"
Dude: "Is that Italian?"
Me: "No, they're made in California."
Dude: "Cool, are they fast?"
Me: "Yes, very."
Dude: "What's the gas mileage?"

Yes, there are lots of clueless people out there.

Omg, I smell a short seller. Tesla doesn’t need advertising. Everyone on Twitter, YouTube and TMC knows about Tesla, so this is obviously untrue FUD.
/s

 

[dc_h]

Sadly I agree on a couple of their complaints. (The private conference calls where they don't even release the recording immediately afterwards, when it should have never been private to begin with.)

If you throw out a 100 or so observations, you’re going to have something stick and then hope you get a sense of objective truth on all the whompy wheels nonsense. Imagine the pressure they put on the media where they don’t have to pretend to be objective and how creepy to be naming anyone on twitter who’s a fan. You wonder if Tesla fans and sustainable energy fans around here don’t need a dedicated PAC to represent us and file class actions against some of thie dishonest FUD and bullying.

 

[JusRelax]

People forget, the mobileye in Tesla's on AP1 are highly leveraged using Tesla code. Meaning off the shelf mobileye won't do what AP1 does today and AP1 keeps improving as well..

Can anybody understand Vietnamese? This is what mobileye can do off the shelf:

 

[Reciprocity]

So how is Tesla exactly solving corner cases today with city driving when the feature is not available therefore no disengagements are logged from let's say an incorrect turn from an intersection? Are they just comparing what the ai would do vs what actually happened in shadow mode?

I use AP in the city all the time. There are two kinds of disengagements in this case as well. There are AP disengagements and there to shadow mode which drivers in the background and anytime a human does something different then what the fsd computer would do, it can be logged and reviewed. Clearly there are billions on instances where this would happen so it's a huge undertaking to create a system to separate the wheat from the chaff. It appears that is exactly what they have done. Found a way to use the mountain of data to find and train those one in a million instances. The more cars, the more miles the more examples.

 

[Singuy]

Omg, I smell a short seller. Tesla doesn’t need advertising. Everyone on Twitter, YouTube and TMC knows about Tesla, so this is obviously untrue FUD.
/s

I would say very few Americans know correct information about Tesla thanks to FUD and noise. Not too many people follow the company religiously as if they have money on the line. I don't blame them because understanding Tesla is pretty close to a full time job. It takes a few hours to dissect the noise from a FUD article...and Tesla gets a barrage of them daily.

 

[mongo]

OT Semi
Let's rehash at
Lets work out the Tesla Semi-Truck Technical Specs

Tesla semi has a massive transmission that is just as heavy as the differential. Possibly heavier since the motor has more torque.

The drive units are 3 or 4 stage reduction gears direct to half shafts. No transmission, but the final gear or two have more torque on them which impacts their size. The 3's differential and gears are uncluded in the <200 lb weight.

 

[Revhappy2]

What to do with these options... sitting on massive losses right now

View attachment 407545

This post has found its way on bagholder quotes:
Bag Holder on Twitter

 

[Singuy]

I use AP in the city all the time. There are two kinds of disengagements in this case as well. There are AP disengagements and there to shadow mode which drivers in the background and anytime a human does something different then what the fsd computer would do, it can be logged and reviewed. Clearly there are billions on instances where this would happen so it's a huge undertaking to create a system to separate the wheat from the chaff. It appears that is exactly what they have done. Found a way to use the mountain of data to find and train those one in a million instances. The more cars, the more miles the more examples.

Yeah but AP in city driving is pretty much lane keeping. It doesn't recognize signs and can't do any of the actual self driving stuff like intersections and turns. I mean we can all agree teslas can drive a straight line pretty well when there are line markings( and not so much when there isn't).

 

[Tslynk67]

A brief warning - noone in the know expects this to happen, including Elon, JB, and Jeff Dahn. What has been driving the EVolution is small gains [in cost - chemistry - longevity - charge&discharge performance - capacity - weight] that add up over time.

By the way, great news from VAG where the board has OKd a 1B EUR investment in a European cell factory. However, Diess inserted some riders concerning cost and political support.


Now, once again, the focus on the trade deficit is looking at a symptom. Back in the mid 80s when I was a teen, I thought it would spell our doom. Learned something since then. Big damage will ensue from going loco - without a cure in sight. From April 2018 [!]:

It was an easy example to pull out of MA. Tesla can potentially drop big news without warning, that was my point.

 

[Reciprocity]

Yeah but AP in city driving is pretty much lane keeping. It doesn't recognize signs and can't do any of the actual self driving stuff like intersections and turns. I mean we can all agree teslas can drive a straight line pretty well when there are line markings( and not so much when there isn't).

How do you know what it does and does not recognize in shadow mode? Green has shown that it does see traffic lights today. So before we get to use the features, it runs in shadow mode. Also there are mounts of examples where the car would have done something differently then the human, this creates examples for training. I think your not understanding what I am saying. AP is always engaged even if you are not using it. When you turn the wheel to avoid something that AP would have proceeded to go through, that creates a delta. if the Delta is large enough, say the car moved x feet off the path the car would have taken, then its flagged for someone potentially to look at. Now they cannot look at everyone, but as they work through the process they will have less and less to even consider.

Elon mentioned and they showed a picture of a new system they are building. I believe its based on the returned DrivePX computers that they will get out of the cars as they upgrade them. They are going to rack them and use them for training. Imagine 100k AP2.0/2.5 computers in a rack that are crunching corner cases. Overtime those would be replaced with FSD1 and FSD2 computers once the costs come down and maybe a die shrink so that the costs of run the damn thing will save enough to pay for them.

 

[Skryll]

If Tesla ever makes an ad, what would be the key messages?

1) fast and spirited driving
2) FSD/AP
3) environment / emission
4) cost of ownership
5) reliability
6) OTA/ tech
7) American auto taking back the lead after many decades
8) spacex package in roadster
9) mobile servicing


I would primarily focus on #s 4, 5, 9, and 2. While wrapping in 3 as why EV. Other points are well known or not relevant.

To me the top points that most people don't know, or even fully understand when you tell them once:

1) convenience:
- why go to gas stations, oil change and smog check appointments and frequent maintenance repairs as the car gets older, unless you are retired and need a reason to get out to meet and interact with people. I prefer to just plug in like my iphone, over night, and be full in the morning, every morning.
- Never jump start your car again because you left the lights on
- Never scratch ice in the winter again, just preheat from your iphone, before even getting up from the breakfast table
- road trips much easier rwith autopilot and when coordinated with food and bathroom breaks with supercharging stops there is no waiting around, just plug in and walk away do what you actually want to do and come back when it is done
- ease of navigation from calendar entries (perfect for realtors showing people around), sending phone contacts to your car, voice navigation google style ('navigate to a pharmacy by broadway' lists rite aid, walgrerens, etc), showing the itinerary including charging stops for long distance trips
- quiet serene ride, no humming, vibrating, shaking, grunting, moaning even when accelerating swiftly,
- storage space in the frunk instead of car parts

2) total cost of ownership
- electricity costs about a quarter of what gas costs the same distance driven, even less when adding solar to your home.
- less maintenance cost, brake pads last many years, regeneratve braking
- battery pack warranty 8 years and most people still have long range after that while cost over the next 8 years comes down massively

3) safest cars, keep your family safe,
- not just because of massive amount of airbags and crumble zone,
- but also because of autopllot taking over hourrs of rmicrromanaging lane and distance keeping, avoiding rear ending in stop and go traffic etc
- climate control (dog mode) to keep your pets safe, cabin overheat protection
- dashcam / sentry mode features

 

[Singuy]

How do you know what it does and does not recognize in shadow mode? Green has shown that it does see traffic lights today. So before we get to use the features, it runs in shadow mode. Also there are mounts of examples where the car would have done something differently then the human, this creates examples for training. I think your not understanding what I am saying. AP is always engaged even if you are not using it. When you turn the wheel to avoid something that AP would have proceeded to go through, that creates a delta. if the Delta is large enough, say the car moved x feet off the path the car would have taken, then its flagged for someone potentially to look at. Now they cannot look at everyone, but as they work through the process they will have less and less to even consider.

Elon mentioned and they showed a picture of a new system they are building. I believe its based on the returned DrivePX computers that they will get out of the cars as they upgrade them. They are going to rack them and use them for training. Imagine 100k AP2.0/2.5 computers in a rack that are crunching corner cases. Overtime those would be replaced with FSD1 and FSD2 computers once the costs come down and maybe a die shrink so that the costs of run the damn thing will save enough to pay for them.

I am wondering how miles are being logged as training material. It's obviously not disengagements when it comes to intersections and stop signs. Which means Tesla must be comparing what the driver is doing to what the AI would have done.

 

[lascavarian]

They are going to rack them and use them for training. Imagine 100k AP2.0/2.5 computers in a rack that are crunching corner cases. Overtime those would be replaced with FSD1 and FSD2 computers once the costs come down and maybe a die shrink so that the costs of run the damn thing will save enough to pay for them.

I appreciate the thinking here but I don't know if this has had any substantiation from Tesla?

 

[Remus]

To me the top points that most people don't know, or even fully understand when you tell them once:

1) convenience:
- why go to gas stations, oil change and smog check appointments and frequent maintenance repairs as the car gets older, unless you are retired and need a reason to get out to meet and interact with people. I prefer to just plug in like my iphone, over night, and be full in the morning, every morning.
- Never jump start your car again because you left the lights on
- Never scratch ice in the winter again, just preheat from your iphone, before even getting up from the breakfast table
- road trips much easier rwith autopilot and when coordinated with food and bathroom breaks with supercharging stops there is no waiting around, just plug in and walk away do what you actually want to do and come back when it is done
- ease of navigation from calendar entries (perfect for realtors showing people around), sending phone contacts to your car, voice navigation google style ('navigate to a pharmacy by broadway' lists rite aid, walgrerens, etc), showing the itinerary including charging stops for long distance trips
- quiet serene ride, no humming, vibrating, shaking, grunting, moaning even when accelerating swiftly,
- storage space in the frunk instead of car parts

2) total cost of ownership
- electricity costs about a quarter of what gas costs the same distance driven, even less when adding solar to your home.
- less maintenance cost, brake pads last many years, regeneratve braking
- battery pack warranty 8 years and most people still have long range after that while cost over the next 8 years comes down massively

3) safest cars, keep your family safe,
- not just because of massive amount of airbags and crumble zone,
- but also because of autopllot taking over hourrs of rmicrromanaging lane and distance keeping, avoiding rear ending in stop and go traffic etc
- climate control (dog mode) to keep your pets safe, cabin overheat protection
- dashcam / sentry mode features

Half way through the quarter and no Raven cars, I think Tesla has a production problem

 

[schonelucht]

Apparently Tesla just settled for $13M in Tapia v Tesla Motors, related to injuries sustained by a janitor pinned between two cars on the Fremont test track. Trial was just about to start (jury selection completed). Since Tesla is self-insured this goes straight to the bottom line.