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

[heltok]

What I'm confused about is what will a robotaxi do when a situation arises that even a human does not know what to do. e.g. crash ahead in the rain and police signalling to go some other direction than planned. Will they have someone in the mothership connect to the car and resolve as a remote driver? Can someone explain the plan?

I had one of those yesterday. At the roundabout I see that the exit was blocked due to construction. Drove out and tried find some other way around but it was the only way out of the peninsular I could see on the map so I drove back to the roundabout to have a look. Saw an orange sign saying ”tutti direzioni” which I thought should translate to ~all directions in Italian(I was in Switzerland but they speak Italian there, btw Swiss is not even a language that google translate recognizes). Had to follow like 5 of those orange signs and take a large detour to finally get out.

The bad news is that current path finding algorithms are not there and that it will require a lot of more general intelligence in the neural network to solve the problem the way I did. Not only does the neural network need to be able to read letters in images, it needs to be able to translate words and figure out the instructions in them and also know if they are relevant or not. GPT3 might be able to do one of these steps, combining this into the visual neural network will be tricky, and then we are talking 175B parameters, which might be tricky to run on an embedded computer today. There are some really clever people working on this I remember this podcast:

Start watching 20min. Elon and Ilya founded OpenAI from where Elon later poached Karpathy, so expect a lot of these ideas to have reached Elon and Andrej.

In many countries the traffic authorities will inform the map providers and automakers of these road blocks. The good news is that if there is enough Teslas on the road, there is a likelihood that one Tesla had driven the same detour as me before me. At some point I assume that Tesla will use all their vehicles for map making to make their own maps.

Third Row Tesla Podcast
@thirdrowtesla
Apr 13
Theoretically it should be possible to improve smart summon’s routing algorithms if Tesla had their own database of GPS points and driving data for Teslas in parking lots then they could just query to get a sense of where the lanes are, as long as Teslas have driven through
Elon Musk
@elonmusk
Apr 13
This will happen

Tesla today are using MapBox for maps. Tesla’s board member Ehrenpreis also is a board member of MapBox. Maybe Tesla should just aquire them, $1B market cap. I assume that Tesla could improve their business using their fleet and at some point just the value of the data that all Tesla cars gather will be a huge asset in itself.

 

[S3XY]

There was at least one video posted this week with FSD slowing for speed bumps, so that is already solved.

It worked where there were signs marking them. I saw another video (Teslraj I believe) where it didn't recognize them and there were no corresponding signs.

 

[Dutchie]

So what would be a better investment?TSLA or FSD?

 

[wipster]

Starting out about 420 in Germany and my Seahawks lost so I'm going to bed!

 

[StealthP3D]

Sharing thinking of writing more calls for Oct 30 435 as max pain seems to be 425. Reason being is that it seems likely that it won't be breached due to no impending news.

What do others think?

In my experience it's silly to say there is "no impending news". News will happen when it happens and the kind of news that moves markets in a big way is NEVER expected. That's why it moves the market. Personally, I wouldn't put my shares on the line on the assumption there will be no news. One tweet from Elon could make your day very bad.

 

[MC3OZ]

Gee, I wonder for how many miles each of those robotaxis can find customers and make money driving after say five years when there is 100 million robotaxis.

No doubt many more people in less affluent developing countries will be able to afford to hire personal mobility.

A 25K Robotaxi working 24 hours can do a lot of trips... Fuel is cheap, maintenance is low.
Even if a driver is required, that is no big deal.

But how much can these customers afford to pay?

In more wealthy countries Robotaxis can change more. But geo-fenced solutions may work for popular high volume routes, and public transport is a competitor. In this case a convenient high quality solution means Robotaxis sit idle for some part of the day.

Tesla will have a very low cost base, that will make it hard for the competition to win on price.
Because of low fleet costs, Tesla will have a large fleet add in the ability to go anywhere, Tesla should win on convenience.
In terms of quality Tesla will have good in car entertainment and productivity..

There are still important areas like cleaning to address, but otherwise it is hard for a competitor to gain a competitive edge other than novelty.

In terms of price, a competitor with deep pockets can certainly try, they may be prepared to lose money for a time in an effort to capture market share and/or they may get some assistance from local regulators...

Robotaxis will help recession proof Tesla.

In terms of the income per car per year, we need to know miles travelled and income per mile, it is very hard to put numbers on that.

 

[Buckminster]

Let me update this for ARK:

The Golden Goose. 100%. 100%. $22k (=$4400 =10 bagger in 4 years)

 

[Buckminster]

I was in Switzerland but they speak Italian there,

4 national languages - so they use a fifth on their currency (Latin).

 

[FrANce]

Let me update this for ARK:

The Golden Goose. 100%. 100%. $22k (=$4400 =10 bagger in 4 years)

Weren't these pre-split? Although 4.4k would warrant a happy goose dance, too!

 

[MABMAB]

I recall npv autonomous vehicle 200k per musk.
Elon Musk: Tesla cars should be worth $100k to $200k with Full Self-Driving package - Electrek
Musk's radical rethink of electric transport, and the value of Tesla and its cars

 

[Carl Raymond]

I'll repeat this again for effect: Nobody will pay $100K for FSD software per vehicle. Not now, not ever. Elon never said this would happen.

Tesla is beginning a production ramp that is going to take everyone's breath away. If FSD graduates from beta and becomes a driverless product, there will instantly be tens of millions of FSD Teslas on the road well before annual production reaches 20M. The more autonomous vehicles become commonplace, the less autonomy as a product will cost. Your $100K is a pipe dream.

Elon said one day everybody will have autonomy and Tesla's main advantage will become production. In other words, autonomy won't even be a distinguishing factor for Tesla.

Do you get it now?

Maybe, maybe not. I think FSD is going to be harder than most expect - a moonshot project. The illusion fooling people is that impressive feats of self driving can still be far far away from never requiring an interventionist driver.

When the US landed on the moon they probably expected Russians would be there within a few years, and space tourists soon after that.

So yes, it’s a pipe dream that Tesla succeeds in the robotaxi business, and it’s so hard that nobody else may be able to achieve it. Monopoly for many years is a possibility.

 

[CyberDutchie]

Weren't these pre-split? Although 4.4k would warrant a happy goose dance, too!

Yes, that’s what @Buckminster was implying with his $4400 10x in four years (2024) comment.

ARK’s been saying $22K pre-split for their bull case for a while now. More recently, Kathy mentioned she expects 45% YoY growth for the stock for the next five years. That translates to $1860 in 2024, based of today’s price.

I think ARK Invest is waiting to see the broader FSD rollout before updating their research.

 

[WarpedOne]

The illusion fooling people is that impressive feats of self driving can still be far far away from never requiring an interventionist driver.

No, the illusion fooling people is that "no intervention ever" is the only acceptable state of affairs.
Because no driver ever has needed any outside help on the road. Ever, not a single one.

Truth is this is pure FUD.

FSD is and will never be perfect because any naysayer will ALWAYS find a fault.
Its success will be demonstrated through statistics with non-zero fault rate.

 

[CyberDutchie]

No, the illusion fooling people is that "no intervention ever" is the only acceptable state of affairs.
Because no driver ever has needed any outside help on the road. Ever, not a single one.

Truth is this is pure FUD.

Correct. The Tesla Network no doubt will include provisions for dealing with stranded cars in their fleet.

 

[Tslynk67]

The 10-Q just dropped: 0001564590-20-047486 | 10-Q | Tesla, Inc.

 

[heltok]

The 10-Q just dropped: 0001564590-20-047486 | 10-Q | Tesla, Inc.

So +11B in assets, +2B in liabilities = $9B in 9 months. But almost all of this was through equity, ie we the shareholders are have taken the hit so that our shares could increase in value?

 

[jerry33]

I just encountered that exact situation in my AP1 car and was asking myself the exact same question. Essentially, the cars that wrecked had been moved to the side of the road, but debris was still in one lane. There was no officer directing traffic, but everyone was just driving into the emergency lane to avoid the debris. Obviously, I had to disengage AP, but my thoughts were the same as yours - there are no programmable rules for this situation. I just followed what everyone else was doing. How will a computer ever know how to make that decision?

This is an edge case. It's one of the things that the NN will look for when enough examples are there for it to learn from. No programming required.

 

[UkNorthampton]

Dolphins - in addition to bats as others have mentioned - have excellent vision and echo location.



I was a ride share driver in the detroit area in 2016 and had an apple employee in my car who laughed at my positive comments about teslas self driving program and was confident apple would rock it.

Also briefly worked in roof sales in that time period in the same area and had a ford engineer as a customer who had similar reactions.

What a difference 5 years makes.

Edit: fun fact, I did both of those jobs because I was doing some career transitioning and wanted to learn about the ridehailing and roofing markets first hand to get a sense of what tesla was potentially disrupting. Came away from both of those jobs feeling very confident in Teslas future. Took awhile tho.



Share price is fairly meaningless in this context. Market cap is the figure you should be looking at.
Cokes market cap is roughly half of Tesla.
​

A bit like Peter the Great learning shipbuilding. A great way to understand each area.

Czar Peter House (Netherlands) - Wikipedia

 

[heltok]

Possible Model Y MIC sighted:

5:40 into the video

 

[jerry33]

No, it absolutely is not. That's a myth.

Individual cars never ever learn. (troubleshoot'd be a nightmare for one thing- nor does the car have remotely the power, or tech, to label and train NNs onboard)

Changes in behavior only happen with firmware updates from Tesla- all AI learning/training happens back at HQ.

On the current wide-release versions of firmware, when you disengage only a very very tiny report goes to Tesla- No pictures, no video, just location is coordinates, how did you disengage (brake, wheel, stalk), speed, heading and time. That's it. Under 1kb of data in the report.

Useful to tell them "Hey AP has trouble with this one location" when a lot of reports in the same spot come in, but that's it.

I don't believe the OP meant individual cars. But I'm pretty sure a disengagement shows in the logs which can be parsed and eventually followed up on.