Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
The next big milestone for FSD is 11. It is a significant upgrade and fundamental changes to several parts of the FSD stack including totally new way to train the perception NN.

From AI day and Lex Fridman interview we have a good sense of what might be included.

- Object permanence both temporal and spatial
- Moving from “bag of points” to objects in NN
- Creating a 3D vector representation of the environment all in NN
- Planner optimization using NN / Monte Carlo Tree Search (MCTS)
- Change from processed images to “photon count” / raw image
- Change from single image perception to surround video
- Merging of city, highway and parking lot stacks a.k.a. Single Stack

Lex Fridman Interview of Elon. Starting with FSD related topics.


Here is a detailed explanation of Beta 11 in "layman's language" by James Douma, interview done after Lex Podcast.


Here is the AI Day explanation by in 4 parts.


screenshot-teslamotorsclub.com-2022.01.26-21_30_17.png


Here is a useful blog post asking a few questions to Tesla about AI day. The useful part comes in comparison of Tesla's methods with Waymo and others (detailed papers linked).

 
Last edited:
During the Tesla SHM today, Elon basically defined his meaning of achieving FSD: human level safety without any supervision.

My opinion: Tesla would have to get to at least 5x human safety to deploy robotaxis. It might actually be easier to get to 5x than we think though...

 
During the Tesla SHM today, Elon basically defined his meaning of achieving FSD: human level safety without any supervision.

My opinion: Tesla would have to get to at least 5x human safety to deploy robotaxis. It might actually be easier to get to 5x than we think though...

We’ve used 4 pages of debate on the earth being round. Getting 5x smarter then a human might be easier then we think.
 
Estimating the FSD Beta miles shown at 2023 Annual Shareholders Meeting, 11.x has really increased the monthly rate with highway driving:

View attachment 938496
That’s a great example of ‘lying with statistics.’ While technically true, highway driving is completely different from city driving when it comes to miles and technical difficulty. The addition of highway to FSD (which was already largely accomplished via the AP routines) simply serves to skew the data. At the very least they should have a marker noting when highway miles were added but to make it meaningful they need to separate out the two sets of data.
 
That’s a great example of ‘lying with statistics.’ While technically true, highway driving is completely different from city driving when it comes to miles and technical difficulty. The addition of highway to FSD (which was already largely accomplished via the AP routines) simply serves to skew the data. At the very least they should have a marker noting when highway miles were added but to make it meaningful they need to separate out the two sets of data.
This is incorrect.

Yes, city driving provides different and more difficult tasks, but highway driving being on Beta instead of NoA still trains and provides useful information to add to the NN. Think about all of the edge cases that still take place on the interstate...it's crazy to say that's not useful.

Do you think any ADAS company will separate the two? no, because that's ridiculous.
 
I think you're more right than me on this part.

I still think there are multiple interpretations of the description. If it's L4, there's no ODD specified. If it's L5, it doesn't say in all weather conditions or unlimited ODD.

I agree there's no precise ODD specified, but they do say "in almost all circumstances" so announcing "won't work at night" or "won't work if there's any rain" wouldn't be possible with that having been said... and they also say "designed to conduct both short and long distance trips" so they can't announce "only for drives of 5 miles or less" or something.... and they do list a bunch of specific things it explicitly will handle including urban streets even without lane markings, complex intersections with lights, signs, and roundabouts, and densely packed freeways at high speed. Plus driverless parking and summon at destinations. So they can't say "won't work without clear lane markings" either.

I suspect they could reasonably get away with delivering an ODD that has exclusions like ground-covered-in-snow or during a hail storm or even in countries Tesla does not sell cars... but otherwise the list of things it WILL include that they provided is pretty long, and broad.

That's why there's potentially even an argument to be made that it's L5.... but I wouldn't love your chances in court if you tried to base a lawsuit on that idea.... whereas I'd lay pretty solid bets on one putting forward the idea they promised at least L4 with a reasonably broad ODD.


Increasingly, and especially with Elon saying it wasn't economically reasonable to upgrade existing cars beyond HW3, I believe this will end up with the pre-march-2019 folks getting a full-with-interest refund of FSD because I don't believe Tesla can deliver what they promised them on existing HW. I suspect Tesla will offer other things as alternatives that might be appealing- like a free one-time transfer of FSD to a new car- but ultimately will need to have a full refund as at least an option offered.

The tricky bit comes with the post 3/19 buyers... Tesla clearly changed the wording to greatly reduce the amount they were promising to delivery during the sales process...So if you ONLY consider that- fsdb once it's officially wide release to everyone- means they have been fully delivered what they were promised during the purchase.... but Elons presentations and statements in both interviews, quarterly earnings calls, autonomy day, and elsewhere only continued doubling down on the "You will have a money making robotaxi with your current car!" rhetoric that went WELL beyond the promises to buyers post 3/19, so that'll either mean some messy lawsuits when they can't deliver that on current HW, or they have to offer a lot broader refund/transfer options to people.



During the Tesla SHM today, Elon basically defined his meaning of achieving FSD: human level safety without any supervision.

My opinion: Tesla would have to get to at least 5x human safety to deploy robotaxis. It might actually be easier to get to 5x than we think though...



FWIW, in the original pre 3/19 FSD description it promised to deliver "full self-driving in almost all circumstances at what we believe will be a probability of safety at least twice as good as the average human driver"

That said, there's no actual requirement to reach a certain x safer level to deploy- legally speaking. There certainly might be for it to make economic sense from a liability perspective though.
 
That is NOT a possible option in the Full Self Driving Levels. Until its is SAE L3 or above you are required to pay attention all the time, hands off or hands on is irrelevant. You are in full control of the car.

View attachment 938238
I agree FSD is just cruise control, nothing more except hype, that is not what the Kickstarter huckster sold me, but that is all I will ever get is driver assistance cruise control. After 4 years it still can't park in my garage with vision, radar, ultrasonics, OpenAI, and billions of miles of bigdata to optimize the model. Tesla has lost the battle of the nines.
ChatGPT is still alpha at best. In real world test it fails. I asked it to write a script to convince my girl for special favors, I did not need a computer to figure out how to be slapped.

They should market FSD as Full partnership with the computer to assist the driver, even though Tesla takes no responsibility for the hardware or software and the FULL RESPONSIBILITY for your vehicles safe diving is the responsible of the owner of the vehicle (the insured).

So far I have not seen the telemetry data used to do anything else but to put blame on the driver and excuse Tesla's vehicle and advanced driving systems (show me one accident where Tesla admitted the car even shares responsibility)
 
Last edited:
Chat GPT hasn't surpassed human reasoning. It doesn't even know what is fake or real on the internet when it scours for answers.

But looking forward to Robotaxi is a pipedream at this point.
I wonder how many Tesla owners actually want to use FSD to make money using Robotaxi? I suspect the percentage is extremely small. Maybe 1-3%. The overwhelming value of FSD has nothing to do with Robotaxi. Potential Robotaxi fleet owners have an interest but that's about it.
 
I wonder how many Tesla owners actually want to use FSD to make money using Robotaxi? I suspect the percentage is extremely small. Maybe 1-3%. The overwhelming value of FSD has nothing to do with Robotaxi. Potential Robotaxi fleet owners have an interest but that's about it.
I don't know, but between here, Reddit, and FB groups there is a vocal group that believe it's coming soon and can't wait.

Many stock owners feel like it's a big deal because Elon says it will increase the value of your car 5x.
 
  • Like
Reactions: MARKM3
For those of you that read this, it will require a little bit of a seeing the forest for the trees.

Logically speaking, FSD has to allow paid ride sharing to increase the value of the car. Even if the car drives around for you, there is no highly economic impact to TSLA stock or car valuation other than making Tesla's a little more desirable to those that appreciate driver assistance features. It is possible that will drive some sales. Tesla charges far too much for FSD.

In fact, the current price for FSD and the monthly subscription price, prices out over half of Americans that live paycheck to paycheck. The business model economics of a Tesla car and FSD fundamentally disagree with one another - Tesla's are becoming more and more affordable and cheaper to operate than ICE while FSD becomes more and more unaffordable and is priced equally to a used car or 1/3 the value of a new Tesla.

For many of the Model 3/Y owners, having your car drive you around without intervention is just not worth 1/2-1/3 the value of your car, 15k, or $200/month. Sure it's cool - but like Elon said, you use your car maybe 10 hours a week, an extremely small portion. Why would anyone fork out the money for it with such little usage if the economics of FSD never materialize into your car being a Robotaxi? There is very little personal gain ROI other than the cool and shock factor to friends/family.

There is a lot of work to be done with the FSD business model, and a step change valuation will only be possible with a materialized ride sharing program.
 
For those of you that read this, it will require a little bit of a seeing the forest for the trees.

Logically speaking, FSD has to allow paid ride sharing to increase the value of the car. Even if the car drives around for you, there is no highly economic impact to TSLA stock or car valuation other than making Tesla's a little more desirable to those that appreciate driver assistance features. It is possible that will drive some sales. Tesla charges far too much for FSD.

In fact, the current price for FSD and the monthly subscription price, prices out over half of Americans that live paycheck to paycheck. The business model economics of a Tesla car and FSD fundamentally disagree with one another - Tesla's are becoming more and more affordable and cheaper to operate than ICE while FSD becomes more and more unaffordable and is priced equally to a used car or 1/3 the value of a new Tesla.

For many of the Model 3/Y owners, having your car drive you around without intervention is just not worth 15k or $200/month. Sure it's cool - but like Elon said, you use your car maybe 10 hours a week, an extremely small portion. Why would anyone fork out the money for it with such little usage if the economics of FSD never materialize into your car being a Robotaxi? There is very little personal gain ROI other than the cool and shock factor to friends/family.

There is a lot of work to be done with the FSD business model, and a step change valuation will only be possible with a materialized ride sharing program.
Most FSD owners aren't looking for a Robotaxi. They want the car to drive them, it's an upgrade like different color seats, nicer engine, etc.

There's a long list of reasons why most don't believe a Robotaxi while you are at work will work logistically; charging, what if you need your car, damage to your car, issues with the car, liability, licensing for it, etc. or mechanically; the current camera placement will always require the car to creep into dangerous situations, weather, no rear cross-traffic alerts, etc.

Robotaxi is akin to colonizing Mars. It's a nice talking point, but not feasible in the near future.
 
  • Like
Reactions: navguy12 and kabin
Most FSD owners aren't looking for a Robotaxi. They want the car to drive them, it's an upgrade like different color seats, nicer engine, etc.

There's a long list of reasons why most don't believe a Robotaxi while you are at work will work logistically; charging, what if you need your car, damage to your car, issues with the car, liability, licensing for it, etc. or mechanically; the current camera placement will always require the car to creep into dangerous situations, weather, no rear cross-traffic alerts, etc.

Robotaxi is akin to colonizing Mars. It's a nice talking point, but not feasible in the near future.

You're exactly correct, so there is no meaningful impact with the current design to Tesla profitability as they move to more affordable cars. The more affordable the car, the less affordable FSD is to that owner so less take rate.

The take rate will scale with the price (price/demand) and unless robotaxi's happen, there will be no meaningful step change to the stock price like Elon mentioned.

Just like ChatGPT, it creates a lot of buzz but creates little value to most consumers.
 
You're exactly correct, so there is no meaningful impact with the current design to Tesla profitability as they move to more affordable cars. The more affordable the car, the less affordable FSD is to that owner so less take rate.

The take rate will scale with the price (price/demand) and unless robotaxi's happen, there will be no meaningful step change to the stock price like Elon mentioned.

Just like ChatGPT, it creates a lot of buzz but creates little value to most consumers.
Eventually Tesla will release a robotaxi car that will compete with Waymo and Cruise. They will use all the FSD Beta training to drive it, but as far as current fleet being able to turn into a robotaxi, it's complete fiction.
 
  • Like
Reactions: pilotSteve
For many of the Model 3/Y owners, having your car drive you around without intervention is just not worth 1/2-1/3 the value of your car, 15k, or $200/month.

Strong disagree.

$200/mo that could let you entirely eliminate owning a second or third car for example if it can be remotely sent places or remotely summoned... could take your kids to/from school and sports events and friends houses and the movies and anywhere else without you needing to drive them or pick them up OR worry about your teen driving themselves and crashing... -- not to mention folks being able to go out for an evening and drink without concern for getting home safely, or enable your car to drive you while you sleep on a long trip?

Mind you- I don't think current HW is actually capable of it-- but if it were it'd be a massive driver of Tesla revenue even before you got to ridesharing.
 
Strong disagree.

$200/mo that could let you entirely eliminate owning a second or third car for example if it can be remotely sent places or remotely summoned... could take your kids to/from school and sports events and friends houses and the movies and anywhere else without you needing to drive them or pick them up OR worry about your teen driving themselves and crashing... -- not to mention folks being able to go out for an evening and drink without concern for getting home safely, or enable your car to drive you while you sleep on a long trip?

Mind you- I don't think current HW is actually capable of it-- but if it were it'd be a massive driver of Tesla revenue even before you got to ridesharing.

That is ridesharing and robotaxi. With your example you gave, there is material financial value to the car not needing a Driver.

That is the value of FSD, however, in it's current state it has not yet come to fruition nor is there a clear path to that reality.

You're not disagreeing with me, you're restating the value proposition of FSD.
 
  • Like
Reactions: Mark II