Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Elon: FSD Beta tweets

This site may earn commission on affiliate links.
Speculation: the more I look into Tesla Insurance data being collected, it looks like Mr. Musk is planning on a pricing beta for the 3rd Party (Tesla Insurance) for using FSD. So that they can claim that the only insurance who can provide you better rates would be Tesla Insurance with all the bugs in FSD.
Not sure I understand. Can you explain further or provide example(s)?
 
It

it’s clear that they are manipulating this stat just like they are manipulating other stat and using airbag deployment as the only eligibility of an accident and there’s a thousand way you can crash without an airbag being deployed.

Before we get into the whole "are they manipulating this statistic" we first need see if there is anything about that tweet that is true at face value.

It wasn't until mid this year that Tesla doubled the beta testers to 2000. So where does the 2000 for almost a year even come from?
FSD Beta has never performed to the degree required for any of the beta testers to get comfortable with it. Sure we can debate about latency in take over events, but it simply isn't good enough for it to be anything than a "device under test" experience where you're way more vigilant than normal.

Now that threw out most of what Elon said it comes down to whether there has been an accident or not over that time period.
  • Is 1000 vehicles going without a reportable accident after a year of driving statistically all that unlikely to begin with?
  • FSD Beta has only included City Streets so that only includes a portion of driving.
  • What one person defines as an accident isn't what someone else would. Like I claim I've never had an at fault accident, but one time I barely clipped someone else's vehicle while trying to park while I was rushing (before I had a Tesla). I only remember the incident because I didn't have a pen, and I was desperately trying to find a pen to write a note while also needing to get myself to the train station. I've also been involved in "accidents" where someone gently rear ended me where we're both looking for damage trying to decide what to do about it.
  • We know there have been curbed rims so obviously these aren't being counted
  • We haven't seen any unexplained damage suddenly show up on any of the FSD Beta vehicles, but that's substantially less than 2K beta testers they supposedly have.
I haven't been able to answer the first question to my own satisfaction as a lot of the factors in an accident aren't present with FSD Beta testing. Like most testers aren't going to be testing it in a snow storm, and there are no teen drivers that I'm aware of.

What I do find amazing is there hasn't been any FSD Beta accident caused by overloading the driver with work, and multitasking. Not just that, but the idiots that wait far too long to take over. So I'd say its pretty extraordinary that there have been no accidents within the 100+ FSD Beta drivers that we know of.
 
Last edited:
... why Tesla Insurance only ?.
When Elon mentioned Tesla insurance calculator, I think he just meant to mean an algorithm for calculating how safe your are. In other words, has little or nothing to do with insurance and just a method of figuring out who are safe drivers and who is dangerous.

I myself feel I'm dangerous, but hopefully I can hide that from the calculator. My wife drives with either the accelerator or the brake on, so I doubt she will get a passing grade. I told her about the calculator and that she should NOT tail gate. She said sounds good, then proceeded to tailgate.
 
Last edited:
When Elon mentioned Tesla insurance calculator, I think he just meant to mean an algorithm for calculating how safe your are. In other words, has little or nothing to do with insurance and just a method of figuring out who are safe drivers and who is dangerous.

I myself feel I'm dangerous, but hopefully I can hide that from the calculator. My wife drives with either the accelerator or the brake on, so I doubt she will get a passing grade. I told her about the calculator and that she should NOT tail gate. She said sounds good, then proceeded to tailgate.

I wish there was a way Tesla could do this calculation without telling everyone explicitly. Whats the point of the calculation if people are going to cheat ?

ps : May be they should just ask for permission to look at the driving record over the last 6 months ... (or whatever history they still maintain).
 
Last edited:
  • Like
Reactions: Nvbob
I wish there was a way Tesla could do this calculation without telling everyone explicitly. Whats the point of the calculation if people are going to cheat ?

ps : May be they should just ask for permission to look at the driving record over the last 6 months ... (or whatever history they still maintain).
Well I still think that once you say its a calculator, you don't need to disclose the score a driver gets or to show if you are eligible or not. there by its a feature that everyone who has paid can request, but still only available to the "Elite's" I will wait to see if anyone with my configuration gets is (MC1/AP3).
 
  • Like
Reactions: Terminator857
What impact do you think it would have on the stock price if one of these FSD releases is actually mind blowing, meaning all of the testers, in week one, never had to take over? Where people are saying, OMG, this looks like production FSD, go read a book while your car drives you? And your answer may depend upon when this happens: a) this year, b) next year, c) 2023, etc.
 
What impact do you think it would have on the stock price if one of these FSD releases is actually mind blowing, meaning all of the testers, in week one, never had to take over? Where people are saying, OMG, this looks like production FSD, go read a book while your car drives you? And your answer may depend upon when this happens: a) this year, b) next year, c) 2023, etc.
2000 testers driving 300 miles each? That's sleep in the back seat full self-driving.
I feel like a pretty good chance of solving FSD is already (incorrectly) factored in to the stock price... Maybe it would triple?
It seems like it would depend on if patented technology were used to achieve it and how enforceable it is.
 
  • Like
Reactions: Microterf
2000 testers driving 300 miles each? That's sleep in the back seat full self-driving.
I feel like a pretty good chance of solving FSD is already (incorrectly) factored in to the stock price... Maybe it would triple?
It seems like it would depend on if patented technology were used to achieve it and how enforceable it is.
There is nothing to patent. There is difficulty and scale in implementation. Stock price would go bonkers as people would start paying for FSD again.

The one beef many FSD owners, very legit IMO, have with Tesla is that they sold something they just could not deliver. Excuses for city street this or that doesn't excuse that the best of AP today is not FSD, not close. Even the highway AP is buggy. Makes a drive much more enjoyable, yes- much more. Will average joes fork over 10k for that - no they have better sense and god bless them because that signal forced change. The longer the wait the less uptake they had. Credibility on this was becoming an issue.

But the good news is that Tesla has made a commitment to actually solving the FSD issue, another solution could have been refunding a substantial portion of the FSD price. So, we are seeing an expansion of the tester program to maybe 10k testers and then another several months and then maybe a big beta release of 50k or something and then....we'll see. Clearly the program has some velocity it lacked before. The use of youtube to show the public progress is brilliant. The uptake rate on FSD had fallen to virtually nothing and they clearly had a credibility issue. How to overcome that? Remove the curtains. Why is this good for Tesla stock holders is that it is putting public pressure on the development team. It also shows that Tesla believes they will solve FSD otherwise they would not take the risk. As an investor this is a great positive buy signal. What isn't getting videos is more interesting.

Dojo

The dojo program is frankly beyond just interesting and no, it is not priced into the stock. It is poorly understood. It may result in solving the generalized problem of AI. The stock doesn't even price in the potential of disrupting the entire electrical grid in North America and Europe. Replacing big oil and Utilities, that's a wowzer by itself and worth a good chunk of the stock price. The value approaching technological or synthetic singularity? Not even a whiff of that is priced into the stock. This is however, what is implied as possible. Solve FSD, really solve it and you have solved AI. Not some hack and whack a mole like Waymo.

Dojo buildout is going to be helpful in training and fine tuning but how long will that take ? Are we talking 1 year? Will they be able to actually release FSD before the dojo is finished? Will countries outside of NA have to wait for the dojo to be finished? Not sure on any of this. This is what's not in the videos and frankly this is what is interesting. All we get to see is one end result.

So no, FSD is not priced into the stock because the implication of solving FSD far exceeds the entire current valuation.
 
It

it’s clear that they are manipulating this stat just like they are manipulating other stat and using airbag deployment as the only eligibility of an accident and there’s a thousand way you can crash without an airbag being deployed.
Well they certainly aren’t mentioning the 50k manual interventions or reports of bad actions that happen with Beta. How many times did the system get intervention of the driving by the driver to prevent accidents?!?! We see this in every Beta video except a few specific drives some YouTuber fanboys post to lump up Elon. They leave out all the nonsense. But many do list the bad things and whether they are improving. Is that reported?
 
What impact do you think it would have on the stock price if one of these FSD releases is actually mind blowing, meaning all of the testers, in week one, never had to take over? Where people are saying, OMG, this looks like production FSD, go read a book while your car drives you? And your answer may depend upon when this happens: a) this year, b) next year, c) 2023, etc.
Those with some industry experience know this isn't happening soon. Maybe some very limited L3 by 2023. Jump in the back seat isn't happening by 2025. Lots of work required. Tesla has to get releases down to one day from two weeks, or be able to shutdown remotely when an unhandled case is known. Pot holes still aren't handled. Waymo finds most of their issues on new releases using simulator. How long before Tesla gets that good? Will Tesla eventually have the option to remote control cars? So much to do.
 
Last edited:
What impact do you think it would have on the stock price if one of these FSD releases is actually mind blowing, meaning all of the testers, in week one, never had to take over? Where people are saying, OMG, this looks like production FSD, go read a book while your car drives you? And your answer may depend upon when this happens: a) this year, b) next year, c) 2023, etc.
Unfortunately we’ll never find out, because the improvement will be gradual.
 
Dojo

The dojo program is frankly beyond just interesting and no, it is not priced into the stock. It is poorly understood. It may result in solving the generalized problem of AI. The stock doesn't even price in the potential of disrupting the entire electrical grid in North America and Europe. Replacing big oil and Utilities, that's a wowzer by itself and worth a good chunk of the stock price. The value approaching technological or synthetic singularity? Not even a whiff of that is priced into the stock. This is however, what is implied as possible. Solve FSD, really solve it and you have solved AI. Not some hack and whack a mole like Waymo.
Delusional... I forgot what it’s like to be in TSLA investor threads. Please don’t bring that nonesense here
 
Dojo

The dojo program is frankly beyond just interesting and no, it is not priced into the stock. It is poorly understood. It may result in solving the generalized problem of AI.
The contribution to solving generalized A.I. if it happens in 10+ years will be little by current dojo.

The stock doesn't even price in the potential of disrupting the entire electrical grid in North America and Europe.
Can you expand on what you mean? How is dojo going to affect the entire electrical grid? Are you talking about massive power draw?
Replacing big oil and Utilities, that's a wowzer by itself and worth a good chunk of the stock price.
Don't understand.
The value approaching technological or synthetic singularity? Not even a whiff of that is priced into the stock.
That's because there is zero chance. Can we put you into the dreamer category? Or where did you get this nonsense?

This is however, what is implied as possible. Solve FSD, really solve it and you have solved AI.
Lets look at it from a different angle. A.I. is not going to be solved so neither will level 5 FSD in the next 10+ years.

Not some hack and whack a mole like Waymo.
If only Tesla were as good as Waymo from a technical perspective.

Dojo buildout is going to be helpful in training and fine tuning but how long will that take ? Are we talking 1 year? Will they be able to actually release FSD before the dojo is finished? Will countries outside of NA have to wait for the dojo to be finished? Not sure on any of this.
Dojo isn't what you think it is. Compare it to Google TPUs and it falls short. If Tesla wanted dojo capability today, they can have it, just rent Google TPUs.
This is what's not in the videos and frankly this is what is interesting. All we get to see is one end result.

So no, FSD is not priced into the stock because the implication of solving FSD far exceeds the entire current valuation.
Unrealistic Tesla FSD dreams have been running rampent since 2015 and none of those dreams have been true.
 
Last edited:
  • Like
Reactions: Bladerskb
Solve FSD, really solve it and you have solved AI.

No. FSD is a subset of AI. Solving FSD would only solve a part of AI, not the whole thing. FSD AI is very specialized, it is just AI designed to solve the specific task of driving a car. It is not general AI. You can't use FSD AI to solve language AI problems or math AI problems.

Not some hack and whack a mole like Waymo.

The idea that Waymo is not really solving real FSD is another one of those myths that Tesla fanboys spread. Waymo uses machine learning on Google's supercomputers to develop deep neural networks for camera vision perception, lidar perception, radar perception, prediction and planning. According to Waymo's Head of Research, they have solved "vast majority of perception cases" and are now working on the long tail of edge cases. They are also working on problems like finding the best intermediary representation of perception to maximize efficiency of behavior predictions, prediction problems like anticipating intent of other road users both individually and as groups, best interfaces between perception/prediction, and prediction/planning, doing joint prediction/planning, building intelligent simulation agents to make simulations more realistic, to better train AV for interaction with other road users and much more. This is all according to Waymo's Head of Research, Anguelov. So, Waymo is solving the problems of real FSD.