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

[scaesare]

I think that doc is a good watch and can inform thinking but it's healthy to keep in mind they are not very comparable.

AlphaGo and the subsequent software is an AI that plays a game and has a very clear identifier for declaring an improvement, it wins.

AlphaGo can play itself, or its predecessor in a virtual environment to improve.

FSD is - at least as I understand it - two things. An AI developed to interpret visual input - Tesla Vision-, And hard coded rules for driving.

Tesla Vision can't really play against itself as I understand. I guess its possible that they could feed two models through a video clip and then whichever model had a higher degree of confidence about what it saw wins. But to validate it it needs to roll out to a test fleet and have users say "Hey this is way better, but it tried to drive through the concrete pillar under the monorail".

That process is much slower than a computer playing itself 24/7.

AI Day could prove me very wrong though.

My understanding is that, while there's no doubt that the "rules of the road" are specified in the system, the diving policy/decision layer is not "hard coded" in that there is a procedural code layer implementing these decisions... at least not any more.

At Autonomy day, Karpathy talked at some length about moving the policy/decision making process in to a neural net... and the "Software 2.0" model. As such. there's a NN for vision, and one or more NN's for the decision making process(es). And those decision-making NN's would take in to account the specific rules of the road.

I think we see evidence of this in the FSD videos where the car will cross the dividing line when no oncoming traffic is present to allow pedestrians more space, or similar.

 

[Carl Raymond]

Yesterday I rewatched the documentary AlphaGo. I had watched it a couple years ago but didn’t fully appreciate all the information that was provided and how it relates so well to FSD.

The Wikipedia page is also very informative and shows all the versions, timeline and performance.

The key take away is that as the time between version reduces dramatically, the performance improves dramatically.
It should be similar for FSD.

View attachment 697937

I’m not sure comparisons to Go are applicable.

Here are the rules to Go:
Black goes first. Take turns placing one stone. Surround territory.

Now write down the rules to driving…. Does anybody even know them? Your set will be a little different to my Aust. set, and my NSW set will differ slightly from a Queensland or Victorian set.

When turning at a lights intersection that has a bike lane, should you wait IN the bike lane, blocking it, or leave room for bikes with the risk that a skinny car sneaks up the inside? No responses to this question please, OT for this thread, merely an example of a situation where the rules will vary, and knowledge of the rules will be poor.

Agree that versions from learnings are happening fast now. I just hope they are not learning fast about a new unsolvable situation - a damned if they do, damned if they don’t type problem.

 

[StealthP3D]

I watched the video pretty carefully, and while he does not understand basic things about the car (like biohazard mode) , I did not think he was too alarmist about the prospect of the performance being too much. Besides that 'does not require skill' line, he is correct, Tesla puts extreme performance at a much lower price. I would say other than the paint stuff, he was positive through most of the video. No question in my mind, his video will net overall sell the Plaid. His reaction to the speed alone will sell these cars.

His statement is non-sensical: "It doesn't require skill anymore to afford a car this fast."

Your ability to afford a fast car has zero relevance to how much driving skill you have.

 

[Beltsbear]

His statement is non-sensical: "It doesn't require skill anymore to afford a car this fast."

Your ability to afford a fast car has zero relevance to how much driving skill you have.

And I agreed with you by saying he was NOT correct about that line.

 

[geometro]

Cramer says "I would buy Tesla now"

Still find Cramer to be a clown.

 

[StealthP3D]

Most cars this powerful has pretty strict pre-requisites...and that's the ability to shell out a few million and actually get one before the production of like 50 cars run out. Doug is right, this is the first car anyone can have access to with a little conviction when it comes to the leasing agreement. We have already see just model S running into trees from showing off. I don't see it as FUD and he is not the only one who said this car should be illegal (believe dragtime said the same thing, and a few of his buddies also said it too).

It was a silly statement - just admit it. "It doesn't require skill anymore to afford a car this fast."

Money does not equal skill.

 

[Skipdd]

Cramer says "I would buy Tesla now"

He saw the SALE sign..

 

[StealthP3D]

I can’t remember exactly, but was the FUD this bad just before the 2020 sky rocket we had right around June through Jan 2021?

I would say there was not a moment in 2019 that the FUD wasn't more prevalent and at least as thick. The primary difference now is a LOT of people have figured out it was all made up. There's still a long ways to go but fewer and fewer people buy into it.

 

[Knightshade]

At Autonomy day, Karpathy talked at some length about moving the policy/decision making process in to a neural net... and the "Software 2.0" model.

He's been talking about that long before AD, but it's always been aspirational, not what they're actually doing today.

My understanding is that, while there's no doubt that the "rules of the road" are specified in the system, the diving policy/decision layer is not "hard coded" in that there is a procedural code layer implementing these decisions... at least not any more.

At least as late as the V8.x FSD beta (and in all wide-release code up through today) they were using traditional code for all of that.

ONLY perception was NNs (though it used quite a few of them, not just one, to do that job)


In Elons recent tweet about 9.2 it seems they're finally, starting in 9.2 (and only for the FSDBeta folks), using NNs for something other than JUST perception- specifically:

V1 of the Multi-modal prediction for where other vehicles expected to drive. This is only partially consumed for now.

So they're running non-perception NN for one thing, prediction in one specific line, it's version 1 and only in the city streets beta, and they're not really trusting it's output yet (which makes sense given it's V1)... that's a pretty long way away from using NNs for actual driving policy.

 

[Artful Dodger]

Here's the problem with posting screen shots of text instead of the text itself: TMC "Search" can't find it. Blue Origin's embarrassment is forever hidden/lost.

Your 11 cry/laff emoji's however live forever in infamy. Just nobody will ever know why.

 

[22522]

It was a silly statement - just admit it. "It doesn't require skill anymore to afford a car this fast."

Money does not equal skill.

Money can buy driving school.

 

[Singuy]

It was a silly statement - just admit it. "It doesn't require skill anymore to afford a car this fast."

Money does not equal skill.

The statement doesn't really make sense if you take it at face value..but the meaning is clearer if you substitute the word "skill" for "effort"...because getting the fastest production in the world used to take a lot of effort when there are only 20 or 50 made..all will appreciate instantly when you have it..sometimes double in price. Without proper connection and ass kissing its nearly impossible even when you have the cash.

I guess when you saw the word "skill" you thought driving skill...while I thought the skills required to navigate the complex road in getting a very limited edition car that fast before the plaid.

 

[Big Time]

Futures up. See y’all tmrw

 

[engle]

My video below is in jest of course, but the fact that we are seeing premium gas ~$4.50/gal here in Californication and the price is high to a lesser extent across the USA is an awesome demand driver for Models 3 & Y!! [*] As an investor, I hope the price only keeps going higher and never returns to what it was before. I would love to see gasoline prices on par with those in Europe. It will be good for our planet and great for my net worth! The timing could not be better with Austin Terafactory Model Y entering production "real soon now" in Elon time. Hopefully, ramping Y in Austin and Berlin will be much easier since they already did it in Fremont and Shanghai. However, I understand they will be casting the entire underbody? And plan to transition to the new cells when they have enough of them available down the proverbial road...

Tesla Model 3 Owner in Beast Rental Spends $100 for California Gas

[*] The other major demand driver is all the tesla-on-tesla crashes in Fremont.

 

[Favguy]

It was a silly statement - just admit it. "It doesn't require skill anymore to afford a car this fast."

Money does not equal skill.

Whilst I agree his statement makes no sense, I think the point he was trying to make is valid. This being that this new level of performance previously reserved for a very few of the uber wealthy, is now available to, if not the average Joe, certainly the doing alright for him/herself Joe. Without a doubt this is going to lead to an increase in deaths as more idiot drivers find ways to do 0-60 in 2 seconds into immovable objects.

Part of me thinks it's irresponsible to just let anyone who wants to buy one, drive off in a Plaid. I'd like to see a mandatory high performance driving course at a racetrack be part of the deal, where you do this first to get a feel for and respect for the sheer power and potential danger of the thing. If you don't display the ability to handle the car properly, you don't get to buy it. (Pretty sure I'm going to cop some flak for that one...)

 

[WarpedOne]

Part of me thinks it's irresponsible to just let anyone who wants to buy one, drive off in a Plaid. I'd like to see a mandatory high performance driving course at a racetrack be part of the deal, where you do this first to get a feel for and respect for the sheer power and potential danger of the thing. If you don't display the ability to handle the car properly, you don't get to buy it. (Pretty sure I'm going to cop some flak for that one...)

Yes, you will, because you are pretending Plaid is first such car that has such a "problem".
You are having a knee-jerk reaction to its abilities.

Every single new car sold today is capable of braking the law and killing you if you so desire.
It is onto you to use it responsibly and abide the law.

The car offers "Chill" mode if you do not trust yourself enough.

 

[Buckminster]

Message to newbies. If you aren't already reading it, @Papafox has a daily update here:
Papafox's Daily TSLA Trading Charts
If you only read one post a day - make it this one. No replies in thread please.

 

[Artful Dodger]

Yes, you will, because you are pretending Plaid is first such car that has such a "problem".
You are having a knee-jerk reaction to its abilities.

Every single new car sold today is capable of braking the law and killing you if you so desire.
It is onto you to use it responsibly and abide the law.

The car offers "Chill" mode if you do not trust yourself enough.

Yup. Any bubba with $25K can buy a Bowtie 572 Inch crate motor with 700 horse, and stuff it into their 1968 Nova with 4-wheel drum brakes... :/

This isn't new. Demuro is scare-mongering for clicks. That's also not new.

 

[petit_bateau]

My thoughts on AI day:

1. Affect on stock price: nil to slightly positive. I hope. Don't think anyone has hyped it this time, and price obviously hasn't risen ahead of time due to expectations of some big move in the stock.

2. I suspect the AI Tesla is investing beyond their vehicle fleet relates to factory automation--the machine that builds the machine. What is Tesla's mission statement? Accelerate the advent of sustainable energy. Best way to do that? Get the price on their cars and other products down as low as possible. How do you do that? Focus on the machine that builds the machine.

So I suspect Tesla is working on in-factory AI. Machines that transport raw material and other goods around the factory. Machines that visually inspect panel gaps and approve/reject vehicles based on the measurements. Any way they can increase the production rate or reduce production cost.

Not explicitly stated in the Tesla mission is "do stuff that is necessary for Mars", but I don't think it is an accident that everything Tesla is working on also happens to be necessary for Mars.

So reducing the number of humans required to build & operate a factory or mine or etc is definitely on the roadmap. So too will be the ability to put the minimum set of these factories / 'farms' / etc on Mars. So expect to see Tesla put together a small chip 'fab' at some point - not necessarily to do bleeding edge 7nm stuff, but sufficient to churn out more basic chips at relatively low volume. Ditto for application of general physical-navigation navigation to task-robots for agriculture. All these things are hugely enabled by the sort of capabilities that are going into the FSD puzzle box, which needs to be able to move around in reall space, and move around humans.

Somehow I think it is going to be quite a while before Tesla runs out of industries to disrupt.

 

[stef]

His statement is non-sensical: "It doesn't require skill anymore to afford a car this fast."

Your ability to afford a fast car has zero relevance to how much driving skill you have.

He obviously means skills to make enough money … like having a YouTube channel to review cars while making dumb faces…