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

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

This site may earn commission on affiliate links.
What Ford has said is they have L3 working "pretty regularly" on a prototype vehicle that is too expensive (thusfar) to actually deploy to consumers.

That's not at all hard to believe. Load up with expensive sensors and L3 that usually works is entirely doable without any magic.

Mercedes has a consumer vehicle with L3 for sale today though with a narrow ODD.

Waymo has had an L4 version, too expensive for consumer cars, for years now. As do a number of Chinese RT companies.


I vaguely recall someone even had a comment on the whole "expensive prototype" thing?


And Mercedes of course only has Drive Pilot on their S series, which is very expensive, and Drive Pilot is an additional $2500 on top of the cost of an S Class for a one-year subscription then pricing beyond that is unknown afaik.

Teslas approach seems most viable, with using only cameras and having people buy the vehicles up front then produce revenue for the company by using them as robotaxis, but the most viable option can still fail.
 
  • Like
Reactions: ShareLofty
I've not spent a lot of time reading RT valuations, but can any market be "ridiculously lucrative" as folks claim if there are competitors in the space?
There will be no RT competition. Seriously. No competition for a very long time.

This all assumes that Tesla's vision-only system with end-to-end neural nets actually works. But success seems more and more likely every day.

Tesla's system drives operating costs to rock bottom which leaves everyone else at an enormous price disadvantage. That's why nobody will be able to compete with Tesla for any market they decide to enter.

Nobody else has what Tesla has and nobody else is likely to have it in the next several years.

Very few people understand this. If you are one who does, it's time to take advantage now.
 
Now I understand why some people hate Elon. It's not enough for him to destroy free speech, now his other company is hellbent in destroying the International Space Station?

/s
Now I understand why I hate some News Media portals, their 'Headlines' misrepresent the true story. Space X gets the contract to tow 'end of life' Space Station back to Earth and sink it in the Pacific.
 
Rivian's Engineering Lead and Senior Manager, Abdullah Zaidi, has also presented on the benefits of adding LiDAR and the importance of better localization sensors for Level 3 systems. This was older info, in Nov, 2022, that's not long ago folks.

This is from Rivian today:

"We developed the Rivian Autonomy Platform in-house to enable optimal performance of our system and increased scalability for future feature updates. Powered by entirely new perception hardware and software, the system fuses inputs from an array of 11 internally developed cameras and five radars performing over 250 trillion operations per second, an industry leading level of compute power. This multimodal sensor approach excels at perceiving different types of information, offering redundancy for improved detection."

I assume "Radar" is just for general object recognition and not LieDar. Maybe they figured it would be handy for the odd tree in the way, but watch out for the flying squirrels! I don't need to elaborate on the risks of sensor fusion. But 5 Radar units are an added cost so it must be important for something right? Makes me ask, what is wrong with their current vision system?

I'm just trying to figure out their autonomy strategy, especially now since VW is onboard. IMHO, without demonstrated autonomy in just a few years, they're all toast. Just my opinion.
 
People slam things like LiDAR and sensor fusion as if Tesla has proven the alternative is at all viable for actual robotaxis — they haven’t, but they’ve proven it’s ok for a Level 2 eyes/hands-on ADAS where a human behind the wheel owns liability.

I seriously hope August 8th doesn’t reveal a more robust Robotaxi-specific sensor suite if only to save people from playing mental Twister trying to contort themselves around such a development.
 
over 250 trillion operations per second, an industry leading level of compute power.
Oh... there's some Marketing Hype. Are they adding up the all computers, how many we talking about? (Also cute - not the "leader" just "leading level").

1/4 Trillion Ops/sec, lol! The NVidia H200 is at 1.83GHz (Boost clock speed, likely requiring multiple clocks per Ops).
 
  • Like
Reactions: wtlloyd
People slam things like LiDAR and sensor fusion as if Tesla has proven the alternative is at all viable for actual robotaxis — they haven’t, but they’ve proven it’s ok for a Level 2 eyes/hands-on ADAS where a human behind the wheel owns liability.

I seriously hope August 8th doesn’t reveal a more robust Robotaxi-specific sensor suite if only to save people from playing mental Twister trying to contort themselves around such a development.
Yup, we do.
You should sell before 8/8.
Not advice.

I'm going off the most Recent Data. They've been in my town for nearly a decade, but not nearly as long as this pole from the 1900's.
Fusion breaks down.

1719509267985.png


OTOH, humans have different sensors. The difference is that we learned using all of them together AND they don't overlap (except for ear balance + vision maybe, but then look up Vertigo for the human equivalent to bad sensor Fusion, or reading in a car... or the Boeing Max for the stretch case.)

Instead, we puke, spinout, or fall over stupid. That's sensor fusion conflict resolution gone bad. But for some sick reason, knowing which way is up was critical to survival, can still run while blind at least.
 
Have you forgotten all the times a Tesla has been spotted with Lidar?

It doesn't mean the company will use lidar on the fleet.
Tesla has taught us to dismiss LIDAR completely for FSD. While true with Tesla's FSD, this is really processing of ANY kind of inputs and provide an output.

If some entity chooses to do LIDAR and echolocation that's fine. Need to process all and spit out FSD.

Tesla (Elon) had chosen not LIDAR due to its costs. A LOT have gone into software and compute instead. All points out to Tesla being fully able to resolved FSD, I'd not dismiss LIDAR or any other input parameters completely.

Somebody has to be 2nd.
 
  • Like
Reactions: Mike Ambler
The problem with Ark is not that they're overly optimistic on their valuations model, but that they're overly optimistic on their targets as well. If you look at their model for 2024 done in 2020, it's miles off.
Is it OK to add that ARK are a bunch of kids in their 20s with absolutely no experience in anything?
 
Yup, we do.
You should sell before 8/8.
Not advice.

I'm going off the most Recent Data. They've been in my town for nearly a decade, but not nearly as long as this pole from the 1900's.
Fusion breaks down.

View attachment 1060305

OTOH, humans have different sensors. The difference is that we learned using all of them together AND they don't overlap (except for ear balance + vision maybe, but then look up Vertigo for the human equivalent to bad sensor Fusion, or reading in a car... or the Boeing Max for the stretch case.)

Instead, we puke, spinout, or fall over stupid. That's sensor fusion conflict resolution gone bad. But for some sick reason, knowing which way is up was critical to survival, can still run while blind at least.
This was not a sensor failure, the Waymo detected the pole but had assigned a low damage score to the object and proceeded into it because the pole was within the driveable space and not separated by a hard curb or something along those lines.

Here’s the recall report:

 
I honestly don’t even know what type of real analysis Ark does on their holdings and with their price targets in terms of looking at real-world limitations or realities. Reading through their stuff, it kinda sounds like they just put some data together and sandwich it between two pieces of bread before throwing it in the deep fryer — a Monte Cristo analysis I think they call it.

It’s like they don’t even pick real targets, they’ll have a blurb saying their base / bull / bear cases are just representative of the output from 2million possibilities coming out of the fryer. And they collect fees for this.
ARK is nothing more than a bunch of college kids doing reports for school. I don't know why they get any coverage or investor dollars. I guess they have Tesla to thank for that.
 
People slam things like LiDAR and sensor fusion as if Tesla has proven the alternative is at all viable for actual robotaxis — they haven’t, but they’ve proven it’s ok for a Level 2 eyes/hands-on ADAS where a human behind the wheel owns liability.

I seriously hope August 8th doesn’t reveal a more robust Robotaxi-specific sensor suite if only to save people from playing mental Twister trying to contort themselves around such a development.
It has absolutely been proven that LiDAR is not needed to drive a taxi. Humans drive just fine without it.

The only thing yet to be proven is if you can build a neural net that makes driving decisions as well or better than a human.

Nobody but Tesla is positioned to build that neural net. That's what Elon is betting on. That's what I am betting on.

August 8th will show a vision-only system. There is 0% chance of anything else.
 
ARK is nothing more than a bunch of college kids doing reports for school. I don't know why they get any coverage or investor dollars. I guess they have Tesla to thank for that.
Incorrect. They might feel like it to you because you're "getting on a bit." But they are not a bunch of "college kids". https://www.ark-invest.com/our-team/

And just because one becomes older, it doesn't make them wiser. Certainly not when it comes to up and coming technologies and industries where historic analysis is less relevant.
 
It has absolutely been proven that LiDAR is not needed to drive a taxi. Humans drive just fine without it.

The only thing yet to be proven is if you can build a neural net that makes driving decisions as well or better than a human.

Nobody but Tesla is positioned to build that neural net. That's what Elon is betting on. That's what I am betting on.

August 8th will show a vision-only system. There is 0% chance of anything else.
Oohh even doubling down on Vision-only, I like it but would be a bit cautious after Tesla had introduced a brand new HD Radar in HW4 but never used it.
 
“The longer the base, the bigger the breakout”.”

So easy to just pull random extreme numbers out of thin air, or to make crazy assumptions to justify your targets. Maybe she should be quiet for a while until she’s actually right on something, rather than being one of the biggest wealth destroyers of the last 3 years who still somehow has a megaphone to yap nonsense.
Incorrect. They might feel like it to you because you're "getting on a bit." But they are not a bunch of "college kids". https://www.ark-invest.com/our-team/

And just because one becomes older, it doesn't make them wiser. Certainly not when it comes to up and coming technologies and industries where historic analysis is less relevant.
 
Oohh even doubling down on Vision-only, I like it but would be a bit cautious after Tesla had introduced a brand new HD Radar in HW4 but never used it.
That's right. Vision-only. Consider me doubled-down.

There is no way Tesla will integrate the HD radar any time soon. It would be extremely counter-productive and delay the development of FSD.
 
This was not a sensor failure, the Waymo detected the pole but had assigned a low damage score to the object and proceeded into it because the pole was within the driveable space and not separated by a hard curb or something along those lines.

Here’s the recall report:

Thanks, I hadn't read it. However, I think this is the only truth we can be assured of:

"Description of the Safety Risk : Insufficient ability to avoid pole or pole-like permanent objects within the drivable surface may result in an increased risk of a collision."

Their Defect explanation is the answer a software PM would say - quite Boolean in word choices. My follow up to that deflection would be to ask the obvious: "Why would you hit a pole even if you could see it?" The real answer is again, wait for it... Sensor Con-Fusion (but only if we also consider the maps as sensor data input).

Facts:
1. Lidar saw the pole (no physical reason it would miss it).
2. Cameras saw the pole (no reason they would not).
3. Drivable space was beyond the pole. (Vision or Maps incorrectly concluded it's a useable road surface).
4. The pole did not move.

So what happened at #3? Actually, I think it's worse in that the maps won over 2 types of vision systems combined. I don't think this was sensor Fusion in the usual sense. But if we look at Map data as sensor input, that's where the "fusion" broke down (meaning that various sources of info are not aligning so a decision is needed).

My next question... What if I were to place a 5M diameter steel ball in the road. Write code to detect a steel ball? What if a brown man stood still? Maps cannot rule the day, it's crazy what happened.

Recalling AIDriver showed us when FSD found a better route than what the maps said. (His Tesla turned left at the light vs the map saying a right then a U-turn. So his Nav didn't make sense, but the car figured it out.)

OK, enough said, sorry for dragging this down an FSD rathole but it's fairly critical to understand for obvious investment reasons. I seriously don't know if any of my input is helping. But for me, this is THE hinge factor. I don't just base this off Elon's religious Lidar comments, I understand motion sensors more than most and how the compass can lie to the gyro's at anytime. Even that sensor fusion is complex but SpaceX clearly figured it out.

So sensor fusion is necessary to some degree of overlap, but there's got to be only one Ego, no schizophrenia.
 
It has absolutely been proven that LiDAR is not needed to drive a taxi. Humans drive just fine without it.

The only thing yet to be proven is if you can build a neural net that makes driving decisions as well or better than a human.

Nobody but Tesla is positioned to build that neural net. That's what Elon is betting on. That's what I am betting on.

August 8th will show a vision-only system. There is 0% chance of anything else.
I almost clicked agree. But 0%, even I won't go that far.
What if he slaps a monkey brain with a Neuralink on the hood just play that steer the car game? 🤷‍♂️

Hey, here's an old corporate work joke... How do you make a monkey shake! 🙈

One scoop of monkey brains and two scoops of ice cream in a blender!
(IDK, it was funny in the 80's anyway).
 
  • Funny
Reactions: CarlS
So basically they will argue the plaintiff legal fees on July 8th and then discuss the recent vote end of July or early Aug.

Makes lots of sense to argue legal fees before a final decision on pay package/s.

This looks to be setting up to a pay the lawyers a huge fee and then rule the pay package is now valid with the recent vote.