Elon reveals Cybertruck details / production expectations on Q3 2023 call

[PilotPete]

My take is that FSD is likely to be solved near the end of this decade. Let's say 2030.

I think 2030 will be way too long. I think we have L3 by the end of 2024, L5 within 24 months after that. Why? You pointed it out, Tesla has 5M cars out there, all providing video for FSD AI to learn. And that number continues to grow. EM said each car CAN provide over 19 TB of video/data every day. So that is;
19,000,000,000,000 bytes of data times 5,000,000 cars daily. Let's put that into perspective.
That's 95 with 18 zeros behind it, every day. If you had that many seconds to live, that would be over 3 TRILLION YEARS. If you were that far away from a CT IN INCHES and could travel at the speed of light (excluding the effect on time, just a distance example), it would still take you over 255 YEARS to get to it.

And with the 20,000+ GPU systems they are running, FSD is coming far faster than we can imagine. I've seen just a few flights fed into a GPU for learning, and the capabilities are amazing.

You want to talk about an advantage on development? No one else is in the same league, much less the same ballpark. And they keep adding cars every day. Each one is adding to that data stream.

Sponsored

 

[cvalue13]

So think about this:

Competition for Tesla EV never came, yet you think competition for FSD will come hard and fast? Even though solving FSD is 10x more difficult than creating an EV?

I strongly believe FSD will be solved, no later than by 2030. Elon said 2020 and I added 10 years. I can somewhat see your point that there will be other players who will solve autonomy eventually, but they won't be able to scale like Tesla. And that is the advantage that Tesla has.

Feels like (but not clear) that you're thinking here of auto companies.

The companies that will disrupt FSD are not auto companies.

They are technology/software companies that already exist, are far larger than Tesla (value and global reach), and far more advanced in this space than people appreciate.

Success is never guaranteed (for them or Tesla), but there are some 800lb gorillas in the dark corners of this autopilot/FSD room that most folks have yet to notice** and will soon have some 'ah ha' moments.




**my business is in this space of edge compute / AI, and so I'm privy to at least a couple of these 800lb gorillas pipelines

 

[Arctic_White]

Feels like (but not clear) that you're thinking here of auto companies.

The companies that will disrupt FSD are not auto companies.

They are technology/software companies that already exist, are far larger than Tesla (value and global reach), and far more advanced in this space than people appreciate.

Success is never guaranteed (for them or Tesla), but there are some 800lb gorillas in the dark corners of this autopilot/FSD room that most folks have yet to notice** and will soon have some 'ah ha' moments.




**my business is in this space of edge compute / AI, and so I'm privy to at least a couple of these 800lb gorillas pipelines

Interesting. Nothing is ever guaranteed then. That is why I hedge.

Who else has the same amount of relevant real-world data? PM me if it's too sensitive to discuss.

 

[Arctic_White]

I think 2030 will be way too long. I think we have L3 by the end of 2024, L5 within 24 months after that. Why? You pointed it out, Tesla has 5M cars out there, all providing video for FSD AI to learn. And that number continues to grow. EM said each car CAN provide over 19 TB of video/data every day. So that is;
19,000,000,000,000 bytes of data times 5,000,000 cars daily. Let's put that into perspective.
That's 95 with 18 zeros behind it, every day. If you had that many seconds to live, that would be over 3 TRILLION YEARS. If you were that far away from a CT IN INCHES and could travel at the speed of light (excluding the effect on time, just a distance example), it would still take you over 255 YEARS to get to it.

And with the 20,000+ GPU systems they are running, FSD is coming far faster than we can imagine. I've seen just a few flights fed into a GPU for learning, and the capabilities are amazing.

You want to talk about an advantage on development? No one else is in the same league, much less the same ballpark. And they keep adding cars every day. Each one is adding to that data stream.

I say 2030 just to be on the safe side. I wouldn't be surprised if it's solved by the end of 2024, nor would I be shocked if it takes them until 2030.

There is data and there is relevant data. I think sorting through good data requires an insane amount of computing power, which I believe to be the bottleneck for now. Let's wait until Dojo is onboard. Tesla is also working on creating one of the world's fastest supercomputers. That will certainly help solve this problem sooner.

It's exciting times!

 

[PilotPete]

I say 2030 just to be on the safe side. I wouldn't be surprised if it's solved by the end of 2024, nor would I be shocked if it takes them until 2030.

There is data and there is relevant data. I think sorting through good data requires an insane amount of computing power, which I believe to be the bottleneck for now. Let's wait until Dojo is onboard. Tesla is also working on creating one of the world's fastest supercomputers. That will certainly help solve this problem sooner.

It's exciting times!

Tesla just added 10,000 H100 chips to the farm earlier this year. And the new OS makes the A100 and H100 way more efficient. Dojos are coming on line. That’s why I say end of 24. Tesla has 4,000 A100s sorting through the data to determine relevance, leaving 6,000 A100s to work with the H100 and new Dojos

When FSD12.X drops, we will have a far better crystal ball to work with

 

[Arctic_White]

Tesla just added 10,000 H100 chips to the farm earlier this year. And the new OS makes the A100 and H100 way more efficient. Dojos are coming on line. That’s why I say end of 24. Tesla has 4,000 A100s sorting through the data to determine relevance, leaving 6,000 A100s to work with the H100 and new Dojos

When FSD12.X drops, we will have a far better crystal ball to work with

Wow... if it does happen by the end of 2024, that'd be amazing!

 

[Crissa]

When NASA outsourced, they forgot how to go to the moon.
It's taking an 'insource' company to go to Mars. (Spagettiengish ;ol)

NASA has always outsourced. They didn't make the Saturn rockets, they just flew them. It was always political.

There's nothing wrong with outsourcing - Tesla buys robots and vans and tools and metals from other companies. It's just about focusing on the right things.

-Crissa

 

[Crissa]

Related, in 2016 when Tesla was in need of help in making assembly lines better, they could have contracted with a company known for such things. Instead they bought one of the best in the world, Grohmann.
https://electrek.co/2017/05/10/tesla-acquisition-grohmann-engineering/

FF four years, and the fact that Tesla required Grohmann to cut ties with former customers caused issues for Mercedes, and maybe others who were not as candid about it.

https://insideevs.com/news/395392/tesla-grohmann-acquisition-taking-toll-mercedes/

This insinuation that Tesla force Grohmann to sever ties has never been confirmed, to my knowledge.

Grohmann was susceptible to buyout because the other automakers were cancelling orders with the company.

-Crissa

 

[Bill837]

This insinuation that Tesla force Grohmann to sever ties has never been confirmed, to my knowledge.

Grohmann was susceptible to buyout because the other automakers were cancelling orders with the company.

-Crissa

Confirmed, no, but I think it's more likely than not.

https://m.slashdot.org/story/325521

 

[charliemagpie]

NASA has always outsourced. They didn't make the Saturn rockets, they just flew them. It was always political.

There's nothing wrong with outsourcing - Tesla buys robots and vans and tools and metals from other companies. It's just about focusing on the right things.

-Crissa

You are right, nothing wrong with outsourcing, many good reasons for it.

 

[Crissa]

Those negotiations are like when you buy something on Black Friday only to find out they jacked the price a month before to look like you got a good deal. Really you still got ripped off.

There's a law specifically against that in California.

I know, we have too many laws.

...But it's because some people can't play fair.

Confirmed, no, but I think it's more likely than not.

https://m.slashdot.org/story/325521

You know me. I don't accept 'likely'. I need things to fit in their fact or not cubbies.

-Crissa

 

[Knucklehead]

I know, we have too many laws.

Yep!

 

[Cybertrip.world]

Some questions and details I'm curious about:

HVAC to bed for camping via rear window or rear vent

Payload capacity (3500 as stated in 2019?)

Clearance for sharp turns with a 5th wheel style trailer over the bed side-walls

Ground clearance measured at lowest and highest

Rear steering turning radius. Example backing up a trailer.

Steer by wire ?

Front approach attack and rear departure angles for off-roading

Lights:
Lightbar at top of windshield under the glass like 2019 reveal


Towing:
Can a hitch ball be bolted directly into a hole in the rear bumper for lighter loads without needing a 2” receiver-style hitch?

Trailer light wiring plug or add after-market

Winch mounting area

Frunk and bed electric outlets

Tire size

Curb rash and off-road rock risk (Is tire wider than rim or plastic wheel cover?)

Side mirrors removable (easy access panel to unbolt it?)

Is there space for a small display screen just inside the left and right A-pillar (inside cabin) so side-camera can display after external mirrors are removed? Will these screens be available after-market

Door opening buttons or fingerprint reader or code? (maybe just a release button like on the interior) I assume Bluetooth proximity activates open button.

Emergency door handle access from outside for rescue?

Air compressor auto-deflate and inflate tires

Tailgate:
Measurements (with a tape measure) of bed width and length

Close-up of mounting points for rooftop cargo rack

Is the top edge of the tailgate replaceable when dented or scratched from a load?

Water drainage from bed

Tailgate integrated steps or ramp

Can tailgate drop down at a low angle, not just stop flat with bed

L-channel close-up and example using it with tie-downs

Tailgate motorized or just soft dampened lowering but manual raising

Accessories

Wiper blade sizes to change (two or three standard blade replacements)

 

[Crissa]

dream of owning AN In-N-Out? Maybe 10.

If they are a California McD's franchisee, they just dream their stores were not located in California anymore. AB 1228

They already paid more than that. They were some of the highest paying fast-food chain.

If they said this, they're lying.

-Crissa

 

[Crissa]

I know, we have too many laws.

Yep!

...But it's because some people can't play fair.

So we gotta have 'em.

-Crissa

Sponsored