Cybertruck Q1 2025 Deliveries (Lowest to Date) [⚠️ ADMIN WARNING: NO POLITICS]

[SCTesla]

So, what does Grok become when it interfaces with FSD?

Also, FSD absolutely is an AGI. It just doesn't talk to you right now because they have specifically excluded that ability. Which is why I brought up Grok.

I don't think you know what AGI is, but Elon doesn't think it's AGI. Even though he called it mini or baby-AGI, once. He also said on JRE that once a company achieves AGI, it will change the world. He said he think Tesla could be first. He said it could come as soon as 2026, but he was iffy on that timeline.

AGI is defined as the hypothetical intelligence of a machine that possesses the ability to understand or learn any intellectual task that a human being can.

Basically AGI is sentient. It can make corrections, write it's own code, figure out problems, and exponentially expand on those. FSD nor Grok, and certainly not both combined are doing that and we don't have a timeline of them doing that. AI talking does not equal AGI. THat's an insane leap. Alexa, SIri, Gemini, Meta's Llama have been speaking for years. Grok speaks now. That's an irrelevant aspect. AGI needs to be able to reason, know that it's a computer, that it exist. Some don't even believe AGI is possible...which is why some bastardize the term to really advanced usage of AI.

No one in AI considers Tesla AGI. It was mocked when Elon called it baby AGI. Karpathy said he understands why Elon said that even if it's not actually AGI.

Sponsored

 

[Sjohnson20]

I think we are going to see some great CT incentives in Q2. Hang tight if you are contemplating a CT purchase.

They will need to move the 2024 trucks. Problem is who wants them? EVs are all about the latest version. If I was ordering a Cybertruck now I would want a 2025. I guess with enough discounting it will happen.

 

[REM]

No one in AI considers Tesla AGI. It was mocked when Elon called it baby AGI. Karpathy said he understands why Elon said that even if it's not actually AGI.

I figured we would already be on the same page about the nuance of definitions here, but I fully understand that a real AGI doesn't yet exist. And it's precisely the reason why Elon called FSD a "baby agi". Because it is.

A more technically correct name could have been "narrow-band AGI". But calling it "baby" makes more sense to the layperson trying to conceptualize what it is. I like narrow-band because it gives a better idea of how they are building FSD to be a full AGI, but for right now they are purposefully placing narrowing limits on how it "thinks" and "acts".

FSD already meets several defining chrematistics of an AGI. It thinks, acts, and communicates. It's just not communicating with you in English. Yet. (what do you think Grok is for?) (what will we call FSD when Grok is introduced?)

No one is claiming to have fully solved AGI yet; but Tesla is absolutely the closest. They are shooting for a goal that makes the following a reality:

User (talking to their Tesla phone app): "Hey Grok, can you warm up the Cybertruck. I'm going to take the family to the Steakhouse down the road. Don't remember the name; I think it's about 3 or 4 miles away."

Grok: "Sure, warming "The Beast" up now. There are 2 restaurants that fit the criteria you are looking for: Outback Steakhouse, located on Broadway, and Roadhouse, located on Peachtree St.

Which one?

User: The Roadhouse. thanks.

Grok: Got it. Ready when you are.

User: *Loads family into the Cybertruck and presses "Start Navigation"*.

Grock: It should take us 15 minutes to get there due to traffic. I'm taking a different route than normal since there seems to be a wreck on the main route. Do you guys want to listen to some music along the way?

When Grok is integrated in the vehicles like this (a feat likely 4 or less years away, by my pure guess), what would you call this system? FSD with added bells and whistles? Or the world's first AGI?

Or even more wild, what happens when someone can say: "Hey Optimus, can you drive us to Roadhouse here in the next few minutes"

 

[YDR37]

They will need to move the 2024 trucks. Problem is who wants them? EVs are all about the latest version. If I was ordering a Cybertruck now I would want a 2025. I guess with enough discounting it will happen.

Telsa is currently offering Free Supercharging on the Model S, Model X, and 2024 Foundation Series Cybertrucks. Would not be surprising if 2024 Non-Foundation Cybertrucks were added to that list.

 

[SCTesla]

I figured we would already be on the same page about the nuance of definitions here, but I fully understand that a real AGI doesn't yet exist. And it's precisely the reason why Elon called FSD a "baby agi". Because it is.

A more technically correct name could have been "narrow-band AGI". But calling it "baby" makes more sense to the layperson trying to conceptualize what it is. I like narrow-band because it gives a better idea of how they are building FSD to be a full AGI, but for right now they are purposefully placing narrowing limits on how it "thinks" and "acts".

FSD already meets several defining chrematistics of an AGI. It thinks, acts, and communicates. It's just not communicating with you in English. Yet. (what do you think Grok is for?) (what will we call FSD when Grok is introduced?)

No one is claiming to have fully solved AGI yet; but Tesla is absolutely the closest. They are shooting for a goal that makes the following a reality:



When Grok is integrated in the vehicles like this (a feat likely 4 or less years away, by my pure guess), what would you call this system? FSD with added bells and whistles? Or the world's first AGI?

Or even more wild, what happens when someone can say: "Hey Optimus, can you drive us to Roadhouse here in the next few minutes"

It's not baby AGI, it doesn't think. It processes curated data and uses that data to write code. No one else is doing it like they are with video and it's advanced, but it's not thinking. When Tesla feeds the car bad data, the car writes bad code, which we've seen multiple times with V13. It's doing that because it can't think, it's a computer program that takes in video and spits out code that dictates driving behavior. Bad in, bad out...because FSD isn't capable of understanding "that's not correct" or "that will break other things we've been implementing".

FSD with Grok is just like FSD with Siri or Google...it's a voice assistance that works with the car. We can do this to a degree with Tesla voice controls, cars in China with voice commands are way beyond this where you can say, "Take me somewhere to eat I've never been that is close to my house"....that's still not in the realm of AGI, just 2 different AI stacks working together. Grok in itself does nothing to further your argument of AGI, which it still sounds like you don't quite understand.

 

[Cybertruck2024]

The only way the CT isn't a sales bust is if the RWD ends up doing big numbers. The issue is the Lightning is already as cheap as any full size ICE pickup truck and it doesn't move, I am not sure CT will either and it will be $10k+ more expensive than the Lightning. I'll be happy if I'm wrong, the CT is my favorite vehicle I ever owned (before the heckling), but I agree with others, there doesn't seem to be market for it.

 

[REM]

Grok in itself does nothing to further your argument of AGI, which it still sounds like you don't quite understand.

lmao, ok pal. I guess you are under the impression that Tesla has a billion "if then else" statements plastered together that makes up FSD? Do you understand that Tesla is rapidly opening up several NN stacks that absolutely lead to it making human-like desicitions on the fly, in the real world? FSD isn't geo-fenced, nor does it have many hard-coded steps like most other autonomous driving systems do. In fact, it's exactly what Waymo does. And it's why they are going to ultimately fail during wide adoption phases.

FSD 12 didn't slow down appropriately at most speed bumps. FSD 13 does. Why do you think that is? You think some engineer somewhere added better code?

Nope. Dojo gave the FSD computer better heuristics to go:

"oh, that's a speed bump ahead!
I'm currently going to fast for that.
Let me slow down before we get to it, then speed back up once both axles traverse it."

^ those types of actions aren't being hard coded. FSD is LEARNING IT.

Elon hinted at this paradigm here:
“The really wild thing about the end-to-end training is that it learns to read. It can read signs, but we never taught it to read. We never taught it what a car was, or a person, or a cyclist. It learned what those things are, what the objects are on the road just from watching videos. It finds order in these things. Just like humans.”

 

[WHIZZARD OF OZ]

Hate to say it but if sales of the CT in the US do not increase significantly I don't think the CT will ever see the international market, importers will ship them at a very high cost to the affluent but the odds of average Aussies, Kiwis, and Brits getting their hands on one will not happen with low US sales in the land of the truck I fear.

Gigahorse, l hear 'ya!
Maybe your words ring true or we together are TROJANS ( GIGAHORSE+WHIZZARD= TROJAN HORSE ) and can get the CT through the GATES OF CYBERHEAVEN DownUnder or Down The Streets Of London and Cross The Ditch...... Let's hope CYBERLIFE don't be a D!TCH
G'Day from the Land of The Boomerang+Kangaroo ?
Guess l'll HOP to it......HOPPING 4 CYBERTRUCK

 

[SCTesla]

lmao, ok pal. I guess you are under the impression that Tesla has a billion "if then else" statements plastered together that makes up FSD? Do you understand that Tesla is rapidly opening up several NN stacks that absolutely lead to it making human-like desicitions on the fly, in the real world? FSD isn't geo-fenced, nor does it have many hard-coded steps like most other autonomous driving systems do. In fact, it's exactly what Waymo does. And it's why they are going to ultimately fail during wide adoption phases.

FSD 12 didn't slow down appropriately at most speed bumps. FSD 13 does. Why do you think that is? You think some engineer somewhere added better code?

Nope. Dojo gave the FSD computer better heuristics to go:

"oh, that's a speed bump ahead!
I'm currently going to fast for that.
Let me slow down before we get to it, then speed back up once both axles traverse it."

^ those types of actions aren't being hard coded. FSD is LEARNING IT.

1. Dojo only handles 10% of FSD training.

2. FSD is being trained by video in and code out. It's programming by the video, but it cannot think. "Learning" can be an adequate description as the more video it gets the stronger the confidence in the model of driving. It's the same way with video analytics in general. People on my team (in my previous job) would upload millions of images and videos of spills and video before spills occur be able to identify when a spill is about to happen and when it did happen. Video in, code out. If I loaded some bad video of a perfectly stable cup that never spilled, it would hurt the end analytic. That's why we get regressions with what Elon calls bad video or too much of one kind of video like we saw with 13.1. Both of these scenarios are because this is not AGI...it cannot think. It cannot tell Elon or the Company "This video actually breaks stop lights (video of people trying to game the green light) or speed controls". It's not capable of that.

 

[tingmo13]

I don’t need the AI for this fight, but let’s speed this along shall we:

“Does Tesla have the most advanced FSD software currently available in China if not who does?” - Me to ChatGPT 4.5

—-

4.5 ??

Tesla’s Full Self-Driving (FSD) software has faced regulatory challenges in China, leading to a rebranding of the system to “Intelligent Assisted Driving” and the suspension of planned free trials pending necessary approvals.

In contrast, Chinese electric vehicle manufacturers have made significant advancements in autonomous driving technology:

• BYD: Introduced the “God’s Eye” system, offering advanced driver-assistance features across its vehicle lineup at no additional cost. This strategy aims to democratize self-driving technology, making it as standard as seat belts.
  
  
• Huawei: Developed the Qiankun ADS 3.0 system, which has been praised for its precision in navigating complex driving scenarios, including tight parking spaces and dense traffic.
  
  
• XPeng: Offers advanced autonomous driving features in its vehicles, positioning itself as a strong competitor in the self-driving technology sector.

Given these developments, Tesla’s FSD software does not currently hold the position of the most advanced autonomous driving system in China. Companies like BYD, Huawei, and XPeng have introduced systems that are not only technologically competitive but also more readily available to consumers.

 

[tingmo13]

Not try to ignore the tragedy overall here ...

FYI, A news station (link) reported that driver (non-owner) was driving with hands off the wheel which is illegal there.
[Aside: I can't tell you the number of times Tesla FSD gave me no notice and flashed red before turning off FSD. I always drive with my hands on the wheel, supervising. They didn't make logical sense either. Some were sun shining thru trees. Some I had NO idea why. I also recall the 3 USA college students from a CT crash several months back that was also a tragedy.]

From another article:
https://carnewschina.com/2025/04/01...mi-su7-claims-three-lives-on-chinese-highway/

• 10:27:17 PM: NOA activated, vehicle speed at 116 km/h
• 10:28:17 PM: Mild distraction warning issued
• 10:36:48 PM: NOA issued hands-off warning: “Please hold the steering wheel”
• 10:44:24 PM: NOA issued risk warning: “Please note obstacle ahead,” with deceleration request
• 10:44:25 PM: Driver took over from NOA, entering manual driving mode, steering wheel turned 22.0625 degrees left, brake pedal pressed 31%
• 10:44:26 PM: Steering wheel turned 1.0625 degrees right, brake pedal pressed 38%
• Between 10:44:26–28: Vehicle collided with concrete barrier
• 10:44:28 PM: Vehicle eCall triggered
• 10:44:39 PM: eCall connected, accident confirmed, police and emergency services contacted
• 10:45:06 PM: Contact established with vehicle owner, confirming non-owner was driving
• 10:47:15 PM: Emergency medical services successfully dispatched
• Approximately 11:00 PM: Emergency services arrived at the scene

 

[REM]

1. Dojo only handles 10% of FSD training.

2. FSD is being trained by video in and code out. It's programming by the video, but it cannot think. "Learning" can be an adequate description as the more video it gets the stronger the confidence in the model of driving. It's the same way with video analytics in general. People on my team (in my previous job) would upload millions of images and videos of spills and video before spills occur be able to identify when a spill is about to happen and when it did happen. Video in, code out. If I loaded some bad video of a perfectly stable cup that never spilled, it would hurt the end analytic. That's why we get regressions with what Elon calls bad video or too much of one kind of video like we saw with 13.1. Both of these scenarios are because this is not AGI...it cannot think. It cannot tell Elon or the Company "This video actually breaks stop lights (video of people trying to game the green light) or speed controls". It's not capable of that.

Bro, there are 2 core stacks that FSD is training with. One of them takes real visuals and associated metadata from the fleet and compiles into several processing lines. The other is a 3D simulation that is constantly running in order to whittle down edge cases.

Your car does that, but at its own scale, and with probability modifiers (handed down by Dojo) to provide guardrails and specific actions for edge cases. How large of a memory footprint do you think each car has? Right. Not that much.

The "FSD" visual you are seeing on the MCU is an actual render (minus most of the tags and telemetry) of what the FSD computer is making decisions on. It's not running from a pre-deterministic stack of code. It used to do that for the most part, but not anymore. It's mostly a series of branch neural nets.

In other words, the car is making decisions. Like humans.

I don't know what you are trying to prove here. Tesla isn't running If - Then - Else statements in the car ?

 

[dalton108]

Bro, there are 2 core stacks that FSD is training with. One of them takes real visuals and associated metadata from the fleet and compiles into several processing lines. The other is a 3D simulation that is constantly running in order to whittle down edge cases.

Your car does that, but at its own scale, and with probability modifiers (handed down by Dojo) to provide guardrails and specific actions for edge cases. How large of a memory footprint do you think each car has? Right. Not that much.

The "FSD" visual you are seeing on the MCU is an actual render (minus most of the tags and telemetry) of what the FSD computer is making decisions on. It's not running from a pre-deterministic stack of code. It used to do that for the most part, but not anymore. It's mostly a series of branch neural nets.

In other words, the car is making decisions. Like humans.

I don't know what you are trying to prove here. Tesla isn't running If - Then - Else statements in the car ?

My car runs on BASIC. OG/CB old code, old horn for LIFE!

 

[SCTesla]

Bro, there are 2 core stacks that FSD is training with. One of them takes real visuals and associated metadata from the fleet and compiles into several processing lines. The other is a 3D simulation that is constantly running in order to whittle down edge cases.

Your car does that, but at its own scale, and with probability modifiers (handed down by Dojo) to provide guardrails and specific actions for edge cases. How large of a memory footprint do you think each car has? Right. Not that much.

The "FSD" visual you are seeing on the MCU is an actual render (minus most of the tags and telemetry) of what the FSD computer is making decisions on. It's not running from a pre-deterministic stack of code. It used to do that for the most part, but not anymore. It's mostly a series of branch neural nets.

In other words, the car is making decisions. Like humans.

I don't know what you are trying to prove here. Tesla isn't running If - Then - Else statements in the car ?

There's a delta from if this than that to AGI. Tesla FSD is not AGI. Neither are the analytics that I produced with spills, intent, or aggression detection. Just because we/they use neural network doesn't mean it's AGI or able to compared bad data or decisions. It still only knows what it's programmed to do. AGI, IMO, is years away. Tesla could be the first, who knows? But FSD isn't there it's not mini or baby or almost AGI and certainly not actual AGI which you originally claimed. Not even Elon would agree with that.

Again, DOJO is a nothingberger in FSD compared to the Nvidia clusters. It's basically a failed expirament. The Nvidia chips turned out to be easier to code and easier to procure en masses and Tesla shifted accordingly.

 

[scottf200]

I think the range is very good. I know Elon's pitch was more, but the range is better than a Y in real world. Quite the achievement for a pickup

However, that is due to the battery pack size of ~75 kWh to ~123 kWh (1.6+ times bigger)

Sponsored