Search results

I remember when he presented it. What a crazy insane feat for a car company to have done all it took to get to the point where they could abstracted it so the customer(the FSD team) only had to change one line of code...

So I was wrong about FSD revenue recognition due to sFSD no longer being beta. But seems that Autopark released some revenue: I am gonna guess ASS will recognize some more revenue in Q2.

$25k car early 2025 or late this year! CyberCab!

130% QoQ compute growth. If we extrapolate it, that's 28x/year. Well above the 10x/year that seems to be standard nowadays. But it came a pretty hefty price. But the reason Tesla decided to do that was because they thought the decision to do it had a positive EV. Given the uncertainty the...

This was a nice interview. Here another leading AI robotics company is tell what strategy they are using. You can interpret this how you want, but my interpretation is that they are validating Tesla's approach: If you are an investor you want Tesla to give away as little information as...

The main advantage of Tesla having a lot more compute is that their engineers can iterate faster and that they can increase the dataset. Meta who has tons of compute has released their newer GPT version and performance has improved a lot compared to last year: Just look at the score of MATH...

We are close to the bubble. Tesla has pushed allin, but their stack covers the others player so they will not be eliminated even if their hand doesn't hold.

They have a list of milestones when they will have certain % in deferred revenue. Going from beta to no longer beta likely happened as they thought that they had accomplished some goals which likely correspond to the goals for recognising deferred revenue.

Doing the layoff in Q2 will add costs to Q2 for severance packages and restructuring. I think deliveries > production in Q2 as they had lots of cars in transit in Q1 which will add a benefit. I have a feeling they will recognize a lot deferred FSD revenue in Q2 as sFSD is no longer Beta. In all...

4680 is NMC for the high capacity batteries that they need for Semi, Cybertruck and future products such as Optimus, eVTOL etc. CATL are doing LFP and Sodium for stationary storage and mass market models such as 3/Y/25k etc. Tesla has realized that they need more LFP made in the US due to IRA...

I wonder if Elon's Auto in his profile is related to AUTO in Wall-E: Maybe Tesla are building Auto?

Learning from Tele Operation works:

Haven't seen any news about the attack on Tesla Fremont:

Auto- (prefix): "Auto-" is a prefix derived from Greek, meaning "self" or "same." It is used in various compound words to indicate self-operation or self-action, such as "autonomous" (self-governing) or "autopilot" (a device for steering an aircraft or ship).

Drew started as a junior at Tesla back in 2006. It was a very high risk startup and most people only do these startups for a year or two. He has been massively incentivised to stay with stock options that has since then gone to the moon. His salary likely was in the millions and his stock...

One nice thing with the trial month. People will not buy any other brand:

FSD choosing the wrong lane:

Milei greeting Elon:

I have. But I do recognize this kind of discussion, it doesn't end well... So let's end it before mods get annoyed.

1. It's true that we currently need more data approximate what humans can do with artificial neural networks 2. It's also true that in many domains we can exceed what humans can do by using lots of data 3. Tesla has lots of data FSD currently has driven >1B miles with 0 deaths. This...

If your neural network can solve it, expect an artifical neural network to solve it. You don't solve it using magic...

1. you input the GPS position into the neural network, train it on a diverse dataset and the neural network learns what regulatory rules are active at different GPS positions 2. you preprocess the GPS position to figure out which jurisdiction you are in, for example "california = 1 ... china 623...

He talks about end2end object detection, not end2end control. His argument is that you will have objects right in space but wrong in lane and that this is problem. Didn't Tesla solve this by having a single neural network doing all the predictions. If not can you just solve it using language of...

Just chance unit from mile to meter and you can add three nines. Imo it's just a figure of speech, probably based on six sigma. Six Sigma is a set of methodologies and tools used to improve business processes by reducing defects and errors, minimizing variation, and increasing quality and...

It's easy for us to pretend to be CEOs of billion dollar companies, but let's face it, we would likely make a lot worse decisions. I think it's better to try to understand why they make the decisions and what it implies rather than trying to figure out which decision had the highest EV. Elon...

So basically supervised FSD has saved many lives and prevented many injuries. And it's been getting better.

Good luck getting all the photon counts from all the cameras to be identical over a few hundred frames from two different time points of a vehicle moving. Even getting the car to move two identical paths is hard enough even without worrying about the light. It's called probabilistic robotics...

Here are some common beliefs associated with the Year of the Dragon: Good Fortune: People born in the Year of the Dragon are believed to be blessed with good luck and prosperity throughout their lives. Strength and Power: The dragon symbolizes strength, power, and courage. Those born in the...

Robotaxi and 25k car will use the same software stack for FSD. Rototaxi is more about making the car comfortable as a passenger, making it easy to clean, making it very durable etc. Basically a car optimized for being in use a lot by passengers and not a fun driver. It might include things such...

Imo Elon wanted unboxed car to be 100% robotaxi because he thinks it's best to go allin on robotaxi. He likes to bet the house because by doing that the company has a higher chance of succeeding at doing the task. Franz et al were unsure if robotaxi software would be ready. Franz convinced Elon...

Back when I developed self driving cars, our solution was to crash into the humans furthest distance away no matter the number of them, if they were old/young, criminals/priests etc. But I guess the Tesla solution to the problem is to do whatever humans do. And if regulators have a different...

When it's working correcly it's not running redundant, that's true. But we don't really how they handle/will handle cases where 1 neural processor fails. I see a few different options: 1. Have a 50% smaller neural network ready to be loaded for these situations, run of the previous output for 1...

I'm not worried about inventory or delta production vs deliveries. I am more worried about inventory minus "in transit to customer". This quarter due to a few external factors in transit to costumer went up a lot, I assume that next quarter it will go down. If it does for no good reason I might...

Don't know how it does currently, but in the end it will be have slightly better than human vision and superhuman attention. Lidar is not much better with fog, at least the non-military grade lidars. Radars do better with fog for object detection, but good enough on sign reading etc to make a...

It's really hard to make a simulator include all edge cases. Doing that is likely harder than solving FSD. A better way is to find rare edge cases IRL then make the simulator create variants of these edge cases, for example same situation but different weather, add traffic, change velocities...

I have been semi involved in crypto 12 years. So many times have I statements like: "I don't understand anything about Bitcoin, have never used it and have only thought about it for a few minutes, but here why it can never work and why all you guys are wrong". Seen a lot of same mentality about...

Timeline for Elon compensation package legal proceedings. Cliffs: late 2025...

Karpathy on how Elon runs companies differently.

Not saying this is false, but can you list some other car manufacturers and how much they are spending on AI compute? My hunch is that they spend a lot less than Tesla on this, but I could be wrong.

Seems like it was a large amount of compute coming online from various sources. Imo this is what Tesla are doing well, being agile. When everyone struggles to get nVidia chips, they have already diversified to alternate suppliers, doing it inhouse and getting things up and running fast. And they...

Can you tell that this is an AI?

Less than half can be a lot less than half. At scale HW5+, cameras, battery pack etc should be pretty low costs. Motors are the expensive part but I am sure they are working hard with their greek team to lower the costs. But I don't think cost is a huge problem at this stage, $1k more or less...

If Tesla has managed extreme improvements from V12.3.1 to V12.3.2 I wonder how they did it. We know that they are no longer compute constrained from Elons tweets. But how would this have worked? Suddenly 10k H100 came online at once? Doesn't seem plausible. And it's not only that the improvement...

Mods are deleting my posts, guess they want this discussion to end so I will not reply here anymore. Mod: This is correct. I also deleted your "last words" because they just continue the argument that I've asked you two to stop, now for the third time. I (as mod) don't care about the correct...