You can install our site as a web app on your iOS device by utilizing the Add to Home Screen feature in Safari. Please see this thread for more details on this.
Note: This feature may not be available in some browsers.
AI is a very hard thing to wrap your head around--especially when we are all so conditioned to the concept of logic-based programming which has dominated computers since their inception.That's in line with most of the responses I've seen. I guess the answer to my question, and really any question about v12 (and later) is "more training."
I had trouble seeing how you could generalize what in my mind was 2 sets of traffic rules: normal ones, and override/emergency/crowd ones. I suppose with enough training and "neurons" to run the inference on the neural nets, it can all be one gestalt. And I suppose we will see, as we are reminded here from time to time: it ain't never been done before.
I am starting to believe 2024 will be the year of FSD breakout. The software architecture is now there, the compute power may well finally be there, and 12 months is a long time in computer training land. Train away!
I hereby volunteer to drive to every football game between now and then (why, for the Mission old chaps!)
I appreciate all the thoughtful responses!
I used to and now Elon gets brought up and I wont defend him. I do spend time every day responding to FUD articles on news aggregators. People ask me why when they are just trolls. I do it because without it people dont see whats real. Yet I refuse to defend Elon antics.You can vent your frustration with sarcasm here and then go out and help by informing people where you find the opportunity
haha. Oh wait?AI is a very hard thing to wrap your head around--especially when we are all so conditioned to the concept of logic-based programming which as dominated computers since their inception.
As long as we keep in mind that neural nets are literally brain simulators, then in theory, if your brain can do it, so can a neural net with enough neurons and enough training data.
The big question that I think you're essentially posing--and you're right, we don't really know yet since it's never been done before--is how much data is needed to cover all these cases? Technically, the car has eyes (cameras) and ears (an in-cabin microphone)--so in theory that's all the sensors needed to drive. It boils down to how much data we need, and how much training.
But what's promising to me is how naturally the car drove in v12. It looked like it drove dramatically better than what I've experienced with v11, and that was without the additional training time and compute resources that Tesla has since brought on. I don't know when we'll hit Robotaxi-level autonomy, but I agree that 2024 will likely be the year that FSD gets good enough that many people will be willing to pay $10k+ for the software...which will be a significant additional contributor to margins. Even without Level 5, a really good FSD system that people want to buy will contribute to the stock price considerably, since margins should go up quite a bit.
Interesting incentive update. But as a spoiled owner of a Model X with lifetime free charging, and not to be ungrateful, but 6 month of supercharging isn't really that great a value, unless:
Homepage, electric cars selected (for me) with several Teslas.
There's some advertising from Tesla. It would cost Tesla money to sell new cars there, just as it cost to sell used cars.Homepage, electric cars selected (for me) with several Teslas.
"New Cars" - straight to Tesla Model 3 and "Electric Vehicles" selected.
Also the same for me in a private window in a different browser.
Same for others?
Presumably this has cost Tesla money - or it's a condition of putting a lot of stock to the website.
View attachment 989611
There’s more to it than learning like a baby.I think we need to stop thinking about this as a problem that is handled with logic and code and merely answer based on how the meat sack in our skull does it.
How do WE identify the officer in charge?
Suppose I were to blindfold you and place you in the middle of a room in a building you've never been inside before. After taking the blindfold off, you would immediately realize that you were standing in the middle of a room in a building, even though you'd never seen that room before and even though the combination of colors and textures around you is unique and something you'd never seen before.
It sounds like v12 operates the same way. It learns much like a baby learns about the world around it--by watching and imitating. What we know about the human brain is that a vast set of neurons make connections with each other in a way that, as a collective, is able to understand and reason. Magically, the neural nets will do the same. I think the main remaining question is: how many neurons are needed to support the full range of tasks required to drive?
And it even had a new feature. It has learned to pull over and park on the street at its destination! I think it was probably not explicitly trained to do this but learned the behavior on its own.But what's promising to me is how naturally the car drove in v12. It looked like it drove dramatically better than what I've experienced with v11.
That is what I did a couple of year ago to leverage their SpaceX holdings You may be right that it's time to increase my holdings.It's a good time to get in on some BPTRX, which holds SpaceX and X Holdings (along with TSLA). Probably gets you exposure to Starlink, X, X.ai, etc.
I increased my holdings yesterday.
Agreed. Monitored autonomy as opposed to robotaxi will be Tesla’s value driver in the medium term. It’s all profit. It’s easy to scale when the tech is ready. Just offer free trials.AI is a very hard thing to wrap your head around--especially when we are all so conditioned to the concept of logic-based programming which has dominated computers since their inception.
As long as we keep in mind that neural nets are literally brain simulators, then in theory, if your brain can do it, so can a neural net with enough neurons and enough training data.
The big question that I think you're essentially posing--and you're right, we don't really know yet since it's never been done before--is how much data is needed to cover all these cases? Technically, the car has eyes (cameras) and ears (an in-cabin microphone)--so in theory that's all the sensors needed to drive. It boils down to how much data we need, and how much training.
But what's promising to me is how naturally the car drove in v12. It looked like it drove dramatically better than what I've experienced with v11, and that was without the additional training time and compute resources that Tesla has since brought on. I don't know when we'll hit Robotaxi-level autonomy, but I agree that 2024 will likely be the year that FSD gets good enough that many people will be willing to pay $10k+ for the software...which will be a significant additional contributor to margins. Even without Level 5, a really good FSD system that people want to buy will contribute to the stock price considerably, since margins should go up quite a bit.
Thanks for this.
”Tesla doesn’t advertise”, yeah right!Cybertruck at Annual Baron Investment Conference (New York) 8K Video
"Premiered 89 minutes ago
Recorded at Lincoln Center, New York. There was a Cybertruck, a Model S Plaid, a Model 3 Performance, a Model Y Performance, and a Model X Plaid. May take some time to process up to 8K, be sure to select it too."
Homepage, electric cars selected (for me) with several Teslas.
"New Cars" - straight to Tesla Model 3 and "Electric Vehicles" selected.
Also the same for me in a private window in a different browser.
Same for others?
Presumably this has cost Tesla money - or it's a condition of putting a lot of stock to the website.
View attachment 989611
There could be 3 reasons why Austin is not currently making Model Ys with 4680 at Austin.