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.
There is a good interview with Dave Lee on this point.


I agree with most of his points - in the short term i.e. about 5 years, GPT is not a competitor for Tesla Vision/FSD. After that, regarding the brains of the Optimus, GPT could well prove to be more suitable for it than what Tesla can create. However, Tesla will not stand still, and it might try to develop its own LLM for the Optimus brain. This will probably be a titanic battle down the road. There was a suggestion that Karpathy left Tesla because he saw the huge potential of GPT, while Tesla was focusing on vision (only-ish).

I suspect Tesla will, and probably already is, putting a lot of effort to create a GPT competitor.

I read what you wrote. It feels like a "jumping of the shark" scenario and way over my head.
 
I read what you wrote. It feels like a "jumping of the shark" scenario and way over my head.

Maybe you are right. It just seems we are on the cusp ( as in a few years) of AGI (without consciousness, that seems to be a different problem), and with AGI EVERYTHING we know changes radically, and probably in an exceedingly fast manner. If you thought we live in interesting times, I suspect we have seen nothing yet...
 
this is a pretty good sales driver headline…

Will this be followed by yet another price drop of $3,750 in Q2?
OR
Will they extend the rules into Q2?
OR
Will they get dream up an excess tire wear penalty, or multi-layer paint EPA disqualification for the next Titanium shortage? 🤣

And does G.J. dare pull out his Demand Warning Sign the day of P&D announcement?
All are possible, stay tuned!
 
Q - how could Tesla leverage GPT-4?

Further, what about Tesla Bot?

Note: I did a search on Twitter for "Electric Vehicles" and "LLM"...almost zilch.

The TeslaBot will absolutely use GPT like functionality. That actually was the missing piece. A general purpose bot is hard to work with unless you can talk to it in English and tell it/show it how to perform a function. Indeed when Tesla announced Tesla Bot, I thought to myself “too early” since I didn’t think AI was advanced enough. Obviously Elon knew better than me. AI researchers (and apparently Elon) knew 3 years ago that ChatGPT was just around the corner.
 
Will this be followed by yet another price drop of $3,750 in Q2?
OR
Will they extend the rules into Q2?
OR
Will they get dream up an excess tire wear penalty, or multi-layer paint EPA disqualification for the next Titanium shortage? 🤣

And does G.J. dare pull out his Demand Warning Sign the day of P&D announcement?
All are possible, stay tuned!
Probably doesn't need to do anything as q1 is worst for demand seasonality. So q2 on demand shouldn't drop that much. Or they introduce a new headlight and demand soars.
 
Maybe you are right. It just seems we are on the cusp ( as in a few years) of AGI (without consciousness, that seems to be a different problem), and with AGI EVERYTHING we know changes radically, and probably in an exceedingly fast manner. If you thought we live in interesting times, I suspect we have seen nothing yet...
Pet peeve of mine. People make statements about consciousness without knowing and defining what consciousness is.

My definition of consciousness is being able to remember and understand at some level the constant biographical stream of events/thinking that is happening to you. And then folding that short term memory of what has happened to you into long term storage.

ChatGPT does the former, but doesn’t have the ability to do the latter. It is like the guy in Memento whose short term memory gets erased every 5-10 minutes.

IMHO, ChatGPT is already conscious for the length of each of its sessions.
 
Pet peeve of mine. People make statements about consciousness without knowing and defining what consciousness is.

My definition of consciousness is being able to remember and understand at some level the constant biographical stream of events/thinking that is happening to you. And then folding that short term memory of what has happened to you into long term storage.

ChatGPT does the former, but doesn’t have the ability to do the latter. It is like the guy in Memento whose short term memory gets erased every 5-10 minutes.

IMHO, ChatGPT is already conscious for the length of each of its sessions.

The question on what is consciousness is really profound, and has vexed the brightest of minds since forever. Personally, and I could be completely wrong, I don't think ChatGPT is conscious or can ever be conscious. Even though it will for sure be able to mimic humans in every possible way, I don't believe it will have a feeling of I. That feeling of I, very oddly, seems completely irrelevant from an evolutionary point of view, as brain experiments have proved that our actions seem completely predictable as a function of inputs into various neurons. So we all could function in an identical manner as we are functioning, but without feeling I, without being conscious.

Perhaps then consciousness is an emergent phenomena of many neurons acting in parallel. In which case ChatGTP will become conscious. Personally, I don't think so. Here I side more with (physics Nobel winner-black holes) Roger Penrose, who believes one needs something extra in physics to form the feeling of consciousness. For him it is the interaction of gravity with coherent quantum systems in our brain - this may be completely wrong but seems the closest theory anybody came up so far which does not involve empty philosophising or religion or similiar.

In any case, we actually might find this out, by watching what happens to GPT. I actually do not know what is more dangerous - if it becomes conscious, or not.
 
Right. focus on first-principles and critical goals. For example, that's why Tesla just introduced Fart mode for auto inspections.
Oh, wait! My Bad.
That is a fair point. I wonder why they couldn't have assigned the programmers who worked on:

Fart mode
Microphone out the exterior speaker
Steam games
Etc.

...on a 360-degree view. If it is patented by Nissan or someone else, maybe that's why. You'd think Elon would have informed us of that in a tweet by now. Maybe he just doesn't want to do it.
 
  • Like
Reactions: bavarian_owner
There is a good interview with Dave Lee on this point.


I agree with most of his points - in the short term i.e. about 5 years, GPT is not a competitor for Tesla Vision/FSD. After that, regarding the brains of the Optimus, GPT could well prove to be more suitable for it than what Tesla can create. However, Tesla will not stand still, and it might try to develop its own LLM for the Optimus brain. This will probably be a titanic battle down the road. There was a suggestion that Karpathy left Tesla because he saw the huge potential of GPT, while Tesla was focusing on vision (only-ish).

I suspect Tesla will, and probably already is, putting a lot of effort to create a GPT competitor.
Ya made me think.... What about Optimus talking like it is ChatGPT? Sitting watching the game with Optimus.
EDIT: Ninja'd by Cosmacelf
 
  • Funny
Reactions: Artful Dodger
Even Elon himself stated if people don’t want it we can build a conventional truck.
A differently shaped and built Cybertruck would have a different combination of specs and price.,

The best feature of the CT is the ramp, and if I was doing any kind of work, everything possible is being wheeled down the ramp rather than lifted.

It is going to get progressively easier to build EVs in any shape and still achieve a great combination of specs, efficiency and price. But currently we are not quite there, form factor and weight and strength matter.
 
There is a good interview with Dave Lee on this point.


I agree with most of his points - in the short term i.e. about 5 years, GPT is not a competitor for Tesla Vision/FSD. After that, regarding the brains of the Optimus, GPT could well prove to be more suitable for it than what Tesla can create. However, Tesla will not stand still, and it might try to develop its own LLM for the Optimus brain. This will probably be a titanic battle down the road. There was a suggestion that Karpathy left Tesla because he saw the huge potential of GPT, while Tesla was focusing on vision (only-ish).

I suspect Tesla will, and probably already is, putting a lot of effort to create a GPT competitor.
I hope so. Elon will have to address the AI risk of wiping out humanity by hiring away the smartest AI developers and building the largest training supercomputer system in the world in-house. This is probably the one thing that keeps him up at night and in his eye the only way civilization has a remote chance to survive. I expect this for him to be similar to wanting to save free speech—but it’s a massively bigger problem.
 
I have no idea anymore than you what percentage of people predominantly, periodically, occasionally, or rarely put in or take things out from the side of their pickups. It’s an awkward maneuver at best and lazy at worst.

What I do know is that Tesla will sell plenty of CyberTrucks and will struggle to meet demand for at least the first 5 years of production - not counting year one of validation and slow ramping.

If you require something special in your pickup truck design that isn’t met by CyberTruck then move along and buy something else rather than pretending you know something you don’t.
I have had a full size pickup truck for a decade. i only side loaded twice (firewood) and rear loading would have been less optimal but would have also worked. Not really important for most people.
 
I have had a full size pickup truck for a decade. i only side loaded twice (firewood) and rear loading would have been less optimal but would have also worked. Not really important for most people.

I hate when people argue about this, it’s like nails on the chalkboard. Honestly feels like both sides are wrong.

When I had my Tundra, I used to side load/ unload a lot. Very likely more than the rear. It is super common when mountain biking to hop up in the bed and lift bikes out of the bed. Likewise the reverse, lifting them into the bed, then you climb in and tie them down. You would do it this way because otherwise it gets tangled with other bikes if you try to slide it out the back.

I’ve also used it quite a bit with loading and unloading tools, rakes and hoes are particularly prone to tangling and I’ve put a lot of them in/ out of my vehicles over the years. If my chainsaw is in the front of the bed, I’m pulling it over the side, not climbing over other tools that might be in the bed.

I think the squat function will mitigate this, as will the frunk. It just bugs the crap out of me when I hear people brush this off. Because a decent chunk of people do in fact load stuff over the side.

No clue if this will be an issue so I done feel it’s worth arguing about, particularly in light of all the trucks out there with massive sidewalls, I just feel like brushing it off is also wrong. Jury is out is entirely fair.
 
Last edited:
Continuing to look good for potential wide release of 11.3.3. A particularly good sign is that my December 2017 Model S just got 11.3.3 (updated from 10.69.25.1). Currently 467 installed and 663 pending on TeslaFi.
Downloading now on my 2018 Model 3. I don't think I've ever gotten an update that wasn't wide release.

Screenshot_20230326_232836_Tesla.jpg