Possibly, but more importantly (as it pertains to us), it may cost Tesla a lot of money.
3 Likes
It also has a lot of significance for the Tesla Network, if such ever comes to pass. The majority of existing Teslas are HW3, and all of the older existing Teslas are HW3. One of the massive advantages of the Tesla Network model was that it could be launched with used cars - which are, of course, much cheaper than equivalent new cars. If you can âflip a switchâ and create a Robotaxi fleet of a million cheap cars instantly, thatâs an enormous advantage. If âflip a switchâ becomes âbuild a bunch of mini-factories and do a material retrofit,â the advantage is significantly diminished.
1 Like
Sounds like a class action suit.
The Captain
Yes, there will be plenty of those. One class action suit was certified in California last year:
âŚbut there will almost certainly be many more. The frank admission that HW3 hardware will never be able to run a fully autonomous FSD suite will probably accelerate the filings.
2 Likes
There were some interesting comments about Optimus. Optimus is expected to be available ânear year.â The AI5 chip will be used in Optimus instead of robotaxi, because apparently the challenges of AI robots are greater than self-driving cars.
When asked when Optimus would go into full production, Musk sounded positively defeated as he explained the difficulties of opening an assembly line for a new product.
He also made an odd comment that Optimus 3 could not be revealed because competitors would steal the design. Iâm not sure what design features are unique to the outside of Optimus. Seems like the good stuff should be in the software and the internals. And should be patent protected anyway.
I think the reason we arenât seeing Optimus 3 videos is that Optimus 3 is not yet performing any human-replacing labor.
2 Likes
You see that with most of the humanoid robot videos, or at least the ones that go widespread enough to reach a larger audience. Theyâve gotten very adept at moving themselves around - very impressive videos of them walking, running, dancing, etc. They can carry things as well, though often somewhat gingerly and a bit slowly. But I havenât seen many examples of robots doing tasks that involve a lot of manipulation of objects to a sufficiently material way that I thought, âAh, that robot is close to being able to do a human job.â
There was a poster who used to put up videos of the latest/greatest humanoid robots from time to time, but not so much recently - so maybe I just havenât seen the advances.
As has been the case, the problem is the data. AI is well-suited for duplicating tasks that require data thatâs very visible to, and plentiful on, the internet. Language, software, images - great stuff for AI. The internet is filled with a gajillion files with words, sounds, images, and video to train on. However, real world manipulation of objects depends on responding to weight and mass and hardness and smoothness and friction and fragility and a host of other factors that donât have massive amounts of data online.
Next batter up:
frank admission that HW4 = AI4 will never run fully autonomous
3 Likes
On the earnings call Musk introduced what he called AI4 Plus. Which means they are already moving beyond AI4.
1 Like
Exactly.
They mentioned a time or two HW4 being unsupervised, but then, by the way, âwe are going to update HW4.â
Well, itâs new hardware now, if it has, say, double the memory.
How can anyone take this seriously?
Or maybe no one takes the products seriously as long as stock has âline go up.â
1 Like