Let me voice an opinion most will disagree with:
Tesla should give priority to developing Tesla Bot over solving FSD.
Most people take an incrementalist view with would posit that Tesla should keep plugging away at solving FSD, then after that it will have the bandwidth to go after the more ambitious project of training Tesla Bot. This assumes that the framework necessary for solving FSD is somehow foundational to higher AI, as if you can just build one on top of the other. Maybe.
A more radical view is that we don't really know what is foundational to humanoid AI, but we can reasonably assume that whatever grand framework can truly make Tesla Bot AI successful will be sufficient to solve vehicle autonomy. Indeed a fully trainable Tesla Bot can trained to be a valet for parking lots. If so, driving highways is not that much of a reach for further training. What Tesla is currently doing with FSD will look like machine learning by comparison. FSD will look like child's play.
So what are the big intuitive leaps that Tesla Bot needs to be a successful? I don't know, but the sooner Tesla discovers this grand framework and builds up all the supporting tech, the sooner FSD will become a trivial exercise.
Moreover, every useful task that Tesla can train the Bot to do has market potential, and much of that potential can be brought to market with minimal regulatory impedance. And much of it is much lower risk from just about any vantage point. The Tesla Bot initiative will yeild a massive stream of marketable innovation along the way to Tesla Bot Singularity.
I believe that Tesla Bot can leapfrog FSD. Sometimes you just need to focus on going to Mars. When you do that, making Earth more sustainable or the Moon more inhabitable happens along the way.
I like your thinking: Having spent almost 2 decades trying to boot-strap and mass-produces EVs, Tesla is just going to say, "Na, we are not into cars anymore - we pivoted to bots".
Bold move!
To recap: There are - at least - 4 possible outcomes of FSD for Tesla.
One is never. Few here believes that, I would recon.
Off course, we don't know
for sure that FSD can be achieved, but it is becomming increasingly likely that is really is a matter of time and effort.
Another is "late as in 10-15 years. And further, due to this long horizon, Tesla won't be necessarily be first, or not by much, or not best, because it will take such a long time that many will catch up. So late could mean that Tesla won't make a killing from FSD or robotaxis, but will have a more modest revenue. And bots are just another level which is beyond difficult.
(I don't think this scenario is realistic, but ... I cannot ignore it entirely. Tesla could become the victim of a a lot of really bad luck - a black swan event which would derail their lead, macro black swans - things like that.)
Yet another is "soon". Soon often means near-future or 6 month to 2 years. It is tempting, but I don't quite believe this.
The last option is "sort of close".
Could be 1 year, could be 5 years. A lot of uncertainty.
The problem is, for each new victory, each new special-case that FSD suddenly is able to pull off, other hard cases arise. Not because they were hidden, but usually because we wouldn't have thought it remotely possible before that FSD could handle such a complex case. And we discover that it has to handle these things in order to really drive well.
There is a sub-case of "sort-of-close" in which self-driving is doable, but really, *really* hard. ("Close-but-hard-cubed")
Hard as in landing and re-using rockets. Hard as in only team Elon can do it - and it still takes half a decade. Good thing: No competition because it is just at the very edge of human abillity to pull off.
That scenario lend itself to very peacefull co-existence re. ressource allocation wrt bots versus FSD. Because the best path toward FSD merges with the best path towards bots.
My guess is that team Elon has concluded that the above scenario is the most likely. And that solving FSD is so very hard that you almost have to built a
digital clone of a primate in order to succeed. (I have avoided saying digital clone of a human on purpose)
And when you reach that, the bot-body is just another 'sleeve' as they say in Altered Carbon.
Dump the car-sleeve, don the bot-sleeve . Download bi-pedal locomotion sub-routines.
Good to go (pun intended)
TLDR:
You have a good point, and I kind of agree the bot is the most important thing.
But given the "close-but-hard-cubed" scenario, which I find the most likely, the paths merge between FSD and bot.
And doing it that way allows Tesla to operate in sort of de-facto stealth mode re. the bot untill they are ready - and continue on plan with making cars and lots of money.
Very few people are taking the bot seriously right now.
But they will.
www.autoblog.com
