-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
2daMoon
Mostly Harmless
Twenty or so years ago I had conversations with folks about recycling and talked about how I wanted to watch for companies who developed the tech for mining landfills for the resources stored there.
That time is getting closer all the time. What was considered wasteful and unconscionable to bury trash will become a mining operation for pre-processed raw materials. The Tesla Bot could be instrumental in this endeavor.
Good review! One thing that might spoil the show though is 4680 cell supply not ramping up fast enough (IIRC Tesla said not needed until end of 2022 but afterwards). There could also be some additional ramp up delays when introducing structural pack in Berlin. Shanghai had the advantage that they didn´t have to ramp two technologies in parallel.
I feel like the general population is going to not believe it until it’s well under way since we’ve been lied to about recycling for decades in its current form. But redwood proves that there’s real value to recycling stuff we haven’t even tried to recycle before.
2daMoon
Mostly Harmless
One aspect of the presentation that sets Tesla Bot apart from so many others was its ability to recognize an object (watering jug), grasp it, and then go find a planter to water with it. Subtle, but freaking amazing. Especially when the visualization was shown from the bot's perspective (like an FSD view for the car).
Another, the pre-planning for actuators purposefully in order to reduce the number of them to six in order to streamline production. Brilliant! Particularly, doing so on the 2nd prototype. Having teams of designers who are cognizant of including this in their plan from the get-go is so very much what makes Tesla different from those companies whose R&D folks are kept isolated from their production teams.
Looking forward to seeing what they accomplish with the Catgirl version...
Another, the pre-planning for actuators purposefully in order to reduce the number of them to six in order to streamline production. Brilliant! Particularly, doing so on the 2nd prototype. Having teams of designers who are cognizant of including this in their plan from the get-go is so very much what makes Tesla different from those companies whose R&D folks are kept isolated from their production teams.
Looking forward to seeing what they accomplish with the Catgirl version...
I thought it interesting that Elon was explicitly backed up by one of the engineers/leads on stage when he made his predictions. It was certainly different than the usual narrative of Elon making pie-in-the-sky predictions while everyone working for him groaned in the background.
This is the result of uber bulls who spew nonsense and also know nothing about the subject. Elon of course doesn't help when he take a dig at other companies when his own tech is not yet finished. These behaviors flame the hate.
You may have misunderstood the purpose of Optimus: Doing dangerous, boring, or repetitive jobs.
So, not a sex bot
Skin is not so important for factory work or house cleaning.
Knightshade
Well-Known Member
He was pretty clear it was pushed back to actually be able to demo the robot
And given how they said the presentation was the very first time it had walked untethered it seems they legit needed 100% of the extra time to make it happen... because after the CT window demo I know I sure would've tested it even an hour before the show if there'd actually been a spare hour to do it.
I don't know that the presentation would have moved the stock significantly either way even absent P&D coming likely tomorrow.... the folks who already knew Tesla was working on amazing stuff beyond cars aren't surprised and if they're slightly more or less impressed than expected won't change their investing significantly either way, and the folks who insist it's all accounting fraud and fake product announcements will be telling you how this is another example of that all day today and into the future.... already seeing tons of that from tech press (Ars Technica for example) and all the heck over Twitter.
So, not a sex bot
He literally said there'd be a cat girl model.
Last edited:
The Accountant
Active Member
For AI Day 2, I was not looking for a working Optimus. I was looking for Ambition, Commitment and Talent and I was very happy with what I saw.
Especially the talent. Many young engineers on a world stage and they impressed me.
The ingredients are there for success.
Especially the talent. Many young engineers on a world stage and they impressed me.
The ingredients are there for success.
They killed it. This is a giant flashing sign for all the brilliant, young people out there. Here you will matter.
I wonder how many of those 6 Optimus actuators will appear in Cybertruck in some form. Might as well ramp up production.
Since this is the investor thread, it should be pointed out that Optimus is currently immaterial to the valuation of TSLA, and will remain immaterial for at least three more years, and likely many more than that.
If Optimus was a separate startup it would be a very intriguing investment. As a pre-revenue company it might already be valued at $10 billion with a very real chance of a 10x return in 5 years.
BUT, inside a $1T high growth car and truck company, $10 billion just isn’t material, folks.
Fascinating, yes. Tantalizing, yes. But irrelevant to our TSLA investment thesis.
If Optimus was a separate startup it would be a very intriguing investment. As a pre-revenue company it might already be valued at $10 billion with a very real chance of a 10x return in 5 years.
BUT, inside a $1T high growth car and truck company, $10 billion just isn’t material, folks.
Fascinating, yes. Tantalizing, yes. But irrelevant to our TSLA investment thesis.
Most lay people are going to judge Optimus by how pretty it walks. Big mistake.
The goal from the start is for the bot to be useful, not graceful. In traditional Agile, their minimum viable product (MVP) would likely have features such as:
- don't get in the way / collision avoidance (don't be anti-useful)
- navigate its surroundings (prerequisite for being useful)
- lift things up and put them down (the most fundamental step of being useful)
All these things require a high degree of perception and understanding of the environment, with further constraints to solve like power-to-weight ratio, repetitive motion stresses, and of course, energy efficiency.
When you contrast that with Boston Dynamics (who I've admired since their initial buzzing dog videos), they are optimizing for fluidity of motion and lifelike kinematics. But their programming is AFAIK completely based upon heuristics, which means they will never be able to scale in a complex environment, like a factory. They are as dumb as the Kuka robotic arms. In other words, the robot will unlikely make a decision to do a back flip in response to an environmental change, unless programmed specifically to do so in very specific and rigid conditions.
To me, this makes Tesla's "useful first" approach vastly superior in the long term.
Couple this key difference in approach with the disruptive Agile practices of Tesla (look up Joe Justice vids on how Tesla is taking Agile to the next level), and it shouldn't be long before the astute observer realizes that the competition isn't coming.
The goal from the start is for the bot to be useful, not graceful. In traditional Agile, their minimum viable product (MVP) would likely have features such as:
- don't get in the way / collision avoidance (don't be anti-useful)
- navigate its surroundings (prerequisite for being useful)
- lift things up and put them down (the most fundamental step of being useful)
All these things require a high degree of perception and understanding of the environment, with further constraints to solve like power-to-weight ratio, repetitive motion stresses, and of course, energy efficiency.
When you contrast that with Boston Dynamics (who I've admired since their initial buzzing dog videos), they are optimizing for fluidity of motion and lifelike kinematics. But their programming is AFAIK completely based upon heuristics, which means they will never be able to scale in a complex environment, like a factory. They are as dumb as the Kuka robotic arms. In other words, the robot will unlikely make a decision to do a back flip in response to an environmental change, unless programmed specifically to do so in very specific and rigid conditions.
To me, this makes Tesla's "useful first" approach vastly superior in the long term.
Couple this key difference in approach with the disruptive Agile practices of Tesla (look up Joe Justice vids on how Tesla is taking Agile to the next level), and it shouldn't be long before the astute observer realizes that the competition isn't coming.
The Accountant
Active Member
and there appeared to be much camaraderie. . . they all interacted well together . . . I sensed that they work 14 hour days not because they have to but because they love it. There was much pride and enthusiasm on display. I am really excited to see continued innovation from this team.
Sorta. Obv it's a long way from brining in profits, but as efforts like this pile up it will be harder for WS to claim that Tesla is "just a car company" and that does go a long way in valuation etc.
It was an impressive presentation all around. A bit dry and "boring" at parts but incredibly interesting and full of info if you paid attention.
The rapid progress on the robot is impressive. I loved all the details about it's physical design, it feels like the hardware is mostly solved already and they only need software dev to make it useful now. And they plan to sell it around $20K per unit? The demand at that price point will be insanely huge. Optimus is going to make us long shareholders SOOOO much money in 3-5 years.
FSD continues to be impressive, but the FSD presentation felt a bit like a repeat of last years AI Day to me, with a few new details.
Honestly one of the things which excited me most was Elon stating they might start doing monthly podcasts on Optimus progress. That's a very appealing notion and I'd love to see this happen.
I don't see this impacting the share price at all in the near term, but for those of us who can understand what this means from a future revenue perspective, and just how soon that could be happening given their progress, DAMN was this an exciting couple of hours! The Optimus program has immense financial potential, and from what they showed that potential is becoming very much a real and tangible possibility. Like Soon. It might be time to start adding some theoretical Optimus numbers to my spreadsheet model now...
The rapid progress on the robot is impressive. I loved all the details about it's physical design, it feels like the hardware is mostly solved already and they only need software dev to make it useful now. And they plan to sell it around $20K per unit? The demand at that price point will be insanely huge. Optimus is going to make us long shareholders SOOOO much money in 3-5 years.
FSD continues to be impressive, but the FSD presentation felt a bit like a repeat of last years AI Day to me, with a few new details.
Honestly one of the things which excited me most was Elon stating they might start doing monthly podcasts on Optimus progress. That's a very appealing notion and I'd love to see this happen.
I don't see this impacting the share price at all in the near term, but for those of us who can understand what this means from a future revenue perspective, and just how soon that could be happening given their progress, DAMN was this an exciting couple of hours! The Optimus program has immense financial potential, and from what they showed that potential is becoming very much a real and tangible possibility. Like Soon. It might be time to start adding some theoretical Optimus numbers to my spreadsheet model now...
larmor
Active Member
The predictions of what it can, and can’t do are pie in the sky, but it literally showed it picking up a box and taking it across the room. That in of itself with a little bit more programming and self-awareness could be used for simple tasks around the factory. This alone is its own company worth a lot more than many companies in of themselves. They cannot be serious about the price if that is really the price, Amazon must be drooling right now.
Artful Dodger
"Neko no me"
At least 3 things came out about FSD during AI Day 2022:
- Elon is driving 'single-stack' right now, and thinks it will be ready perhaps by Nov/Dec,
- Single-stack is already better/safer than Production Autopilot at Highway Driving, and
- FSD Beta v.11 (single-stack) will replace production Autopilot as soon as it is demonstrably safer (Elon says its morally indefensible to not switch ASAP).
Another side benefit to Tesla is that, once all (Hdw 3+) cars are running FSD under the hood, the size of the fleet for the purpose of data/event capture will increase greatly. This will coincide with DOJO reaching operation status for training this larger amounts of data in mere hours instead of weeks.
Chairs!
My thoughts:
The Good:
Tesla basicaly reached #4 and #5 on my list. So this is definitely awesome progress in one year, let alone in 6 months (Feb prototype).
As expected Tesla was able to leverage FSD vision tech quickly into Optimus. Congratulations Optimus, you already have state of the art perception!
The robot is quite less clunky and less weight than Boston Dynamics' Atlas robot.
The Questionable:
The control algorithms used for learning tasks seemed pretty vanilla / industry standards so far. As I wrote previously, they generate an internal model of the robot which then estimates what torques are needed at each moment to achieve the task. They know those estimates will have errors, so then they rely on feedback about the center of mass stability to alter control in real time (regulation of forces around CM is what I focused my PhD on). You could see in the simulations how they were able to achieve faster locomotion than in real life (as of now). I am hoping they evolve into using deep learning for more advanced kinematic state estimation / control between now and the next AI Day.
Sensors - they are big on showing actuators which makes sense because that's what they are good at - but where are the sensors for feedback? They casually slip I believe that the actuators will also act as sensors. This is true that any pressure will produce an unexpected torque on a motor and be sensed feedback, and that is probably good enough for many tasks. But not for anything requiring high sensitivity like handling delicate objects. Again, that's okay for the first generation which I imagine will be focused on factory tasks.
The Ugly:
They claimed a target of 500 W for "brisk walking". This may simply be best than can be done the robotic form of non-human walking, but that is god-awful. With a 2.3 kWh battery pack, that's about 4 hours of walking. Don't expect the first generation to be used for tasks requiring lots of locomation. Factory tasks are still achievable and can require less energy actually, so not a problem for this version.
It depends on bodyweight of course but a human can walk using closer to 50 Watts. Intense running or cycling for a 6ft man like myself uses 300 Watts, 500 Watts is the output for maximal sprinting for like a minute. Part of their much higher cost is the weight distribution is much worse than for humans (more weight spread out distally closer hands / lower legs requires more energy to swing around) and their control algorithms are far from optimized.
Overall
It's great progress. Amazing to turn that hardware into that functionality in 6 months, for any company of any size. You should be bullish that Tesla can take this prototype technology and turn it into a useful robot for a set of tasks, perhaps in factories to start with. There are definitely issues that will need to be resolved over years to evolve this into a more general purpose robot, but that's okay. The one thing I want to see by next AI day is improvement or more clarity in their approach to control algorithms.
Similar threads
- Locked
- Locked
- Locked
