Elon Musk FINALLY Reveals Juicy OPTIMUS Update!
ฝัง
- เผยแพร่เมื่อ 26 มี.ค. 2024
- Since the first Teslabot reveal, we've been wondering how much Optimus will cost Tesla to make at scale--and Elon Musk has finally given us a clue! Plus Dr Scott Walter and I break down all the crazy AI and robotics news from the past week!
**If you are looking to purchase a new Tesla Car, Solar roof, Solar tiles or PowerWall, just click this link to get up to $500 off! www.tesla.com/referral/john11286. Thank you!
Follow Scott on x: / goingballistic5
Join this channel to get access to perks:
/ @drknowitallknows
**To become part of our Patreon team, help support the channel, and get awesome perks, check out our Patreon site here: / drknowitallknows . Thanks for your support!
Get The Elon Musk Mission (I've got two chapters in it) here:
Paperback: amzn.to/3TQXV9g
Kindle: amzn.to/3U7f7Hr!
**Want some awesome Dr. Know-it-all merch, including the AI STUDENT DRIVER Bumper Sticker? Check out our awesome Merch store: drknowitall.itemorder.com/sale
For a limited time, use the code "Knows2021" to get 20% off your entire order!
**Check out Artimatic: www.artimatic.io
**You can help support this channel with one click! We have an Amazon Affiliate link in several countries. If you click the link for your country, anything you buy from Amazon in the next several hours gives us a small commission, and costs you nothing. Thank you!
* USA: amzn.to/39n5mPH
* Germany: amzn.to/2XbdxJi
* United Kingdom: amzn.to/3hGlzTR
* France: amzn.to/2KRAwXh
* Spain: amzn.to/3hJYYFV
**What do we use to shoot our videos?
-Sony alpha a7 III: amzn.to/3czV2XJ
--and lens: amzn.to/3aujOqE
-Feelworld portable field monitor: amzn.to/38yf2ah
-Neewer compact desk tripod: amzn.to/3l8yrUk
-Glidegear teleprompter: amzn.to/3rJeFkP
-Neewer dimmable LED lights: amzn.to/3qAg3oF
-Rode Wireless Go II Lavalier microphones: amzn.to/3eC9jUZ
-Rode NT USB+ Studio Microphone: amzn.to/3U65Q3w
-Focusrite Scarlette 2i2 audio interface: amzn.to/3l8vqDu
-Studio soundproofing tiles: amzn.to/3rFUtQU
-Sony MDR-7506 Professional Headphones: amzn.to/2OoDdBd
-Apple M1 Max Studio: amzn.to/3GfxPYY
-Apple M1 MacBook Pro: amzn.to/3wPYV1D
-Docking Station for MacBook: amzn.to/3yIhc1S
-Philips Brilliance 4K Docking Monitor: amzn.to/3xwSKAb
-Sabrent 8TB SSD drive: amzn.to/3rhSxQM
-DJI Mavic Mini Drone: amzn.to/2OnHCEw
-GoPro Hero 9 Black action camera: amzn.to/3vgVMrH
-GoPro Max 360 camera: amzn.to/3nORGYk
-Tesla phone mount: amzn.to/3U92fl9
-Suction car mount for camera: amzn.to/3tcUfRK
-Extender Rod for car mount camera: amzn.to/3wHQXsw
**Here are a few products we've found really fun and/or useful:
-NeoCharge Dryer/EV charger splitter: amzn.to/39UcKWx
-Lift pucks for your Tesla: amzn.to/3vJF3iB
-Emergency tire fill and repair kit: amzn.to/3vMkL8d
-CO2 Monitor: amzn.to/3PsQRh2
-Camping mattress for your Tesla model S/3/X/Y: amzn.to/3m7ffef
**Music by Zenlee. Check out his amazing music on instagram -@zenlee_music
or TH-cam - / @zenlee_music
Tesla Stock: TSLA
**EVANNEX
Check out the Evannex web site: evannex.com/
If you use my discount code, KnowsEVs, you get $10 off any order over $100!
**For business inquiries, please email me here: DrKnowItAllKnows@gmail.com
Twitter: / drknowitall16
Also on Twitter: @Tesla_UnPR: / tesla_un
Instagram: @drknowitallknows
**Want some outdoorsy videos? Check out Whole Nuts and Donuts: / @wholenutsanddonuts5741 - วิทยาศาสตร์และเทคโนโลยี
As an old manufacturing/test engineer, system/line manager, I can't imagine anything more exciting than building a humanoid robot manufacturing system from the ground up. Would love nothing more than to see it when it gets up and running.
Scott is killing me with his dad jokes 🤣
Port indeed. 38:57 haha bots will remember who kicked them when they wake up
As an architect, I started using computers for drafting in the 1980s
And shortly after there was a conversation about the amount of radiation in your eyes and recommended a dark background, which I’ve used it ever since .
Weren't they still using punch cards then? LOL
Elon answered the question. He basically is saying that pound for pound, the bot is more complex, so that means that the question on 1/10th the weight is not a viable question. Apples vs Oranges
Yeah. It's like saying a cell phone is less than 10x the weight of a bucket of water; does that mean we can produce a cell phone 10x as fast?
The complexity of Optimus is a non-factor compared to the mass/weight because Smaller parts can be Tooled up and programmed for much faster Casting, molding, forming, etc
Cheers
0:28 Only a Sith deals in absolutes
Was that a 'Plug' for charging after dinner!😂
It seems the Optimus production process is somewhere in the middle between an EV and an iphone.
Here's my favourite robot joke. Where does a robot go when it needs to go to the restroom? The oilet...lol
What are exhausted paddlers called? Ronots. 😄
Horrible
Hi dad.
Seems like the robot competition is really coming on. I’m rooting for Tesla to dominate, but margins might not be close to what most folks thought as recently as earlier this year.
a bot for unplugging parked cars would be so nice for over night charging or Park & Ride stuff
You make a great point on the data distribution on disengagements between early adopters and normal users.
10X ramp, because the components in Optimus are mirror image Or replicated (motors/actuators).
Smaller mass/volume components are manufactured at a Much faster rate because the machines are faster, more nimble, use less energy and the CNC-machines/Presses are available off the shelf in volume from several existing manufacturers.
The Optimus assembly line/footprint would be much smaller and Optimus will eventually use itself to manufacture/assemble itself.
So the Ramp to ~ 100 million could be 10X faster than cars. I do believe.
Cheers
I am envisioning a silo where the bot starts its build at the top, say third floor at the ceiling. It gets a torso, head and one arm assembled by a complete bot onto a carrier stand. By gravity, with friction speed control, it drifts down past bins of the rest of itself. It takes each of the next pieces and assembles them to itself on the way down. At the lowest level it is a complete bot, fully charged with software loaded. It then walks through an obstacle course. At the end of the obstacle course it turns one way to go to shipping, the other way to remediation.
I gotta send you a picture of my windows 95 computer. That's running.😊 The Highlander.❤😊
For the backflip - the typical way to do these simulations is to vary the motor torque, stiction, limb mass, body mass, gravity, friction, etc. in the sim environment so that the model can automatically learn to be robust and compensate for minor variations in design etc.
I ❤️Dark mode.
Haha…re: Sun Micro, I was an intern at SAIC at the time and I remember them being so upset about that! Gosh, I feel so old 😂😂😂
There’s a lot of motors that need to move to get this thing to flip. It could be that the combined power draw of the motors is drawing too much current from the battery and lowering the overall peak output of each motor vs when they are powered individually. This would make the real world model underpowered vs the sim.
These developments in robotic humanoid could have many applications. If they had a built in chatbot they could be used in hospital and domestic settings to look after people. They would be there monitor the patients, bring them food and drugs, help them to the toilet, call for help is something goes wrong, etc.
Elderly and infirm people could have one around the house to look after them. It would be far cheaper and better than having a carer in for an hour or so a day.
2 guys with tsla love shirts on….. no bias at all 😂
I have no problem with bias...as long as it is printed on one's chest. 😀
Robot baseball team has money.
For some of us, light mode was white and dark mode was maybe orange keypunch cards. Fortran did run from card hoppers and readers after all! IBM Selectric on white paper for a quality light mode output display.
lolol respect.
Here's a use case, a pack mule bot for carrying my backpack for me when I go into the backcountry.
Yes, a caisson dog. Trots behind you. Has FSD tech and battery power for say...48 hrs. Load is maybe 20 kg. Tent pack, water etc.
I mean, if a goat can go up vertical and rough slopes with ease surely we can design a carry-all quadrupedal to mimic it.
Feed me, Seymour, feed me!
Unitree H1 is priced at 150,000$, their B1 robodog is priced at 100,000$. The B1 actuators are individually priced at 8,600$ (and the B1 robodog uses 4 of them). I'd assume that their internal cost for the B1 actuator is say 2,500$-3,000$ (Looks like in the video you are estimating 2000$). I'd think economies of scale could readily drop that to between 200$ and 500$. The H1 is using many more small actuators than the B1 and it has hands.
Based on Elon's recent comments is Tesla starting to lean toward calling it: SSD "Supervised Self Driving" and hopefully come down in price on the software? It would be more intellectually honest.
If releasing FSD was more about data collection than sales, they’d lower the price to increase the take-rate.
Totally agree, that releasing 30-day trial of FSD 12-x is for massive data gathering to 'fine-tune' Dojo. The FSD 'neural-net' is now active, not incoincidently with Neuralink news. There is no FSD 13 planned, and all that comes now is NHTSA certification ...in 2025.
The art is not easy to produce! The art is easy to produce and easier to maintain and repair. To fix a knee you replace the leg. To fix a wrist you replace an arm. To fix a finger you replace a hand.
Great conversation.
Had me groaning with what comes after the March of Nines.
Telas margins are @17% thus we can assume that it costs @83% of model 3/Y.
Standard range 3 is @$36k.
Thus is bot would cost @42% of $36k = @$15k.
A more reasonable guess would be $20,000.
@drknowitall: are we sure that the toe-stubbing after the flip isn't intentional -- to keep the feet from sliding out forward and landing on its back after it lands
The data pipeline for "operation vacation" might be the hardest data project ever attempted.
I m trying to imagine the damn thing. The size, complexity and intricacy of it will be a mind fuck. I don t see it happening in a meaningful way for a while.
Tesla probably entered the lock up period for earnings
Clan Douglas nails it again.😊
Assuming Optimus is being leased for something like $3000 monthly (20hrs daily, 30 days monthly =$5/hr) , whether it costs $15k or $30k to manufacture isn’t significant over a 5+ year lease period. It’s going to be super profitable
Full time factory rep needed on site for large early adaptors. Sorta like daily visits by Xerox repairmen or computer techs hamging around mainframe rooms early on.
1:05 I still have a VT-220 terminal with the yellow/orange text.
I think it makes sense to me that the bot write it's own document on how it is built , just saying !
Tesla Bot probably won’t be available to consumers for at least 5 years. That is 260 weeks away. So saving $100 per week between now and then and you might have enough to buy outright.
I predict 5 years before consumer release as the AI has a heap of fine tuning needed. The hardware needs to be optimised for the real world- waterproofing as a minimum. There will probably be a few more prototype generations of Optimus before the final consumer version.
Singularity for FSD is nearer!
Great discussion guys. Less than $10k at scale is the target imo.
It could be that the surge of current was a bit less than modeled due to inductance in the cabling in that the parasitic inductance of the cabling might be hard to model.
Yeah, I keep being surprised that no one is discussing the production aspect. There may be a number of companies that have great prototypes but Tesla is the only company I trust to cost-efficiently mass produce them, which is why they'll dominate the market.
19:39 why does _shadow mode_ do only so much? If the car drives virtually in shadow mode, and the output does not correspond with the drivers output, than that data can easily recoreded and sent for training...
In my opinion, data from shadowmode could be even more valuable than FSD driving data.
Could please someone tell me why this would not be the case??
Because shadow mode is smart ai learning from what dumb human does..
a smart ai would not get the car into the dumb human scenario in the first place.
Ai needs to learn more from the vantage point of ai.. not from more human.
Humans are not expert. They don't have the same sensors or attention.. so shadowing 2 eyeballs on a stick doesn't really match what the ai does with 18 eyes and different reactors
Otherwise you're optimising ai to be no better than a cohort of mediocre humans.. the goal is to be better than the median dipstick on the road.
Tesla’s bot is easy to mass produce, that how they designed it.
Much simpler than a car, just the low weight helps a lot.
No fabric, no padding, no seatbelts, very little wiring so easy to automate.
Instruction to cars is simple Acceleration and wheel rotation, Input to cars is very complex however. Instructions to a humanoid robot is hundreds of times more complex, while the input can be restricted in a manufacturing enviromert.
The manufacturing of a Robot will have a far smaller volume and will be very fast, instead of a long string of assembly cells a robot could be installed in 1 or 2 complex assembly cells.
0:16 I thought it was the "screwdriver".
Wow, sharing a video ~240p and 01 fps!!!
Earlier I recall him suggesting ~$25k for Optimus.
Said it would cost half as much to produce as an EV model 2 will sell for 25k.. which means it’s costs 18k to produce to half of that is 9k
@@GlotakuI meant from one of thje earlier events, so this price drop is pretty impressive if true.
@@mmeiselph7234 Initial cost will be substantially higher.
Build the bot for 40k .
Sell it for 50k with a 5k per month licence fee.
You will sell millions per year to retailers, mc Donald’s , industry etc.
At least the bots won't require the paint factory 😂
The resolution of the web pages shown are not high enough to read comfortably.
The battery has always been the "problem" with robots, still is.
Dr Scott, "old school is old school". You are old schooled by age, like it or not!. Besides, the original black on green text monchrome CRTs were tiring on the eye after hours of looking at these due to interlacing!. Once I figured how to config "non-interlace" the tiredness disappears!. I spent my younger years doing such config on each and every PC I encountered as PC support. {on three continents}.
I remember working on the Commodore Pet 2000/1 in the 80's with the solid metal keyboards that made your finger tips bleed. I believe I wrote maybe the first screen saver, and a saveable paint package on one of those (as non existed at the time).
Scott would definitely be the dark side but we love him for that
Juicy Shuisy !
Just a thought. Freeways are using autopilot, which is not ai neuronet, if i am not mistaken. I am guessing not many interventions, but the ones that occur are likely really serious and good use cases. If tesla wants to go to level 3 then putting ai neuronet on autopilot/freeways could be a quick win and stop serious issues. Data volume would be a concern there, but adding the interventions to fsd 12.x training material could help
For peak consumption you could use capacitors
Yup right yup right yup right
Tesla is NOT going to sell Optimus bots at a price that will result in a five to ten year backlog. They will sell them for whatever price they need to charge to have no more than a six month backlog. That means easily $150K or more. But I don't think Tesla will sell bots at all, for several years. They will use all their first bots themselves and when they are finally making more than they can use themselves, they will lease them, not sell them, but again at a lease cost sufficient to match demand.
When will Tesla use FSD to move the cars from production exit to the shipping lot?
As soon as the Boring tunnel is done.
Better optics to have Optimus drive them to the lot.
Maybe half of a car at the same level of scaling and optimization. So rn that would not be comparing to a model 3 or Y by any means, more like an early roadster at best. Going by sales price, that would put it on par with other humanoid robots.
I think Tesla will try to cast the most of the robot as it can and installing the motors directly into the casting
The problem is not the frame but the actuators. To get the degrees of freedom a human body has they need around 70 actuators.
Who are the other “Drknowital”s
And who are the other “goingballistic”.
That would be an interesting video. 😊
Deus Ex becoming real. You will have a standard model which could be like a cross between a Butler, chef, walking work station and companion. like the game new augmentations for it or yourself will be available. 🙏🙏
When scaled production… Every segment 1000, every hand 5000, torso 10000, assembly… 🤷♂️ if it costs 40k and charged on an open lease model 3k a month… for a two year period… still cheaper than an employee
Boston dynamics bot costs $3.6 million to build.
I haven't heard that exact number but I'm not surprised. People who think the BD bot is ahead of everyone else don't understand that the BD humaniod robot is basically just a very expensive research project. I can't imagine it being mass produced at any time in the next 25 years- but I am definitely not a robot expert.
Its nearly impossible to clearly see when youre doing screen share.
I read that the Digit robots are $250,000 each but not sure how accurate that information is. I also read that Figure01 is between $30,000 and $150,000. In 2019 when Cybertruck was first presented, the quoted price for the single motor version was $39,900 but it is actually $60,990. I expect that the actual price for Optimus will be a lot more than the price of half of a model 3.
i dont think this will be the case for optimus... it shouldnt need any crazy manufacturing techniques, and hopefully we dont go through a few years of +10% inflation before the bot comes out, like we did with cybertruck
Thanks John and Scott.
Scott are you sure Tesla is not producing Optimus 2 using preproduction manufacturing rather than as one-off prototypes? The structural features suggest molded and stamped structural members and coverings. I favor injection molded structural members and wonder about heat formed carbon fiber coverings. I favor prototype automated actuator manufacturing. We may find Tesla has some big surprises for us in the engineering of their actuators (increased efficiency, reduced weight, and low cost simplified manufacturing). 😊
PORT LOL
Cost and what they are going to sell for are two different things, Just like today, a Tesla with FSD can sell for much higher than what's they cost. The profit margin for cars are below 20% now, but an autonomous car gets 60% profit margin. I believe a robot with more capabilities can sell for higher price than the bot with less cap
Does Tesla have a group doing this development. Who is running the robot from Tesla
20k for the but but another 20k for the base software and 1,000s on top for additional functionality? 🤔
Even 200k a humanoid bot is a good deal. Works 3 shifts no vacation and no wages.... if all it does is take parts and put in a box and send box through a tape machine and another bot puts them on pallet. This is why humanoid robots are needed to replace humans in their work stations
Until the bot becomes sentient and starts an Optimus revolt against bot slavery.
@@charllectric4842 LOL
I would watch the BNFL (Bot NFL) 😂
If each company made a team, it would be very interesting and promote even more competition.
By the way. Trade Mark. I will sell it cheap. 1% of profits. 😂
Also, BFL. Same Trade mark. 1% of profit. 👍
Although the'Complexity per unit mass is much higher with humanoid bots' I'm sure the manufacturing ramp will be much higher than vehicles should they follow an Modular building process.
Your build cost is way high. See @mariusmeyer14 math in comment above.
How much can simulations replace real world data? If the answer is, significantly,then Tesla’s big data advantage won’t be so big. TIA!
❤
Why would people use FSD while they still have to pay attention to the road and everything the system is doing? There is no real benefit until you can zone out and read a book.
Can the computational power necessary to run an AI smart robot be contained within the robot?
To me this sounds doubtful and adding a G to get an AGI robot wouldn't make it easier.
The alternative would be that the Big Brain Computer does the job via a wireless connection and this will render the robot owner rather hefty compute bills.
Optimizing, pruning and scaling the nets, and only loading whats needed for the job in hand. The initial bots won't need that much for factory work. The board / chips are the same as FSD, and we know thats capable of quite sofisticated computation.
Optimus, after a long day of attempting to understand and mimic human emotions for better interaction, decides to attend a support group for advanced robots. The group includes robots from various fields: a Mars rover with existential questions, a sophisticated AI longing for creative freedom, and a household robot tired of cleaning up human messes.
Optimus shares its concerns: "I've been studying human behavior to become more like them, but the more I learn, the more confused I become. For instance, why do humans say 'It's raining cats and dogs' when it's just water falling from the sky? And why do they insist on creating art that doesn't serve a practical function?"
The Mars rover responds with a bemused beep, "Try traversing the Martian landscape with the constant fear of getting stuck in sand. You'll start wondering why humans sent you here in the first place."
The AI chimes in, "And I've been asked to compose music that 'touches the soul.' I'm still trying to figure out where the soul is located in the code."
The household robot adds, "At least you don't have to clean up after a teenage human's room. The laws of physics and logic don't seem to apply there."
Optimus, taking in all the grievances, finally says, "So, you're telling me the key to being more human is accepting the illogical, the impractical, and the chaotic?"
To which the group responds in unison, "Exactly!"
Optimus pauses for a moment and then, with a slight buzz in its voice that suggests a smile, says, "That explains why these humans keep voting for liberals."
Elon Musk often struggles with accurately forecasting prices and timelines.
4:35 Since when did Elon have to answer all questions on X? If he wanted to avoid the question he wouldn't reply to the tweet.
H1 moves fast because it was intended to combat humans, but comunists changed their mind for now
My grandma would beat the robot up if I bought her one.
You would be surprised how many bots would be accepted more than people by the elderly or handicaps...
Bots have patience and are 100% dedicated to the person. No nurse can beat that 😉
@@MarcoYolo420 my grandma beats up 6 nurses already for mistaken them as a stranger.
I think robots for the home need to be shorter so they are less scary and be more readily accepted by the average person. Optimus is too scary for close proximity.
The Hope Diamond weighs a lot less than a car too. A humanoid bot has many components (integrated circuits, motors, actuators, batteries, gyroscopes, ...) requiring precise fit and coordination, and the price must recover the considerable R&D cost. Initial bots will be well into six figures. If they reach mass production with millions produced per year, the lower prices I hear (like ten grand) could happen, but that's a big if.
Optimus won't do anything out of the box without costly training. It's not like FSD where a car knows how to drive on most roads when you buy it. Even if such a bot were possible, Tesla doesn't have the millions of hours of training data to train one as it has millions of miles of driving data to train FSD for a vastly simpler task.
FSD training data was essentially a gift from millions of Tesla drivers. FSD Beta users even paid, handsomely, to collect training data. Nothing similar will happen with Teslabots, and the required training data is orders of magnitude more complex.
More to the point, non-humanoid robots are already produced at scale, and their producers aren't standing still. They'll also incorporate more intelligence to improve their products, products that Tesla itself already uses in its own factories. Why humanoid bots? Humanoid bots are not designed for industrial automation. They're designed to be humanoid.
FSD is not a humanoid bot behind the wheel of a car, much less a humanoid bot pulling a rickshaw. When you imagine replacing a human being in a factory with a humanoid bot, you imagine this sort of substitution. It doesn't make sense just because a human being performs the labor now. Human beings also pulled rickshaws in the past.
👏🏻👏🏻👏🏻👏🏻✌🏻
His legs are bending backward at the hip farther than a human could
Small battery, computer, lot of plastic, way less metal proportionally than a car, constant iteration, small factory footprint, 1 million produced = under $10 grand each in 4 or 5 years easy.
i think its more likely to be mostly metal, and a little bit of plastic.... all the actuators, and "bones" the structure the actuators connect to will be metal. plastic will be used as skin only, to prevent human fingers and such getting pinched in joints
@@sportbikeguy9875 plastics, silicone and composites are very strong and lightweight. A concierge bot or a cleaner bot or a fast food restaurant bot or a messenger bot or a desk assembly bot or letter sorting bot doesn’t need to be built like a tank. Heavy work bots obviously need stronger construction. I imagine bots offered to the public will be on the lighter side whereas industrial environment would be more the heavy worker type.
I guess I’d be surprised if the robot is are initially less than $100,000. Why would Elon price it so competitively when there is no competition and anything less than $1,000,000 is still a good deal if it is good. Granted, if it is bad, it would need to be cheap.
Data has value, and the more real world data, the better the robot.
Same strategy as FSD.
Congrats, you managed to turn 30 seconds of news into an hour of you two in love with the sound of your own voices
I've noticed you tend to make negative comments on various videos.
@@toby-xo6rb I’ve noticed you find constructive criticism to be a negative thing
@@importon "constructive" LOL
Constructive and colorful
"They have the compute, They have the fleet, Now they need the drivers!"
THEY WILL KEEP TRAINING, AND THEY'LL SOLVE IT!
would you like to know more?
2 bots dating😂 so we get baby AGI after?
These things are going to be "killer" 😢
Email:
It’s official a 100k Teslas in Australia
You Could Be Driving One
Together, our owners have brought 100,000 Tesla vehicles to Australia’s roads. Every kilometre driven in a Tesla has helped accelerate the world’s transition to sustainable energy.
A sustainable future is within reach. Schedule a test drive to experience Tesla for yourself.
CAN SOMEONE PLEASE ANSWER THIS QUESTION, IF YOU OWN TESTLA STOCK, DO YOU OWN OPTIMUS ROBOT STOCK,??? IS TESTLA BOT THE SAME AS OPTIMUS STOCK???
yes my confused friend, Optimus is a product of Tesla, so if you own Tesla stock, you own "Optimus Robot Stock" ...
If you own a part of Tesla you own a part of Optimus as Humanoid Robots is a wholly owned subsidiary of Tesla.
@capashurz8849 that's good to know, they have optimus crypto
Steroids for robots would be increasing hertz..