Tesla Bot Goal to Automate Labor

Elon Musk and Tesla revealed the vision for a general-purpose humanoid robot.

They would start with the most boring, repetitive and dangerous work. The work humans least like to do.

What is dangerous work? Installing solar power onto buildings is dangerous. Roofing is considered the fourth most dangerous job in the United States. Some of the most common hazards for roofers include slip and falls, electrocutions, and burns.

Tesla needs robots to install solar power.

Other dangerous jobs:

Mining, garbage collection and iron working is dangerous. Tesla needs to mine more materials for their batteries and cars. Steel working is needed for Tesla cars.

Logging is dangerous as is fighting wildfires.

Logging and firefighting are not something that Tesla is currently involved but it is a key part of climate change.

Delivery drivers is the seventh most dangerous job. This is clearly matches well with what Tesla is already doing. Couriers and delivery services are estimated to be a $126 billion a year business that is growing at 4% per year. Automated delivery and logistics improvement could expand this industry the way self-driving would expand taxi and trucking businesses.

Highway maintenance work is also dangerous. Boring tunnel maintenance and support would also be dangerous. Narrower utility tunnels would be useful but dangerous.

A few months ago, Hyundai bought Boston Dynamics with a valuation of $1.1 billion. Boston Dynamics was purchased by Google in 2013, then Japanese investment firm SoftBank in 2017, which bought it from Google for $100 million before injecting $37 million more in 2019. Now, Hyundai holds an 80 percent stake in the company and SoftBank retains the remaining 20 percent.

Boston Dynamics has about $50 million in revenue and about 300 employees.

Sarcos Robotics is another player with innovative robots. Sarcos thinks they can make dexterous mobile robots. They are projecting $2.7 billion in sales just five years from now. Sarcos is targeting exoskeletons and a remotely-operated humanoid-type robot that’s anchored to an aerial work platform and remotely controlled by a human.

If Tesla is successful in recruiting 200-300 of the best AI researchers and graduates then they will have the opportunity to match or exceed Boston Dynamics and Sarcos within a year in humanoid robotics. Tesla already has some of the best AI researchers and they have the batteries, vision and other systems functioning already. Tesla and Elon’s other companies can work on several useful niches to generate early revenue. It will be relatively trivial for Tesla to get initial revenues beyond the $50-100 million levels of Boston Dynamics or Sarcos. Elon is far more skilled at scaling revenue.

There were claims and still are claims that Waymo is ahead of Tesla in self-driving. However, Tesla is able to generate over $10,000 to $12000 for about 25% of their car sales with FSD sales and $4000 for autopilot sales. This is over $1 billion in AI software sales every year now that they are selling 800,000 to 900,000 cars in a year.

Tesla’s vision only AI/camera navigation will scale down to the humanoid robot. Tesla has experience with robotic arms used in factories. Tesla will access the utility small labor market with the cybertruck and a cybervan. Tesla can work with automated or semiautomated robotaxi. Incrementally adding a deployable small robot to the robotaxi or a Tesla human driven ridershare fleet would be trivial. Tesla can generate billions in revenue with a 50% complete humanoid robot or even wheeled human scale robots.

76 thoughts on “Tesla Bot Goal to Automate Labor”

  1. Wheels are extremely useful on smooth, flat surfaces (for varying values of smooth and flat). Very efficient and capable of high speed for minimal energy use.

    In our world, we've made huge amounts of smooth flat surface, so it would make a lot of sense to have optional wheels for a robot that isn't going to spend its life in limited, off-road, environments.

    Legs and wheels. Like heelies in the feet.

    Reply
  2. Eh, Bezos was moving towards shelf mover Kiva robots and fixed picker arm stations. Packing items might still be hard though. For now…

    Reply
  3. If all those steel mill accident videos on youtube that usually start with "meanwhile in russia" are any indication, mill jobs are plenty dangerous too.

    Reply
  4. This specifically bit Elon during his Alien Juggernaut phase, as he discovered automotive wire connectors aren't designed for robots.

    Reply
  5. It's not the large labor pool itself (though that does have interesting effects on project management) but that it's a cheap labor pool. With the uplift of the chinese standard of living it isn't that cheap anymore.

    The crossover point for western countries to bring back manufacturing will be a reasonably practical general purpose human stand-in, and more adaptive manufacturing lines allowing faster changes, ideally coupled with work on lights-out manufacturing. Once that happens, then it becomes a raw material access question. If I can mine stuff domestically, or dispatch robots to foreign countries and return raw or somewhat processed materials to domestic manufacturing facilities, then things can accelerate dramatically. At that point, resource rich but poor countries would need novel regulations to keep from getting ripped off.

    Such as a cut of the processed materials stream itself, or timeshare access to the robot fleet for domestic infrastructure projects. Otherwise, they'll end up being hollowed out by a locust army of TeslaBots, with only a meager bank account to show for it.

    Reply
  6. Considering all the crab evolutions, we might see either a full crab, or perhaps a centaur? A centaur with wheels for hind hooves could behave like a Boston Dynamics Handle robot when "standing"…

    Reply
  7. A relevant comparison would be Xiaomi's Cyberdog, which resembles Boston Dynmaic's Spot. Xiaomi are starting from scratch, with access to modern neural network software stacks. They don't have a DOJO class supercomputer but there's plenty of government help there. If the actuators are pure electric, then we might see some apple's to apple's comparisons.

    Apparently Elon has dealings with the RoMeLa lab, which has been developing the BEAR all electric actuator.

    Reply
  8. I think you mean

    soldierbots will continue to be used in human-human conflicts

    I mean, what do you think a predator drone is?

    Reply
  9. On the principle that software is easier to mass produce than hardware, I'd say no to centaurs: Walking with fewer legs is just a control problem, and it's pretty much solved already.

    Reply
  10. Of course it is very hard to predict, but I have a feeling we will develop the ability to create soldierbots a lot sooner than we will develop the ability to change human nature via genetic editing or any other way. Therefore, I think it is much more likely than you would like that soldierbots will be used in human-human conflicts.

    Reply
  11. My thought was only to be able to shrug as part of non-verbal communication with humans, assuming the robots can be capable of such sophisticated communication. Any movement that looks like a shrug to a human would do. I don't believe it matters how closely the internals match how it happens in a human.

    Reply
  12. Centaur. That's the way to go, at least at first. Boston dynamics has the bodies and legs, you just need a humanoid torso/head/arms.

    And by arms I'm thinking several. Either Kali style, or a squid.

    I think squid would be more effective, but like the torso and head, human like (if with more arms) gives better interaction with humans, which is a big part of most jobs. At least on Earth.

    Reply
  13. I'm not sure where and when you are posting from, but in 2021 Earth it is the WOMEN who have been saying for some decades now that their mechanical alternatives do a better job than many men.

    Reply
  14. Now THIS version of your human extinction concern is something that rings true to me.

    And now I'm wondering why I have never seen a SF drama where the hero[ine] is trying to score with some romantic partner but is unable to compete with the "perfection" of androids.

    Yes, Stepford Wives was sort of waving at that possibility, but they still had to keep the robot thing a secret.

    Reply
  15. Why 2-legged?

    Surely boston-dynamics, and 300 million years of evolution, shows that 4/6/8 legs is usually easier and more stable?

    Reply
  16. Only the "Receivers". From there it's encrypted 5G amongst themselves.
    For someone who expresses concerns about AI, Musk sure seems intent on building SkyNet.

    Reply
  17. I hope that when we will have tech to increase our IQ/wisdom like BCI's, CRISPR editing, other methods, we will become on average wiser, intelligent enought to realize how stupid war with your own people is
    (it's literally like cancer). On the other side, having some powerful weapons for potential war with unknown, malevolent alien race exiting somewhere deep in the Galaxy/Universe who are for example trying to kill all races they meet, make sense for me.

    Preparation for unknown, alien evil race = yes
    Preparation for war with your own people, human civilization members – stupidity

    Reply
  18. You seem to be tacitly assuming that only males will be using sexbots.

    In reality, we'd probably require artificial wombs and robot nannies, or at least a REALLY hefty subsidy for parents, to the point where intelligent people would view having and raising kids as a desirable career path.

    Reply
  19. Yeah but if only a fraction of men prefer real women then polygamy can keep us going. Meanwhile, the rest of the men still need danger pay to be able to afford the extended warranty on their cyber-thots. We will have to think of what tasks robots will find repellent and hire humans to do for them.

    Reply
  20. You'd probably want it in Ludus' application, as well as an extra degree of freedom for the hip. Otherwise a robot can get by without it.

    Shrugging is generally a byproduct of the necessity of moving the scapula and clavicle around to get a good range of motion on the shoulder without popping out of the socket. A robot shoulder can get the same range of motion without shrugging, by distributing the motions differently along the arm.

    Reply
  21. You're thinking pronation/supination, which I agree is very important. That's accomplished by twisting the radius and ulna around each other in the forearm. The hand rotates about an axis through the length of the forearm.

    Ulnar/radius deviation is rotation of the hand in the plane of the palm; In one direction you're moving towards the radius, in the other toward the ulna. It only has about 50-60 degrees total movement, tops.

    https://www.researchgate.net/figure/Wrist-and-forearm-movement-a-flexion-extension-represents-pitch-b-radial-ulnar_fig1_331176887

    Reply
  22. Tesla Bot Goal to Automate Labor

    You dont need a humanoid robot for that, that's very last thing one would produce if you're interested in utility. Personal interactions is just about the only area where there may be some utility to producing the humanoid form.

    What is lacking to automate human labor is robust AI and not a humanoid robot .

    Reply
  23. That depends on the cost of the automation, doesn't it? Lower the cost a lot, and the profitable opportunities will expand greatly.

    Reply
  24. While shrugging probably isn't necessary for performing many tasks, it could be useful for communicating with humans. So don't cross that capability off the list completely. Just move it down a bit in priority.

    Reply
  25. Nice reference to classic SF on the part of you and Brett. However, as I said in another comment, I think it is inevitable that they will be employed as soldiers as well. Don't leave that out of your planning.

    Reply
  26. You may or may not be correct to characterize our species as young and stupid, but war or at least lethal aggressiveness is human nature (perhaps primate nature, given observations of war-like planned activities within chimpanzee wild populations). Any plan that ignores human nature starts out with at least two strikes against it.

    Like it or not, some countries are going to try to employ robots like they do human soldiers. Countries that ignore that are in danger of being overrun by countries that recognize it.

    I have a feeling that lack of a power source that is both compact and holds enough energy to run a soldierbot for a few days at a time will greatly slow down adoption of soldierbots, but sooner or later, they very likely are coming.

    Reply
  27. Three days already passed and Brian is still silent about the ignominious defeat of the USA in Afghanistan.
    Maduro still in power despite Brian told everybody that he would be gone in 2018.
    And Iran enriching at 60% despite Brian told everybody that Iran had been contained last year.
    Brian as a futurologist is worth less than my grandmother

    Just .. sayin!

    Lukkkkka

    Reply
  28. I think it will incredibly interesting to compare their progress to Boston Dynamics because, as I understand it, Boston does not use any machine learning.

    Reply
  29. Fair enough; I thought that some types of actuators could have two degrees of freedom, which would give 24. But in any case, this looks like the count of Raptors on Super Heavy, that is, subject to many changes between now and when the first prototypes show up.

    Reply
  30. I'm a mechanical engineer, have designed hobby bots, including at least some design studies for a human equivalent walking robot. (I was curious what it would take to build one.)

     6 degrees of freedom per leg isn't unreasonable; Human legs have at least 7, neglecting the toe motions, (Which are really dispensable for walking purposes.) but hip rotation is redundant for walking, so assigning 12 actuators to the legs is reasonable.

    Arms have a lot more degrees of freedom, (The shoulder alone has 5!) but you could get the job done with three for the shoulder, (The robot has no need to be able to shrug.) one for the elbow, and three for the wrist, that's seven per arm. So they seem a couple degrees short on the arms. Maybe gave up on the ulnar/radial movement, it's kind of limited in range anyway. 

    No way 12 actuators will give you human dexterity in the hands. Not the degree of dexterity a touch typist requires, anyway. But you could probably get enough dexterity to utilize most tools a human would use, to grasp and pick things up. Something a bit better than a high end prosthetic, maybe. I suspect most people have never learned to use all the dexterity a hand has, anyway. 

    I guess you could build a robot that could do most human tasks with that many actuators, but it would be a stretch calling the hands "human equivalent". "Ham-handed human equivalent" probably didn't sound good.

    Reply
  31. 9/11, rings a bell?

    It doesn't matter how smart and advanced your nation is, it only takes a bunch of over-confident and/or fanatical fools running free in some place with enough determination to harm you, to start a war.

    Reply
  32. To be more precise I meant aggresive side(those who attack), other side have no choice but to defend itself. Still the whole concept of human to human war, killing eachother, fights between parts of the same and one Human Civilization is proof that we are young and stupid as a species. I hope that we will soon start becoming wiser exonentially, just as fast as our tech

    Reply
  33. Indeed. This is a much better fit for wi-fi and 5G.

    The humanoid robots will be roaming around human cities most of the time anyway.

    Reply
  34. Well, if we let evolution be our guide, they will either look a lot like us, or like crabs. https://en.wikipedia.org/wiki/Carcinisation#:~:text=Carcinisation%20(or%20carcinization)%20is%20an,Nature%20to%20evolve%20a%20crab%22.

    Actually, by flexible, I was thinking of how I, as their manager, would be able to employ them. A human form would offer me the most options imaginable, given the existing wide assortment of devices, vehicles, and tools intended to be used by humans.

    Maybe you could dispense with pinkies but they might have to use an archaic keyboard (rather than just connect with Bluetooth) at some point. An extra set of arms might be good, but the mechanics of it would require sacrifices and probably make them unable to fit in some vehicle cockpits.

    Reply
  35. Yes, but I think that war is an important enough activity that generals will want specialised robots to do it. A Predator drone is much more useful than a humanoid soldier already.

    Also, Musk has mentioned that he's making a point of having Tesla bots be slow and weak enough that you can "run away from them", and possibly overpower them, so I don't expect them to be exceptional soldiers even if they can be taught to shoot an assault rifle.

    Reply
  36. One slide on the presentation mentioned 12 actuators on the legs, 12 on the arms, and 12 again on the hands. Perhaps these 12 actuators can yield close to the 25 degrees of freedom you mention.

    Of course, the Tesla bot is still at this stage a PowerPoint robot, and we've heard all sorts of cool tidbits about MCT/ITS/BFR/Starship which quietly disappeared as the development advanced.

    Reply
  37. They need to get the human environment navigation right. The bots must learn the home map and facilities (washing machines, doors, charging stations, etc.) on the fly, and adapt to changes (a piece of furniture moving, trash in the way, water on the floor).

    On the street, it has to work along humans without crashing into people or falling over. This would require live mapping capabilities and cooperation with the cloud, something Tesla FDCs already do, but on a different context.

    Reply
  38. I concur. A two legged dexterous robot would work very well for planetary missions. Even as a tele robotics platform for people nearby, when people arrives.

    It would require to be ruggedized for space radiation and vacuum, becoming a new variant of the commercial product on Earth, but it would be worth the effort.

    Reply
  39. My hunch is he wants to send one to Mars on the first Starship lander there as an impromptu astronaut, if it's available.

    Given such landings can realistically happen in 5 years (3 if you are really, really optimistic, next year no way), they have some time to make at least a walking bot, that can be remote controlled to enter into the elevator and take pictures.

    Reply
  40. I'm not so sure that the humanoid form is the most flexible form for a robot. It's flexible and has advantages with environments and tools that were designed for humans, sure. And humans have an intuitive advantage in programming them. But I'd be shocked if evolution had converged on THE best form.

    By the way, the illustration says "human level hands". I doubt that, unless they're being a bit coy about what they mean by "human level". 25-26 degrees of freedom is a LOT of actuators to pack into something the size of a human forearm. 

    I suspect that robots could get by without a pinky just fine.

    Reply
  41. Everything Elon Musk does is about settling Mars. Spaceships are necessary so SpaceX. Better batteries are necessary, so Tesla. Underground construction is necessary so The Boring Company, smarter and more capable humans will help, so Neuralink.

    More capable and flexible automation is also necessary, as the cost of labor (and the danger to labor) on Mars will be, to borrow an overused but doubly appropriate adjective, astronomical.

    As far as making them humanoid? It's probably not as efficient as making a robot for a single set of tasks (like welding on an assembly line) but is by far and away the most flexible form (especially as it can fully use human infrastructure), short of the living metal/nanobot robots in the Terminator sequels, and you only have to design and mass produce one, rather than endless varieties.

    At some point I expect he will make a serious push for radical life extension (if he is not already), both to keep those priceless Martian settlers around longer, and himself, so he has a better chance of living to see it all come to fruition.

    Reply
  42. Good sexbots with social AI are an extinction level event for humanity. We'd be so far below replacement level reproduction you couldn't see it from where we'd be, and so few people would be having kids that the majority would no longer bother accommodating them.

    Technically, good birth control is already an extinction level event, the sexbots are a fast extinction level event.

    Now, if every sexbot came with an artificial womb (Wouldn't have to be part of the bot, it could sit in a closet.) and a subscription for monthly supply of gametes, and had a "perfect mother" software package, that would probably solve the problem.

    I get the impression this is the Japanese government's end solution for their population implosion.

    Reply
  43. It is an obvious application for space. And if you build the robots with a human form, it becomes easier to teleoperate them for special tasks.

    A good deal of the problems they're having with rovers on Mars wearing out could be solved by using legged robots with elastomer sealed joints to prevent entry of abrasive dust. Environments like Mars or the Moon are really hard on rotating seals due to dust.

    Wheels have a good thrust to weight ratio, admittedly, and are computationally easy to use. But legs are vastly superior on rugged terrain.

    Reply
  44. 5’8, 125 lbs and slender is pretty much what Austin Powers would call a fembot. Elon was married to a Westworld Sexbot for a time so the application isn’t lost on him. This specific humanoid design is very much also a “Stepford Wives” replacement for women.

    Reply
  45. Spot on. They are even referred to as humanoid. 5G cloud connection allows them to be as much a single entity. Mass produced they’d end up operating other machines meant for humans much as in SciFi, for example literally driving vehicles, pushing lawn mowers, riding tractors. Mass produced they’d outpace manufacturing of dedicated robotic versions of machines. Yet another retro future scene from Elon.

    Reply
  46. The standard Optimus Teslabot can be handed an assault rifle or a Javelin missile and perform well enough for many military missions. There will always be enough Engineers to adapt this tech for military purposes. Tesla will have automated much of the process anyway.

    Reply
  47. Brian’s right about early commercial application, the most obvious is last foot package delivery riding in robotic delivery vans. One he didn’t mention is Space. I bet there will be a dozen Optimus prototypes onboard Artemis Starship when it lands on the moon.

    Reply
  48. He ought to put half a dozen of them on his first Lunar Starship landing. Have them go out and build a dome so that people have a habitat waiting for the next human landing. THAT might wake some people up about the new opportunities in space!

    Reply
  49. "Iron workers" are the people that assemble the steel I-beam skeletons of sky scrapers. It is indeed a dangerous job. "Steel workers" operate steel mills, for the most part they work around the furnaces.

    Reply
  50. Much of the difference in earnings between men and women caused by men being willing to take dangerous and dirty jobs when the pay makes it worthwhile, since men are judged by what they can provide. A man has little societal value and exists in a permanent state of “what have you done for me lately”. Without these dangerous and dirty jobs, men and women will see their earnings become more equal (except where hours worked and years of experience are crucial) but it will come at a price as society tries to figure out what to do with men who lack opportunities to excel at anything that has a monetary value.

    Reply
  51. "to serve and obey and guard men from harm" would be a good mission statement for this branch of the company. They could use a pair of folded hands as their corporate logo.

    (The illustration even looks like Jack Williamson's Humanoids.)

    Reply
  52. Absolutely worth a lot of money. Personally, I would value a cleaning robot even more than our car. But that is just the start. The market is probably close to insatiable for humanoid robots… You could imagine new industries that are simply not possible due to labor costs…

    Reply
  53. The stock market hilariously values the Tesla AI efforts at +0.45% pre-market. It's to be expected, but it's never the less hilarious. Humanoid robots represent a market that is potentially several orders of magnitude larger than robo-taxis, and that market is valued at more than 12 trillion USD by Ark Invest.

    In part, I have to agree that Tesla has not proven anything substantial with humanoid robots, so there is no reason for the stock to go ballistic yet. But come two years (my guess), and Tesla robots are walking steadily in houses, picking up objects..showing the proof of concept. Tesla may have their first commercial product in 3-4 years for limited use in precisely the dangerous situations that Elon Musk outlined.

    Perhaps a cleaning robot for street use?

    Reply
  54. True.

    Now the question is if Tesla will lend their expertise to make war robots?

    May really be bad publicity for Tesla. Tesla is on the side of killing.. But on the other hand, it will be done by someone, so why not Tesla? And having the weapons industry on your side might be really good when trying to twist the arm of the american bureaucracy to make adaptations of launch procedures, rules etc more in line with new technology for Tesla and SpaceX. And we should not fool ourselves into believing that China will not do it.. As soon as they have the capability, they will..

    Then again, there are many engineers that will simply not work for weapons applications, so Tesla may loose 75-80% (my guess) of the top engineers.. Which may not be worth it over all..

    Reply
  55. I think that Elon punted on the question of using AI for production. Surely, you can use ANN's to make visual inspections, plan production flow, predict production deficiencies from sensory data.

    Really, how come Tesla sends out *any* cars with unacceptable panel gaps? Should be an absolute no-brainer to install cameras that continuously monitor gaps other deficiencies as they occur on the production line. And with their expertise it should be a walk in the park to implement this in their factories.

    Reply
  56. It's an interesting change for society. Now, if robots can perform ever more complicated tasks and do it cheaper than any human, does it make sense to import low skill workers at all?

    And will this negate some of the advantage that China has – enormous labor pool – over the USA, or does Chinas advantage mainly consist in the number of engineers, i.e. something that will not be affected by the Tesla humanoid robot in the next 10-20 years?

    Reply
  57. First off, I'm not sure Boston Dynamics uses ANN's at all. Possibly for perception. I am almost certain that they don't use it for planning and motor control. But I am of course open to being corrected.

    This would mean that Tesla probably can achieve better results than Boston Dynamics using their ANN-approach. Furthermore, they can iterate *vastly* quicker than Boston dynamics. Boston Dynamics does not have access to even 0.1% of the compute that Tesla has.

    My prediction is that the Tesla humanoid robot will surpass Boston Dynamics (future) state of the art robot in less than 2 years. The progress of Boston Dynamics – in my opinion – has not been so impressive the last couple of years. At least if you look at what is publicly known…

    Reply
  58. I dunno, it sounds like a slimmed downed production version of ATLAS with Tesla chips in it. Hyundai could easily contract make that for Tesla with Boston Dynamics IP, but Elon being Elon he'll probably insist they build it themselves. Depends on how much Elon dislikes ATLAS' electrohydraulics versus a pure electric actuator.

    Reply
  59. So, Elon Musk wants to build robots that can do repetitive tasks?

    Let me guess, the Tesla Bot will receive instructions via Starlink and Musk will have emergency override privileges.

    Reply
  60. Startling presentation. Opens with a joke Android (human in costume) dancing. It turns out the ambition is entirely serious and even looks like the fake.

    I’d only understood the potential for Tesla’s tech to do this for a few months and expected it to be several years before this came up. Still amazed he actually decided to go for the full on SciFi humanoid robot.

    Quite a jump beyond Atlas. It will be amazing to see Optimus Teslabots doing dances and obstacle courses. Looking forward to Alien Dreadnought GFs including Optimus Teslabot workers mass producing Teslabots. They’d certainly be worth $100k to start and only use 125lbs of stuff rather than thousands of pounds. There is a market for as many as they can make at $100k for a long time.

    Reply

Leave a Comment