The Value When Tesla Bot Shows Up to Work

Woody Allen said that 90% of life is showing up. Tom Peters said 80% of success is showing up. Tesla Bots will be able to show up better than Boston Dynamic robots.

Boston Dynamics have some agile robots that are able to climb and jump over barriers in rooms. Tesla has a million cars navigating the entire real outdoor world. The Tesla vision navigation is applicable to moving visually around an indoor space.

The utility of Tesla Bots will be higher than Boston Dynamic robots.

Tesla Bots will start with vision-based navigation that adapts to the environment in real-time.

Boston Dynamics uses fiducial markers (signposts) to guide robots around a path. The Boston Dynamic robots are also controlled by joystick controls but they have automated hazard avoidance and crash avoidance.

A Tesla car can navigate from where it is now to another GPS coordinate. There are GPS systems that can accept voice command instructions for target locations.

The Boston Dynamics humanoid Atlas robot is a research and development platform. They have a team of about ten developers babysitting the Atlas bots to make a huge team effort to produce the end product of Parkour videos.

Boston Dynamics does great work but is like Xerox Parc making graphical user interfaces and mouse pointing. It was then Steve Jobs and Apple that took that work and created commercial successes. Boston Dynamics has been working on cool robots for nearly 30 years. Boston Dynamics has been sold and resold many times and now has a few thousand robots and $50 million per year in revenue.

You will notice in the behind-the-scenes Boston Dynamic videos that the robotics experts are talking about solving foundational problems and are not working toward specific high-value use cases.

All of the others working on self-driving cars have Lidar and other expensive sensors and powerful but expensive AI chips. They have tailored their solution to self-driving cars in the $20,000 to $200,000 price ranges. Tesla’s vision-only solutions and custom AI chips can scale down to $500-$10000 robots.

Tesla is able to mass-produce a million units per year. Tesla can make smaller 1-20 kWh battery packs. Boston Dynamics has never mass-produced any consumer solution. Sarcos Robotics has not mass-produced consumer or enterprise products either.

The Industrial Robotic Arms market in the U.S. is estimated at US$929 Million in the year 2020 and the global industrial robot arm market is about $6 billion a year market. An industrial robotic arm can cost anywhere from $25,000 to $400,000. Tesla has over $11 billion in revenue each quarter and this is growing at 50-90% each year.

Steve Jobs also talked about the rigidity and lack of adaptability of big companies like Pepsi, IBM and Xerox. Tesla is able to iterate and change products dynamically every week. Waymo, Huawei, Baidu, Boston Dynamics, Ford, GM, Volkswagen have strategic plans that evolve over the timescale of a decade or more.

SOURCES- Tesla, Boston Dynamics, Steve Jobs
Written by Brian Wang, (Brian owns shares of Tesla)

40 thoughts on “The Value When Tesla Bot Shows Up to Work”

  1. Note the endgame, like Tesla's robotaxi fleet, is not to sell the device, but the service. It's easily foreseeable that TeslaBots will be on-demand leased to end-users, whether they are regular people or businesses.

    You will never get to own one, unless you are filthy rich. The average person will thus increasingly own less things, and lack capital assets to earn for them when otherwise not being used.

    Still a big IF, but if Elon strikes even close, he will be in a very commanding position.

  2. The legislators will be on board the moment they realize self-driving cars can have internal cameras with facial recognition, so that when they put out an arrest warrant on somebody, any self-driving car they step into will lock its doors and deliver them directly to a police station.

  3. Of course, the 'replicants' weren't actually robots. They were genetically engineered clones with pre-programmed personalities and memories.

  4. IF it could do light housework – I'm betting it'll take years of hard work AFTER they get a robot able to simply navigate around any arbitrary home without destroying stuff, to get the robot to do even a handful of useful housework chores in those same homes.

    Yep, a cheaper robot wouldn't be as good, wouldn't last as long – but people buy cheap stuff all the time, especially if a 'higher quality' product doesn't do any more for them, and doubly so if neither the high nor low quality product does a lot of anything yet (i.e. if its mainly a novelty), and triply so if they have an expectation that new and improved versions will arrive in a year or so.

    And my expectation would be that 'cheap' robot technologies would evolve a lot faster than existing, well evolved robot technologies, eventually approaching the quality of the latter while keeping costs down.

  5. If, and it is a big "if", a practical humanoid robot can be built not too long in the future, the advantage to the humanoid form is that it would be able to use all the things already designed for humans to use, and so could be deployed much more quickly. Changing the forms to improve efficiency could come later. I have a feeling that robots as capable as Brian has been touting are a lot farther off than he imagines, so robots tailored to limited tasks probably would be more likely to be successful first, but I don't hold very high expectations for even that any time soon.

  6. Unless the self-driving manufacturers can get such statistics to become common knowledge, self driving cars will be delayed much longer than they should.

    The public and the legislators need to be taught that self driving cars will make different mistakes than human drivers will make, but the self driving cars will make far fewer of them. The legislators need to be on board so they make the laws be such that lawyers cannot go after self driving car manufacturers for typical accidents, unless there is true negligence involved.

  7. Agreed. Especially when such humanoids also can out-think (come up with a larger solution pool-if only as part of web search), remain emotionally neutral (even under duress), are likely physically faster and stronger (if not as dextrous), and have fewer obvious physical flaws. Hard to be without envy and resentment of such mechano-beings.

    I think our mechanical helpers should necesarily be very limited/ focussed in purpose, guaranteed expiry (per replicants in Blade Runner), and not be capable of emotional response/ obvious empathy. Just the facts, ma'am, just the facts.

  8. I think the biggest issue is one that Asimov touched on in the Caves of Steel: Getting people used to ordering around humanoid entities could have bad implications for interpersonal relationships between humans.

    I'd actually prefer that the robots be distinctively not humanoid, for that reason.

  9. In Vietnam the USA got defeated
    In Venezuela the USA got defeated
    In Iran the USA got defeated
    And now
    In Afghanistan the USA got defeated

    Take that!

  10. Well. They're starting to define a certain level of complexity/ awareness that would make certain systems subject to rights' protections. Too simple to call it consciousness or self-awareness exactly… more of a self-determining/ self-preservation threshold… which often is the same level that we like to see in our servile animals.

  11. yes. yes. and could you get one that resembles that bully that tormented you in 4-th grade. Mmm…. With each exploding appendage that mean-sprited Jon Trenton goes down.

  12. well a roomba is one thing – simple system to do a monotonous job, easy to define with a perceived value and cutesy countenance.
    But a robot butler? an undefinable thing that most people would never be fully satisfied with or be continually annoyed by. Who wants a smart pet – we get them to amuse us or condescend to (which some may call freely-given loyalty-affection – but who are we kidding? – we crave the dominance-approval and they want the food).
    That all being said, an irrational population with too much money and too little time may collect and toy with such a product — though likely they may become amusement pieces — perhaps cool to hunt them for sport???

  13. sad creature. Unable to parse the idea that without the US and the top G7 countries this world would be a pre-industrial, medieval back-water as most of eurasia and south america is currently. Envy, jealousy, and covetousness is the last respite of those who have made poor life decisions and backed the wrong socio-political horse. Easier to blame others than fix your own failures, i guess.

  14. Conflicted about this.
    Of course, the human body is a physically inefficient and awkward thing. A sad compromise in all things. Slower, weaker, energy inefficient, high-maintenance parts, and poor weight-distribution for stability -as compared- to all the animals hunting us and being hunted by us – 25,000+ years ago – with our tool-making and community-building, due to recently enlarged cerebral cortex, being the only things that allowed us to leave the savannahs as an intact species. Why propogate this gangly misfit body type? Uncanny valley? disturbing need for the upper-middle-class+ to subjugate a human-like system to its whims and deprecations?
    the more likely body type is a bottom-heavy spider creature/gecko entity with supremely dextrous/ adhesive appendages and highly-adaptable/ sensitive sensors (and perhaps an accessory drone that can lift up to 2 lbs).
    Witness a more likely success in this whimscial adaptation:

  15. $15k would be cheap enough for mass adoption, if it were simply capable of light housework. But I have a hard time seeing the price get that low even with mass production, given the number of actuators needed for even minimal performance, and the complexity of the electronics and software.

    The problem with cheap actuators is that they tend to be inefficient, and not last long. How often do you want it to be recharging, given that the batteries typically have a finite number of charge/discharge cycles?

  16. Probably not, but it's a lot less massive, probably won't kill anyone if it runs them over.

    BTW, Tesla autonomous driving caused around the same number of deaths per million miles as Teslas driven by humans, both about 1/2 to 1/3 the national average.

    Autonomous driving is about 2x better than human driving if you only count deaths where self driving was verified, i.e. not counting unverified claims by drivers or passengers. Gotta figure some are looking to blame Tesla instead of themself.

  17. I doubt they have put in a lot of physical design effort as yet, but they seem to be taking a conventional robotics path, control of fast and precise electric motors with limbs assembled from many parts to insure repeatable control by software without needing to adapt much to the individual robot or to changes over time.

    That will work, and mass production economies of scale will be possible. But even if they invest heavily up-front in a fairly automated factory, it'll likely still cost $15K+. Too expensive for a new software platform with few useful applications available early on. It'll take many years to slide down from 'high end toy for the rich and nerdy' to 'human worker replacement', let alone 'universal necessity of daily life'.

    They should instead start by imagining a robot 'somehow' made by injection molding – as cheap as you could get it – and work backwards to a cheap but viable production process. Maybe cast an entire skeleton in a single piece, to which they can make a few cuts to free it to move. Cheap muscles that are all slightly different but good enough.

    Each robot made that way would be different from every other, requiring software to learn its quirks and adapt to them. It'll be a lot cheaper, but wear out a lot faster and likely be irreparable – costing about the same in the long term, but immediately affordable by millions even if it has only a handful of useful 'apps' initially.

  18. Your hatred to the country that you are living in has turned you upside down. Biden is not going to defeat evil, but evil will be defeated nonetheless.

  19. Concur with the cost idea, however, why do you think that is not what Tesla has in mind? The body seems to be well within scope of current technology. Particularly if using lighter materials and the advances in actuators that Tesla engineers can probably whip up quickly. There are a lot of advanced parts out there and Tesla can buy patent rights (or whole companies) if needed. How much has Asimo etc been held back by having fixed body plans? The challenge IMO has always been power. I expect one of the first things T-Bot will be programmed to do it get itself charged.

  20. And of course it will be connected to Tesla HQ through the upload link, indicated by the red light flashing in its chest…

  21. The thing about a robot with FSD is it's much less likely to crash into things with disastrous results.
    There's still the Twilight Zone aspect of it though, bringing a creepy robot into the house. Would you want that thing around your newborn baby?

  22. Only humanoid robots justify a focus on mass production because they have so many potential uses and only mass production can lower costs and increase quality enough to achieve a robot based economy.

    Elon is doing the same thing with humanoid robots he’s doing with Starship or Tesla Robotaxis. That’s the same thing Henry Ford did with the Model T. Design a machine with wide utility and focus on the machine that makes the machine.

  23. impressive robotic feats of Boston dynamics

    Yet ultimately a toy with limited utility.
    All the promise is in solving AI, this bot is just another flamethrower.

    Tesla FSD serves as a proof

    You might not be aware that they're fully driver-less taxi services in the world, been around for years while Tesla was playing with their toy autopilot. Any decent team and a few 10s of million can produce a state of the art self driving system that can work in a modern and stable town with good weather, but current machine learning architectures are insufficient for lvl5. Tesla is just playing catch-up and is not some pathfinder in the industry.

  24. My belief for a long time has been that the key to getting useful humanoid robots is to get the cost of the 'body' down. Software would improve rapidly in a virtuous spiral (some initial sales create a market for new software that encourages more sales that create a larger market…) if it had an affordable 'body' to run on.

    But apparently that's not what Tesla has in mind.

  25. Tesla has more demand for their cars then they can build, it has been this way for years. They can only ramp-up production so fast.
    Boston Dynamics only sells a robot dog for 75k, not gonna have a line out the door at that price.
    If Tesla can make a humanoid robot for an affordable price, they will sell so so fast, everyone wants a robot to do the laundry, clean dishes, vacuum, dust, clean toilets, wash the car, walk the dog, mow the grass, shovel the snow, etc.

  26. Doesn't Musk's cars have a tendency to crash into stationary ambulances? Placing a dancer on a stage who pretends to be a robot isn't convincing to me. If I were a potential client, I would at least want to see a real robot on that stage, even if it was limited in its current dexterity. Each to their own I suppose.

  27. Boston Dynamics has never mass-produced any consumer solution because there is no mass demand for their products. Tesla does not mass produce 10 million units a year because there is no demand for it. Toyota produced 10 million units in 2019 because there was demand. Mass production history is irrelevant, acquire the demand and you can contract any kind of skills to produce it.

    Always some new goal to get everyone excited and forget about all the old ones.

  28. This could be a game changer for housekeeping and home aid. Mount a dyson backpack with attachments, the robot can vacuum, and other tools for mopping, dusting, etc. For elderly with mobility issues, can fetch a meal (maybe even throw it in the microwave or toaster oven), assist up stairs, from bed to bathroom, etc.

    $30K cost amortized over 20 years @5.615% = $2500/year. Throw in $1500 year maintenance/insurance. So for $4,000/year, a 24/7 maid and home health aide. That would only pay for 200 hours at $20/hr wage & overhead for a human helper. Can use fleets of these Tesla Bots to augment & leverage human help, so each person needs only an occasional check-in. Can really reduce costs/expand assistance.

  29. Not everything Musk does turns to gold, but this one certainly can and is believable. Tesla FSD serves as a proof of where robotics in the real world is going.

    For the linked video. Xerox thieves?

    More like Prometheus, taking fire from the gods to give it to the mortals.

    And so far, the impressive robotic feats of Boston dynamics still have a god-like hefty price tag.

    Or like NASA VR experiments in the 90s and noughties, which cost millions and were out of reach for anyone not on the inner circle. Now you can buy something like it from Facebook for 299 USD.

  30. More of a hype than driverless cars. Don't need all that complexity for repetitive tasks. We already have robots for that.


Leave a Comment