Teslabots Revealed

Tesla revealed a crude Teslabot that was built in 6 months using mostly off the shelf parts. This first version is able to walk and is training to perform tasks in the Fremont factory. Tesla revealed a more advanced bot that uses custom Tesla parts and was developed with Tesla car manufacturing experience. It is more streamlined and is designed for mass production in the millions.

Teslabot will have 2.3 kilowatt hours of batteries which is enough for one day of continuous operation.

They already have it picking up objects like watering cans.

I think Teslabot could be about 12-30 months from having significant financial impact for Tesla.

They are looking at an initial mass produced cost of less than $20,000. this would be $2.50 per hour of labor if it operates 7X24.

They want to get the more refined version 2 walking and performing better than the first robot for performing tasks in the next few weeks.

They will focus on a factory use case and want to nail it to prove usefulness in the real world.

43 thoughts on “Teslabots Revealed”

  1. Sorry burst your bubble guys, but does anybody actually believe this will ever work as advertised? Note, it’s one thing to build a robot that can walk on two legs (it’s been done before) and to build a robot like that which can simply replace humans on a factory floor (which is orders of magnitude harder that anything accomplished to date with AI.)

  2. Not gonna lie, I’m a bit disappointed. These robots still look very far from the ones in the movie “I, Robot”.

  3. Why did they make it dance? Surely it only invites negative comparison to the frankly groovy Boston Dynamics robots?

  4. Hands will be the major issue to solve – the worm drives shown are too weak and wear-prone. Our finger muscles are in our forearms, and trying to match their performance and durability at any reasonable cost may be nigh-on impossible.

  5. Are we sure that this is the best shape for functioning in various human environments and becoming the desired butler, house/ property keeper, assembly-line worker, retirement residence helper? It can’t/ shouldn’t be primarily because of a being a familiar and non-theratening bi-ped form. I think Boston-Dynamics has it better – low and spread-out – maybe even better as a spider or salamander-shape per MIT.

    • No, Elon has already said this is not the best shape for functioning in any particular environment, but it is in fact the best because every environment we work in has been shaped with the ergonomics of this shape in mind. A low-to-the-ground spider-form robot may not be able to climb helical stairs inside a factory, or maybe a ladder up a cell phone tower, but a robot which is shaped like a human can. This is the point that Elon has been stressing repeatedly.

      Otherwise it would have been much better to make a hexapod, or even wheeled, robot, with a special “power tool USB” connector so it could swap out working bits. But then there would have to be special stairways and special tools for these robots to use, they couldn’t just inherit the tools for human use. This is why Optimus is human-shaped. It has nothing to do with being non-threatening, or it would be shaped like a Roomba.

      • Yes. A broad class of robots that are essentially humanoid is important.

        Aside from making them more adaptable to a wide variety of tasks, being able to build new tech so that it works with old and existing environments and tools is essential for rapid adoption.

        We have people living in homes built before the American Revolution that have running water, flush toilets, electricity, central heating and air conditioning, internet, and wi-fi. They may also have refrigerators, electric or gas stoves and ovens, blenders, food processors, dishwashers, clothes washer (and dryer!), and garbage disposals.

        If use of these things required the construction of entirely new houses, or radical redesign and renovation of any preexisting house, it would have severely limited the spread of these technologies. As it is, we might rent a cabin in a remote mountain range and be somewhat surprised, or even dismayed, to find it lacks a dishwasher . . and heavens help us all if it lacks broadband internet.

        My cabin in the mountains is rather remote, with very few neighbors down my road through the trees, yet somehow it has all of these things, and even a sewer connection (which boggled me when I learned of it–again, because it is really remote). I wouldn’t mind having a robot to keep the place clean inside, blow away the leaves, shovel the snow, and maybe even chase away the bears, but a Roomba could never handle it.

  6. Gary, the only multifunctional robot that is already in production is superior. Gary’s Software is a platform for Developers to develop their own code to operate Gary. When it comes to multifunctional robots the Musk does not have the same advantage of being the only one in the field as he had when he started developing luxurious electric cars and reusable rockets.


    • Is that robot *really* in production? The year-old article you linked to mentioned there was still a round of funding to go through, and those things rarely complete on schedule.

  7. It’s what happens when a stores runs out of sales! Not a pretty sight. I think there’s a mathematical formula for it:

    N = R * fp * ne * fl * fi * fc * L

    • And my reply went to the wrong spot. So now it just seems out of place haha. Sorry Brian/Doctorpat/Snake Oil Baron

    • The equation for working out how many sales staff in the galaxy might be intelligent.

      It all makes sense now.

  8. This bot and its various technologies will be a boon for the disabled and housebound. The prosthetics and remote control possibilities are huge.

  9. My predictions for the 2023 AI day:
    Telsa will show a robot that isn’t really at risk of falling, but if it is deliberately tripped and knocked down, can catch itself to avoid damage and get back up.
    It’ll walk stiffly onto the stage and then start dancing almost as fluid and fast as the human in a suit from last year.
    It will do some task in with a conventional factory robot, where it’d be too dangerous to have a human working.
    Then half a dozen more will suddenly walk on stage to demonstrate that they’re progressing on manufacturability, and they’ll work together to lift a piano.
    Then Elon will make a joke about them making the factory robot obsolete, and they’ll work together using power tools to disassemble the factory robot.

  10. Meanwhile most stores have people spending hours adding and removing sales rages from store shelves because no one wants to be bothered investing in electronic shelving signs—which already exist.

    • I’ve seen stores with e-ink price markers on shelves. Not clear if they are long range wirelessly updated like an IoT device, or you have to get up close with a RFID reader/writer to change the display by wirelessly powering it with the handheld device.

  11. Lot’s of significant announcements. I didn’t expect this much progress on a mass production ready version this soon. Should be vast advances by AI Day 2023 with hundreds of these at work in Gigafactories next year.

    7 Exapods. Confirmation of planning for next generation Dojo. Architecture for unlimited scaling.

    • We may just be looking at the leading edge of a singularity, finally. Enable these bots to do enough factory tasks, and you have fully automated factory. Fully automate enough different types of factories, and you have a self-reproducing factory. And while the Teslabot is designed to be general purpose, there are some tasks out there, like agricultural labor, that would affordably use an even cheaper less generalized version.

      The future is dire for anybody who can’t do useful things a Teslabot can’t master. And that’s a moving target that will eventually encompass more than half the population.

      But maybe Musk has already thought about that, with his neurolink upgrade path for humans?

      • Brett, the singularity is about AI exceeding human intelligence and not about building widgets faster and more cheaply.

    • I think it’s intentional, Part of why Musk is doing robotics, AI and neuralink himself is to prevent an I-robot or terminator type situation.

      part of his risk mitigation strategy here is to make mass produced household robots weaker and or slower than the average person to prevent them from being used to harm people or commit crimes.

      If you think that’s too much of a trade off, may I point out to you several pieces of automate equipment I use on a daily basis, a tumbler and ultrasonic cleaner.

      both of these machines are not faster than doing it by hand, but because they don’t require my attention when doing their job, I can do something else instead.

      I am more productive because of a slower, yet fully automated machine, the same goes for any robot
      sure it may take 30% more time to do anything, but YOU won’t be doing it, and if speed is important, then get a second robot.

      • Terminator, shmerminator — that’s fantasy. Why would some machine intelligence need anthropomorphic robots to hunt down human beings? All that would be needed is a virus deadlier than COVID. Killer robots are only for dazzling visuals on the cinema screen, in contrast to silent-but-deadly viruses that kill far more efficiently.

    • Do you realize that 10 years ago, Honda had been working on Asimo for 26 years – i.e. since 1986? And they NEVER figured out how to make it useful for anything other than PR stunts, nor did they focus on making it mass-producible. They finally killed the project this year to focus on more useful robots.

      • Honda never had a coherent approach for integrating the Asimo platform into an actual human environment.

        Walking up and down stairs, giving flowers to little girls on stage and greeting was as far as it went.

        While Tesla has at least one critical additional asset: lots of training data and experience enabling robo-cars to live in the real human environments on actual streets.

        They will leverage that experience to make their bot much more flexible and knowledgeable than Asimo soon. They have never done so in actual homes, but the gradual approach of adding up millions of hours of experience into the latest iteration’s training will remain.

      • No, I expect there was a passive armature in it to keep it from being floppy.

        OTOH, we’d better expect the first ship to Mars to be carrying Teslabots, and a high bot to human ratio to be maintained on Mars. It solves a lot of issues if they can do that.

        • I assume it will be the case.

          It would be a huge waste to send a rocket without some way to roam the corridors, test the elevator, and get someone to inspect it from the outside.

          My huch is the first Mars Starship will have several ready for deploymeny after landing, if they polish biped walking and objet manipulation before that. It’s still several years to that flight and they do have time to get it ready.

          • In the course of trying to come up with some flippant remark, I came to a serious conclusion.

            Robots designed to operate on earth will have:
            – Circuitry designed with air cooling
            – Joints and moving parts vulnerable to freezing lubricants and vacuum welding
            – Parts vulnerable to dust
            – Batteries and components designed for earth surface temperatures

            Put that together and your robots will function best on Mars if they are wearing human spec space suits. And will “die” without them. Albeit slower than a human would.

            That’s… an interesting issue that I’ve NEVER seen in science fiction.

            Of course you can, and people have for 50 years, design robots to operate in a space environment. But an Earth optimized robot is like an earth optimized lifeform.

            • I’ve actually seen that addressed in a webcomic, “Freefall”. (Which is hilarious, but has a huge attention to details.)

              The robots are, except for specialized industrial ones, weak enough a toddler can beat them up, on account of being built as cheaply as possible and still being able to to their jobs, and, yes, most of them can’t take vacuum.

              Robots on Mars or the Moon would absolutely have to incorporate something like a simplified space suit, if only to prevent migration of abrasive dust into moving parts.

            • I was more thinking about some minimal changes compared to the Earth version to allow them to make their job before breaking down. They will need these robots for some actual work on Mars after all, and they won’t be expected to last beyond a few hours outside, taking pictures and returning via the elevator if it can.

              Like basic vacuum proofing it and a few joint heaters (yeah, that actually sounds like some extra years of development).

              But a robot in a human-like spacesuit would be definitely be more funny and probably clever.

              • A much simpler suit, as it doesn’t need to provide breathable air, remove humidity, deal with liquid wastes. And a robot intended for such use could be designed with joints that had room for those special volume constant suit joints.

                Add a bit of liquid cooling for the energy intensive components, and you don’t even need much pressure.

Comments are closed.