James Douma Interviewed About Tesla Bot and AI Day

James Douma was interviewed by Dave Lee about Tesla AI Day. James is an AI expert and notes how Tesla has done significant re-architecting and is charting new paths to solve self-driving.

James was amazed that one of the Tesla AI modules is using 18,000 amps.

James and Dave think that if Tesla is able to make a product in the 2-3 years that it would have to be constrained domain space. They are hyped by the ambition but think a truly general robot will take time.

Nextbigfuture notes the industrial applications can support constrained domain application.

Dave notes that Elon will be focused on productive, useful and affordable. Boston Dynamics does not have that ambition.

Achieving a $25000 or $10000 humanoid robot would be more challenging than getting to a truly useful and long range $25,000 electric car. It will take mass production.

Tesla FSD and Vision navigation gives Tesla a good head start to the Tesla bots.

This is a multi-decade project that will deflect the course of civilization.

Tesla has transportation, energy and SpaceX Starlink has all communication and now Tesla is going for all human labor plus Mars and Neuralink working on expanding human intelligence.

SOURCES- Dave Lee Investing, Tesla
Written by Brian Wang, Nextbigfuture.com (brian owns shares of Tesla)

19 thoughts on “James Douma Interviewed About Tesla Bot and AI Day”

  1. Sadly the only serious open research currently in the field appears to be being done by none other than our resident commentator Daniel Ravenest actually…

  2. Internally it runs at about 1V as per Musk in the presentation. There is nowhere for the power to go but to heat, so all of it will have to be carried of by the cooling. The difference from 18kA to 15kV may be a rounding error in the 1V figure, plus maybe some thermal inertia and the current not running at full strength all through the cycles.

  3. I've thought about this a lot, being a space fan and a tooling engineer. I think we're closer than most people realize to being able to make self-replicating factories, we've been gradually automating as is economical.

    The biggest two obstacles:

    1) A lot of production related knowledge is proprietary, not public domain. You couldn't build our current industrial infrastructure from just public domain information. Too much of the knowledge is hidden or even exists only in the heads of small numbers of engineers.

    2) Early self-reproducing factories would have to rely on less optimal, more general, production techniques. We're not used to doing that, because the scale of our economy allows for things to be optimized, in a way you can't if you wanted to close a small system.

    But, still, if we made a real Manhattan program scale push, I think we could do it in under a decade.

  4. Indeed! It is only now that *factories* of any kind are being practically considered, let alone self replicating. DARPA NOM4D. There is a similar leap from expendable to reusable. But, expendable is so primitive it should be considered a minor phase of development. Criswell has human made factories making simpler factories, but those human made factories cannot make themselves. Yet! It is getting to be a matter of market choice rather than possibility, as, in a sense, humans collectively ARE a self replicating factory. I say use humans to get going, but only as needed, not as non-paying tourists. Paying tourists are always welcome.

    Muskbots would have to run the O'Neill ISMRU at Mars Station, much more than ISS situation. More the reason to build and test it here in LEO. It then would only have to coat itself with shielding for the human areas and be ready to go. Use it for families. Very easy, compared to Mars surface.

  5. I've long been saying that self-replicating factories are THE vital development for our expansion into space. Life in space requires a much higher ratio of infrastructure to population than life on Earth, because there's no pre-existing biosphere to provide the life support. In most of space, there isn't even gravity, you have to provide that, too.

    With self-reproducing technology, we can economically deploy arbitrarily large amounts of resources per person.

  6. It not only is a limit, it is pushed to that limit, as the speed is pushed the heat goes up. And speed is the only thing that matters.

  7. The potential of neurolink as a mind control device scares me silly. But so does the possibility that, without it, I might eventually not have a mind to bother controlling.

    If it were Google behind neurolink, saying "Hell, no!" would be an easy decision. With Musk behind it, I'd worry more about it being subverted, than I would about mind control being the actual purpose.

  8. More to the point about Muskbots, an ISS style O'Neill micr0g ISMRU set up is clearly a vital part of any self sufficient Settlement or Community at Mars, or anywhere. Musk should make one here, even growing from ISS as Axiom plans, and get it tested out before towing to Mars orbit, where Phobos and Deimos can be used to add radiation shielding and get started. After the unique and artistically valuable properties of the moons are recorded. Solar Sats can go along for the ride. Provide g spin for recovery from trip from Earth. It is the easy part, do it first!

  9. Sounds reasonable to me, too. A high power chip running at 3.3V is going to draw a lot of amps. As a practical matter, the only real limit on logic chip power consumption is cooling.

  10. That is the plan, as few human as needed, to the extent possible. However, we have already started O'Neill with the ISS micr0g experiments that are only now being efficiently performed. It is big time, and the tending humans are there almost by accident, as test subjects for Mars related stuff. So, the BIS plan is as you describe, when at Bennu, but for cislunar we can construct things with avail human LEO help now, vastly faster than machines making machines. 70s O'Neill plans were very short on robots! My deeper point is that the asteroid mining itself makes O'Neill trivial. Indeed, cislunar O'Neill will be before the asteroid mining FALC imagines is done. People *want* to go to Space! To escape power addicts!


  11. Remember when the USAF wanted to build a space station for monitoring the USSR in the 50s? Eventually the technology got good enough that they could do the spying without having to send people up. So they did, and the manned space stations of the future fizzled out. Do you think that they cannot get the technology to mine asteroids without having people around to babysit the machines?

  12. 18 kA × 5 V = 90 kW, with 15 kW of heat rejection, gives an efficiency of 83.33%. It just about checks out, except for the physical dimensions of the tile. It will probably need bars instead of wires, though.

  13. "Fully Automated Luxury Communism" is on the way. Well, the plan is *small world*, as they want to bring asteroid mining material to Earth for use in that perfect state, on Earth. I thot it was a great work, something I wish I had written. Get them thinking about asteroid mining. O'Neill then follows, no commies if you can leave! But they don't seem to know about O'Neill, and apparently are a supernova of planet chauvinism. Asteroid mining without O'Neill????

  14. What to buy first? Neurolink to improve my brain? Or an Elon robot to do my chores? Or an Elon car to drive me to work? Or if my job is taken by a robot, a one-way ticket to Mars? Mind you; I'm not being cheeky here. I like having choices!

Comments are closed.