When Technological and Economic Limits are Passed

Brian Wang speaks with humanoid bot and AI expert Dr John Gibb about the surprises in a future where teslabots do everything in factories.

RethinkX projects a 20 year disruption of the global labor market by humanoid robots. RethinkX projects there will be about one billion humanoid robots by 2045-2055. However, this skips over the starting point of the Technological singularity when factories get nearly completely automated.

RethinkX projects humanoid robots will enter the market at a cost-capability of under $10/hour for their labor, on a trajectory to under $1/hour before 2035 and under $0.10/hour before 2045. They projected one billion bots by the 2040s.

I project that one million humanoid bots should be plenty to kickstart the exponential civilization. This means the massive disruption will be clear well before there are one billion bots. The bots need to be able operate and build factories and operate autonomous mining and construction machinery. They will operate factories but also build new factories. There can be unlimited production by constantly and rapidly doubling the number of bots, factories and mines.

We need to create a simplified supply chain of few minerals and metals for batteries and other components. Partial non-exponential aspects like needing humans to build and operate semiconductor chip fabs would not be that limiting on the other aspects that are exponential.

RethinkX says it is impossible to know in advance the full details of how the new labor system will differ from today, but the key feature is: the marginal cost of labor will rapidly approach zero.

Elon Musk described the thousand Starship fleets going to Mars every two years at a SpaceX presentation last month. The fleet can easily transport hundreds of thousands of humanoid robots and all of the factory machinery and mining machinery.

Having no zoning and no Not in my Backyard restrictions on Mars will allow unlimited replication of factories. Production of more rockets, factories and robots can then colonize and develop the entire solar system.

Nextbigfuture Described Truly Exponential Humanity Beyond the RethinkX Humanoid Robot Labor Forecast

RethinkX says it is impossible to know in advance the full details of how the new labor system will differ from today, but the key feature is: the marginal cost of labor will rapidly approach zero.

Elon Musk described the thousand Starship fleets going to Mars every two years at a SpaceX presentation last month.

A million person city on Mars is what Elon Musk and SpaceX have described. They talk about it being sustainable. This means all of the mining and gigafactories need to be there. There must be a simplified human supply chain on Mars. This is the seed that is needed to replicate and build more and more Gigafactories for Starships, humanoid bots and machines.

Twenty-six doublings from the sustainable Mars City and Gigafactories is the 10 trillion bot colonization of the Solar System. This can be completed by 2100.

17 thoughts on “When Technological and Economic Limits are Passed”

  1. All of this is predicated on a robot working how many hours a day between recharge?

    Currently, how many hours of actual working can an Optimus robot operate before requiring a recharge?

    How long does it take to recharge to allow it to work a shift?

    You see the problem? Current battery technology is the Achilles heel of any viable plan to economically displace people. This is exactly the same problem with the EV.

  2. When I was a very young boy, my Dad said something that hit me like a hammer. (AKA: It made an impression) He said, I believe to my question about how machines do certain things, he said: “Bob, machines make machines” Understand, my Dad said this in the 1960’s. He was correct then, and (may he rest in peace), is as correct now. I’ve always believed biology must merge with engineering to give us the best of all possible worlds. Self-repairing surfaces, self-sterilizing surfaces, machines that “improve” themselves each time they “repair themselves” this is how biology works.

    We need to incorporate this “innate sense of self preservation and desire to be better then we are today”, into the current machines and buildings our engineers are now building. This may seem crazy,, but honestly I think it’s inevitable. Shall we move on?

  3. Even if all your rosiest predictions of technological advances prove true, aren’t you being naïve to assume that it will all be put to productive use expanding our wonderful civilization? Seems quite likely that a major use of humanoid robots of some nations will be used for warring against robots (and humans) of real or peceived enemies

    • We are already great at destroying humans and human scale things. SpaceX Starship applied as a hypersonic bomber (mach 20) with 200 tons of payload is far more powerful than billions of humanoid bots. Ball bearings at mach 20 are like high explosives.

    • The sketch of “everything new will be used for war” as an excuse for defeatism, deceleration or flat out Luddism is getting a bit trite.

      It’s technically true, but we already have ways to kill ourselves several times over.

      The potential bad uses shouldn’t stop us from pursuing the good ones, given the concerns are exaggerated and there are other ways to kill us all.

  4. This path of replication was called “clanking replicators” in projections about Molecular Nanotechnology years ago. It’s possible and serves as a causal overdeterminer for exponential growth.

    More likely is that Nanotech machines built to atomic precision will take off in this same time frame aided by AGI. Nanotech robots will achieve the same replication effect just faster, cheaper, better. Living things are a subset of possible Nanotech and the path to molecular scale industry is probably through biotech that steals complex system designs from nature like CRISPR.

    AI seemed like an old technology dream that would take decades or centuries until suddenly it arrived. AGI will deliver another old promised revolution – nanomachine replication and robotics, before clanking replicators get the chance to scale very far.

    The old roadmaps to Molecular Nanotechnology had milestones just recently achieved – including solving protein folding (recently achieved by AI Google Alpha Fold). Dust off the old Nanotech revolution ideas, they are about to get very real and disrupt things faster and more completely than humanoid robots. Of course MNT will make humanoids better, faster, cheaper too.

  5. Humanoid robots, by lowering the cost of labor to near zero, will lead to rampant DEFLATION.
    At first, this seems awesome, as your money can buy more things… but anyone with a huge debt will find that the value of the debt is increasing.
    Many industrialized governments are hugely in debt… this could lead to unforeseen problems since politicians are loath to be the ones to balance the budget.
    The politically expedient behavior would be to introduce legislation or regulations to block or hinder humanoid robot labor.

    • I’d expect the Fed will at least initially prevent price deflation by pumping out money – especially as governments start to see declining tax revenues under deflation.

      Also government bureaus are an obvious place to apply AI to replace white collar workers. There’ll be huge resistance to this at first, but then the Republicans will probably pick it up as an issue, demanding slashing of government spending using AI. So eventually govt costs could fall, while the Fed maintains prices and taxes pull in reventues on the dollar-valued increase in profits. Something similar may happen in other countries.

  6. 1) Some jobs can be completely replaced by robots, but only if their job is realistically and completely definable.

    2) Human problem solving is not easily transferred. No matter how great the (LLM, Neural Net, Fuzzy Logic, GPT, etc.), AI is not equipped for ‘outside the box’ thinking.

    3) Jobs with 90% hard and fast rules (accounting, medical diagnosis, computer coding, customer service, robots in a manufacturing setting, etc.) will use AI to increase productivity, but won’t be able to take humans completely out of the loop.

    4) Unexpected problems always pop up in business, especially on a manufacturing line. Part defects. Incomplete instructions. Breakdowns. Leaks. Poorly phrased requests. Design flaws. And because we always want a newer model, manufacturing lines change constantly, bringing in even more complexity and problems.

    5) The more complex a task, the farther from 100% reliability, efficiency, and transferability you get. Example: 100 lines of code? Easy to completely debug. 1,000,000 lines of code? Almost impossible to debug.

    The biggest question becomes: Given that there will always be unknowns and unexpected problems, how much robot error is acceptable? How many humans will be needed for supervision?

    KIVA robots at Amazon make mistakes, but so do humans. KIVA robots make sense. They perform at a higher level than humans for the simple tasks they perform.

    How complex of a task can humanoid robots handle? If Full Self Driving for automobiles is any indication, then cautious optimism is required. It might take half a decade or longer to get a reliable AI to guide humanoid robots.

  7. The value of almost everything is measured in how much human labor is needed to obtain it. And those that are not can still be acquired by trading or selling things whose value is based on human labor involved. So what remains? Exceptional talent, bankable popularity, political power, energy, resources, land, intellectual property, manufacturing capability (to include robots). I suspect governments are going to have to have a lot of robots if they intend to take care of their mostly unemployable populations (and possibly spearhead off world resource gathering). And governments in places with poor resources, little desirable land, or feeble industry, are going to have an especially difficult time of it.

  8. The human exploration of Mars was a wonderful dream, but delaying our first Mars mission by ~30 years doomed it.
    It’ll likely take about a decade to get to the point of successful Mars landings. Robots and AI will get a LOT better by then, so why send humans? With no humans to take care of, the establishment of industry on Mars will proceed a lot faster, so only a hundred complete robots (one mission) will be delivered. There’ll be plans to return a Starship to Earth, but eventually that’ll be scrapped – they’ll be too useful as Mars orbital shuttles once the space-only versions of Starship come into production.
    A few more missions delivering electronics/sensors, batteries and motors will allow production and assembly of about 1000 robots on Mars, raising Mars’ industrial level enough to make electric motors and batteries. A couple more missions will deliver just electronics and sensors for 100K robots, and a few more missions with other high tech parts will establish robotic manufacturing of 99% of Starship and its launch/landing support infrastructure. Chip making hardware will be delivered shortly after the Mars population hits 0 humans, 100K robots. So only around a dozen Marships will be needed to establish Mars (and Phobos) as the new industrial center of space exploration and exploitation.
    At a guess, no humans will go to Mars surface until the nuclear powered Starship 3 is built in Mars orbit and put in service sometime around 2045. It will deliver rich tourists roleplaying as Mars explorers in an alternate timeline 1995 – modelled on the successful “Moonbase 1985 experience”. Both experiences feature bumbling, wise-cracking robotic sidekicks, in case anything on the ride actually needs fixing.

    • Space is not a friendly place to us large primates.

      Life anyplace off Earth is likely to closely resemble life in a submarine, at least for the explorers and early settlers. Not a promising environment for space tourism. The vast empty spaces onboard the spaceship in the movie “The Martian” was almost certainly just Hollywood being Hollywood.

      I suspect that, other than robots, those folks who actually go there will be cyborgs, AIs, human uploads, or, at the very least, heavily genetically modified beings, all of them spending substantial amounts of their off-duty time in a VR environment.

  9. Ten dollar per day third world labor is cost competitive with robots until you factor absenteeism and labor strikes.

    • Assuming a mass produced robot costs under $25K to make – less than a car – $20K/year leasing seems a very profitable price. Assuming 20 hours a day, maybe 350 days a year (with a few days servicing and other downtime), that’s under $3/hour. I’d guess with competition that’ll fall to $2/hour. And as robots take over and reduce supply chain costs, that should go even lower.

      • Prices aren’t directly related to costs. The prices of humanoid robot labor will be set by the costs of the jobs they can replace. If a robot that costs $25k to make can replace 4 developed world industrial workers at $100k each (3 shifts + one out sick/vacation etc) then it can easily be leased (including all training, maintenance etc) for $250k a year. The user saves $150k/year and can easily do the accounting and make monthly payments with no other costs or complexity. Robots will fill highly profitable niches like this before they come down in price and there are millions of jobs like this.

        • We’ll see. Recent history has innovative companies focusing on very low initial prices (even free) to acquire and lock-in customers as fast as possible, rather than maximizing profits. Competition for market share – and the associated investor and lender willingness to pump money into the company for rapid expansion – is going to be fierce. There’s already over half a dozen advanced humanoid robots announced.

          And initially demand from conservative corporations isn’t going to be all that high unless their labor cost savings are very enticing – like 80% to 90% savings. (They also have to worry about human workers’ fears and possible lawsuits or union strikes.)

Comments are closed.