I think what you are missing here is that we are quickly entering a time where employing humans for almost the entirety of management work will be unnecessary, as will a large swath of low hanging creative work, and in a short period of time, most manual labor. The cost of maintaining a human being will exceed to cost of automation in perhaps as much as 75 percent of the workforce.
How does an economy function when the surplus value of labor is derived from machinery instead of humans?
I know the past answer to this has been that people will just shift to operating or maintenance of the machines, but this shift is fundamentally different because the shift is of mind-power.
We have been able to build machines that can deftly do anything that a human can do, and even replicate this capacity in human morphology. The problem has always been operating these machines in a way that replicates or improves upon human ability. Now that we are building machines that can be trained on existing data and synthesize novel solutions based upon that training, the need for human intervention in automated processes, including in the maintenance and repair of this processes, will drop precipitously.
Let us consider the economics of a robot, assuming a general purpose anthropoid robot that ca be trained in most commercial processes, and presuppose that the mind for such a device will soon be available either by subscription or in an embodied form.
We might see something like this:
Mass produced cost will probably be similar to an automobile, at 20-200x the cost per kilogram of automotive engineering so about 30-300 thousand euros. Service life between overhauls at the indicated forces and speeds should achieve approximately 1 billion pivot revolutions if we assume the surface sliding/rotation movement durability of modern mass produced machines. This would translate to about or 8 billion steps or 1/8 rotation gestures. If we assume 1 such gesture or step per second, that translates into a mechanical service life of about 2.2 million hours, or a mechanical cost of 0.014 to 0.14 euros an hour.
Now let us consider the cost of batteries, assuming that cost per capacity remains flat.
Current production technologies allow for about $100 per kWh of battery capacity, with a 4000 cycle lifespan. If we assume that the robot exerts an average of 500w over its lifespan of 2.2 million hours, the battery maintenance cost of maintaining that output will be approximately 30,000 additional euros in battery replacements. The electricity itself, at .10 per kWh, will cost more at about 100,000 euros over the useful life of the robot.
If we add all of this up we end up with 300k in machinery (200x as expensive as a car, or 1/3 the cost of a fighter jet per kg), 30k in batteries, 100k in electricity, and 3000 months of mind subscription at 100 euros a month per unit, we end up at about 750k lifetime cost for a machine that should be operational for about 250 years if it operates 24 hours day.
Now, let’s assume that those same energy, compute, battery, and mechanical costs are compressed into only 25 years of useful service.
That would be about 3.25 euros an hour. But that assumes that it is 1/10 as durable as modern automotive machinery yet costs 200 times as much per kg, costs 1000 a month in compute costs and consumes 5kw 24/7. -a pretty pessimistic figure.
So we can say that GP anthropoid robots will probably cost between 0.33 and 3.30 euros per hour to operate and is less costly to acquire than a front end loader or a farm tractor. And it will be built and maintained by the selfsame machines.
This is what we are facing.
New solutions are needed.