You're setting up a complete strawman here. Nobody ever said robots are going to be buying stuff, just making stuff. The rest of your argument kind of falls apart once you remove that leg. The claim is that once robots can think effectively enough to displace a human, that human will have extreme difficulty finding work anymore because thinking is the only viable niche for humans left. Jobs won't come into existence to make things, robots will make things. People won't have more effective purchasing power because they won't have jobs in the first place, because robots will be doing all the work.
It's not a strawman, it's
part of why things won't work the way you think.
A robot used for manufacturing
strictly adds value to the economy. I'd say it's a producer, but not a consumer, unlike a human being, but it's not even that. It's a
tool, which enhances the production of human beings.
The fact that a robot has taken over a job a human used to do, does not cause there to be less value, less
wealth in the economy. And in anything approaching a free market economy, wealth doesn't just sit around doing nothing; it is used and spent, so the wealth that
used to be spent paying one worker, is now freed up to pay for a different worker.
And robots will
not be doing all the work. This is, simply put, a baseless assertion, with no supporting evidence beyond speculation about the future. History has shown us time and again, that no matter how many jobs are eliminated because of better methodologies and machines, we
always find something new to do with the freed up resources, and that something new, means new jobs. If those jobs are also highly mechanized and efficient, that means that there are
still more resources floating around, available to make
more new things.
The people of 1700 didn't imagine that trains would make ox-pulled wagons largely obsolete. The people of 1800 didn't imagine that the car would make the carriage a niche luxury good. The people of 1900 didn't imagine that aircraft would make passenger-carrying ships functionally exclusive to luxury cruises. The people of 1950 didn't imagine that calculators would mean that engineers trained fifty years later wouldn't recognize a slide-rule on sight, if they even knew what it was in the first place.
Yet every single time,
every single time without exception, the fact that there is a machine that does part or all of a job better, has just mean
new jobs with
new things that never even existed before came about.
Yes, yes, I've heard the argument that 'not everyone can be an engineer or a physicist.' Yes, that's
fine. You don't need an IQ of 120, or even 100 (though below 90 is probably pushing it) to have practical hands-on understanding of what it means to be a mechanic, or a plumber, or a carpenter. You don't need to be a genius to have specific specialized knowledge in how to operate a couple types of heavy machinery, like a logging machine, or a front-end loader. And there are
continuously more jobs for luxury goods and services being created,
because as necessities become cheaper, people can afford more luxuries.
Even necessities like housing that remain expensive, have more luxuries built right into them.
This also does not account for the fact that the IQ scale has been repeatedly
recalibrated higher. 100 is always the average, so if the average intelligence rises, then they shift the scale upwards. Odds are decent (but by no means certain) that this will happen again in the future.
And as to the argument about machine learning? Let me share something with you:
After decades and decades of AI development, while
some pattern recognition has been managed, a basic reading of things like this will show you that AI does not do well with
meaning. It can be taught the strict rules of grammar. It can be taught many of the weird exceptions and irregularities built in English. That doesn't mean that it can teach itself why you can't 'switch a child to airplane mode.' Exhaustive and highly-detailed programming work from specialists may be able to
manually teach it a lot of the things about what a child is and isn't, but it can't teach that
to itself.
As Doomsought said, you can create and program a machine for highly specialized tasks, but it's very difficult to implement, and each use-case requires different specialization. People thought self-aware robots were around the corner in the 80's (and some decades further back), but that continuously fails to materialize, because
human thought is more advanced and sophisticated than almost anyone understands.
And that's the kind of barrier you need before humans can be outright replaced, rather than just given better economic tools, by robots as a whole.
TL;DR: The 'we'll just keep creating new jobs and new types of jobs' position has two hundred years of evidence of the results of what has happened since the first industrial revolution when things become more automated, the 'permanent 30%+ unemployment' position has 'we speculate because of things in AI development that we see' as the sum total of its evidence.
You do not implement government policy based on a position with the evidence weighted so heavily against it.
Edit: Regarding the 3d Printing thing:
3d Printing is great for specific uses, but what nobody in here except Doomsought seems to understand, is how
ineffecient it is when it comes to time, energy, and sometimes material.
A low-end 3d printer that only works with plastic or resin, will take hours or even
days to create a single toy. And that's a single-cast rigid-shape toy. If it's a complex toy with moving parts, you have to have multiple print runs for each individual part, and then assemble it by hand.
After you find the blueprints, make sure they're compatible with your machine, queue up and execute multiple print-runs, making sure to have the machine kept properly fed for resin/plastc/etc. All of this takes
time, and time is your most precious resource.
Or, you can go down to the store, or hop onto Amazon, and buy an equivalent toy for five to fifty dollars. Depending on what you're getting (like low-quality toy cars) you might get a pack of multiples that combined cost you a dollar or less.
And that's just with one of the
simplest applications for 3d printing.
It certainly has its place, and I'm glad of the modular resiliency that high-end 3d printing adds to our economic system. Some of the really expensive ones can print circuit boards and the like for you at home, and if we have a complete economic collapse, being able to restart the 'building the tools to build the tools to-' process from 80's tech rather than 1920's tech would definitely be a good thing.
Being good at using 3d printers is practically a professional skill in and of itself, and while I expect ease-of-use and operator friendliness to improve as time goes on, that doesn't change the basic economic limitations of it. Some things, like simple spare parts, might shift to the 'just print it' model in the near future, but...
There are
physical limits that mean that mass-production will
always have advantages for efficiency and economies of scale, for many different things. In order to overcome those advantages, we'd have to have technology on the level of Star Trek replicators and the like, which is
not happening anywhere in the near future.