One hundred years ago 60% of the American workforce worked on a farm. Today, it's .6%. One hundred years ago, there were scientists who openly questioned the existence of the atom. Forty years later it was weaponized, and ten years after that started being used as a source of power. The last hundred years have seen the rise of interchangable parts, assembly lines, division of labor, just-in-time manufacturing, and reasearch and development. They have also seen the invention of the automobile, airplane, telephone, television, plastics, computers, and pretty much everything upon which this country now depends.
The corollary to the expression, "Those who do not learn from history are doomed to repeat it" is this; "Those who do not look to the future will end up stuck in the present". The reason that there are so few jobs in agriculture today as compared to a hundred years ago is that a number of devices were invented to automate the process of planting, tending, and harvesting crops. The same thing happened in manufacturing; devices and processes were invented that made a single person much more productive. It also served to reduce the education required to create a complex thing like a car. Before Henry Ford came along, it took someone with years of training a long period of time to make any single complex item, whether that was a car, shoe, or desk. By dividing production into a series of unique steps, all but a few of the artisans could be cut out of the equation. What few artisans were employed were the ones who helped to set up the system that made the cars, shoes, desks, etc. In a free market society, prices drop as efficiency increases.
One of the best ways to increase efficiency is through automation. Machines are, in general, cheaper than people. The biggest tradeoff is usually in start-up costs. For simple (usually minimum wage) jobs, the start-up cost for a worker meaning training, which is not terribly expensive. For a machine, that start-up cost is usually the expense of purchase and installation, if not also the expense of R&D. The trade-off in price comes over extended periods of time; workers require a constant influx of cash, while machines usually do not (maintenence is usually very cheap compared to wages). This obviously means that people are more flexible than machines. If you're talking about something like making a hundred thousand of anything, flexibility does not count for a whole lot. This is why we use machines.
Of course, machines can replace more than just physical labor. The term "computer" used to be used just like the term "driver"; it meant "one who computes". During WWII, we had whole rooms full of people whose job it was to crunch numbers. You needed to be good with math in any job that even remotely required numbers, because it would be grossly ineffecient to have cashiers and the like spending five or ten minutes puzzling over the price of everything that you had bought. With the invention of the calculator, the baseline for education in a number of jobs fell, just like it had with the invention of the assembly line. The same is also happening now with the written word; handwriting has little relevance to any job, nor does spelling (which almost any computer will correct for you).
I've often been asked the question, "So once machines can do everything, what will happen to all of the people who were working at those jobs?" Kurzweil would say that those people will find new jobs that are created by this automation. I'm not sure that I agree with that. What historically happened was that those who worked on the farms got pushed into the factories, and later on those who worked in the factories got pushed in the service industry. Neither of these require much in the way of education. Let's say that there's an invention that obsoletes a large section of the service industry; the natural consequence of that is that unskilled labor (being in greater supply) would drop in price. The problem with that is that minimum wage is fixed, because of the cost of being a person is fixed (or at least has a minimum). If we have a large supply of people looking for work, employers will choose workers as far to the left of the bell curve as possible. What would happen to the uneducated masses whose labor is too expensive?
There are two options. The first is that they'll become wards of the state, in which case the employers and workers will be indirectly paying their wages anyway. The other option is that the uneducated masses are eliminated. I'm not talking about eugenics; I'm talking about aggressive education.
Let's say that with advances in technology we can squeeze agriculture down to .1% and manufacturing down to 5%. This is not unreasonable. That means that the rest of the nation's workforce would be allocated into management, service, sales, transportation, construction, administration, "defense", advertising, science, and various others. Some of these require extensive training (administration, advertising, management, science) while others do not ("defense", sales, service, transportation). The maxim that I'm working by is "the human can be replaced". By looking at those things that humans are better at than machines, we can chart a general course for workforce outlook in the next twenty years.
The big question, in my mind, is "What the hell are we going to do with all of that labor?". When you make a process more effecient, the waste usually just goes away; if product A initially takes x of product B to make, then reducing that to x-1 means that the price for both product A and product B falls. It also makes the price fall on any product derived from product A falls. If we were in a simple system where demand for product A is constant and there's no outside demand for product B, then this might mean that production of product B drops off. But this doesn't work with people, for the simple reason that labor is always there to be used, and it has a definite floor price. Also, unlike other markets, it is incapable of responding to supply and demand. People won't start (stop) breeding because there are (no) jobs for their children, and even if they did it would take about 18 years for the workforce to feel the effects of that. People do sometimes die because they don't have jobs, but that's pretty rare considering the support systems we have set up.
So where does this excess of labor go? We could likely feed, clothe, and house the entire population with only 20% of the workforce. We can put the other 80% to work doing useless stuff, or creating entertainment, or squeezing even more effeciency out of everything else. This is essentially what we're doing now. More on this later.