John Stuart Mill famously remarked that labor-saving machines have not saved one minute of labor. More seriously, Marx argued, roughly, that machines can gain temporary market advantage, but cannot produced "surplus value," which come only from increasing attachment of labor. (Though Marx's value theory is complex and slightly ambiguous.)
Nonetheless, observers, neoliberal and socialist alike, seem to agree that machines are "productive" and "labor-saving" overall. For neoliberals this is productive "innovation" and for leftists "technological unemployment." The idea of a future where "most work is done by robots" seems remarkably widespread. This strikes me as a fallacy for three reasons.
First, in a rough adaptation of Say's Law, someone must obviously make, fuel, and maintain the machines. Someone else must, for example, mine the additional metals, feed the miners who mine the metals, raise the farmers who feed the miners who....etc.
Second, the rise of "labor-saving technology" since the 19th century has come with a huge, five-fold increase in the global laboring population. The "local" increases in per-laborer productivity appear more than offset by the attachment and utilization of increased, lower-cost labor worldwide.
Third, the idea of "robots doing all the work" would seem to violate the Second Law of Thermodynamics. Machines cannot run or reproduce themselves. The only "perpetual motion machine" is life itself. Without humans all machines succumb to entropy almost immediately. This would apply to any level of technology.
I am not arguing that machines make no difference. But is this difference primarily local and conditional? Is the local "productivity" of machines primarily and irreducibly a way of displacing labor globally... and redirecting labor through "machines" from lower-cost to higher-cost zones?
Above all, is it logically possible for machines to produce more "work" in total than goes into them? Again, even apart from issues about "quality of work," I am always surprised at how many people seem to assume this "robot replacement scenario" is plausible.
P.S. Though I would like to respond to answers and comments, the "comment" function is not working for me at the moment, I have asked "Help" about it. My argument from thermodynamics appears to be misunderstood, I will address as soon as possible.