Marx claimed that machines cannot "produce surplus value" but only redistribute labor and provide individual firms with a temporary market advantage. Nonetheless, many thinkers across the spectrum seem to believe in a globalized concept of "technological unemployment" or the idea that someday "robots can do all the work."
While machines can clearly alter the nature of work and redistribute labor, can they globally "replace" labor? All other arguments aside, doesn't this ultimately violate the laws of entropy? After all, "life" is the only "perpetual motion machine," and without humans all machines would quickly cease to output "work." Ideally self-perpetuating machinery like the Von Neumann Replicator or Maxwell's Demon remain mathematical and "frictionless."
Peripheral evidence supports this. Obviously, machines are made, fueled, and maintained by humans. Further humans feed, raise, and support the humans who make the machines, etc. Additionally, since the rise of "labor-saving" technology, employed global labor has only increased tenfold. This also suggests that machines redistribute, alter, and subdivide labor globally, but do not "reduce work" in any absolute, quantitative sense.
While this is slightly trickier, I would also suggest, following Maxwell's Demon, that "mental labor" is not free and can only increase on a pyramid of "physical labor." The machine, in other words, only reduces work locally and always contains equivalent "work" as human labor distributed elsewhere in space and time. Even with "free," nonfossil solar power, the most efficient way to transform it into "socially useful labor" is to capture it photosynthetically and feed it to humans.
Isn't this a perfectly plausible argument? Doesn't the common idea of "robots doing all (or most) of the work" violate the second law of thermodynamics? Isn't machine replacement of human labor in an absolute and global sense ultimately fallacious? Is there really a physical "free lunch" inside the labor capacities of machines?
Note: I posted a similar question on Economics Stack, but I believe it requires a broader, more interdisciplinary approach best suited for philosophy. The logical implications of entropy often appear here.
Note 2. I am having a temporary Java-script issue preventing use of "Comments," so may comment or inquire about answers later.