It's so ironic this is so often pointed out, since the reality (increasing marginal revenue product of labor) is indicative of something good, not bad.
Technology has certainly already disrupted a lot of industries and increased the marginal revenue product of labor (and lowering the number of people employed), and I think we are just getting started. At some point, if we aren't there already, we are going to be in a situation where our long-term labor needs are substantially less than the supply.
In a sense, this is a good thing for our society overall (and very good for the capital-owning segment of society), but what about those whose labor isn't needed? We're not too keen on paying people to not work.
I'm certainly not advocating a Luddite inspired revolt against technology or remaining inefficient for the sake of keeping people employed, but those who argue that the reduced need for labor is all rainbows and unicorns often fail to account for the human element of the labor equation.
My belief is that it is more likely that as we leverage human ingenuity more with machines, that ingenuity, alongside artificial intelligence, will be freed to pursue greater things. Simply projecting time is correlated with mechanization with is further correlated with job supply decay to Pluto may not be valid.
I do not, at the same time, see a future for un-skilled or semi-skilled labour.
What is falling is manufacturing employment. This is due to a combination of rising labour productivity and a jump up the value chain (which produces fewer, more sophisticated products).
Regarding arms, the US is the largest arms exporter. And it is all built here. That's not just because of export regulation (ITAR), but because that precision manufacturing basically only exists in the US, Schweiz, Sverige, Deutschland, and Russia (plus a couple more).