Tuesday, February 27, 2007

Productivity--but not mine, of course

Over a year ago, I see, I was writing stuff that I've said for years: that "productivity growth is a process of learning how to produce goods and services with less labor." And that's a question of

increasingly automatic production of the means of (increasingly automatic) production.

There's an obvious natural limit to that, namely 100% automatic production of the means of 100% automatic production, i.e. self-reproducing factories. This is my limited version of Kurzweil's Singularity. He goes far beyond that, he mingles it with nano-this and artificial-intelligence-that and general transhumanism in a most entertaining way. He may well be right, but for productivity concerns it doesn't much matter if the self-reproducing factories are the size of cells or the size of aircraft carriers. (It does make a difference in the sort of dangers you have to face and the sort of society you create, but not for productivity as such.)

And this past fall, there was a Scientific American 50-years-ago item that had an image which I remembered from childhood; it was an image of self-reproducing machines, aquatic ones. Hmm...I was four; I don't suppose I read Scientific American then, but it's likely that I came across that issue some years later. Or perhaps I read some science fiction based on it. I dunno, but I guess I've always assumed that productivity, in the sense of output value per input labor for a large range of kinds of output value, was headed towards infinity within my lifetime (except that, as I've said, I didn't actually expect the human race to last as long as it has.)

At any rate, though as a PhuDdy geek myself I have talked a lot about Moore's Law, maybe that's sloppy: what I might call the productivity-singularity doesn't necessarily even stay down with Moore's Law because it's more hyperbolic than exponential. I'm sure this is not an original thought, but I don't remember reading it anywhere; so it's probably wrong. But I don't see why. If you think of the productivity growth rate as the rate at which we learn to produce stuff automatically, then that's a rate at which we solve the remaining problems. Of course we can always think up new problems, and maybe we will, but hey, think about food, clothing, shelter, communication, transportation, energy. You may come up with wonderful improvements in food ad infinitum, you may enjoy putting your own labor into cooking, but some of us will be happy with "Tea, Earl Grey, hot!" from the replicator -- or the self-reproducing robot. Any given production problem will be solved, and once it is solved by relf-reproducing robots then it stays solved. (Until they take over, of course, and decide whether to keep us or not, but that's another thread altogether, where Bill Joy and Ray Kurzweil can argue forever.)

So I expect self-reproducing factories, whether desktops or submarines, to handle most of our production problems and I suppose I expect self-reproducing factories that produce solar-engine balloons (or something better) to indirectly supply our energy, and quite possibly extract carbon from the atmosphere, and so on. And I worry that all these productivities will also supply the opportunities for terrorism; I'm worried about too much "empowerment", not about too little.

Every few months, it seems that the RepRap project is significantly closer to a replicator; and this month I see Cornell has an Open-source Replicator project:

The Altair 8800, introduced in the early 1970s, was the first computer you could build at home from a kit. It was crude, didn't do much, but many historians would say that it launched the desktop computer revolution. Hod Lipson, Cornell assistant professor of mechanical and aerospace engineering, thinks a little machine he calls a Fab@Home may have the same impact.
Well, I think he might be right. (And the Altair was a bit more than twenty doublings ago. But maybe I should stop thinking in Moore's Law terms.)

But the economics profession is still doing its productivity-growth projections, as I noted in the aforecited post a year back, in a linear way. That does make productivity exponential, doubling in a couple of generations, assuming that some magical productivity-growth pixie-dust parameter will keep it growing at the average rate it's been growing at. And this doesn't make sense at all.

Just for the fun of it, I go back to my grandfather's Economics text (Shorey Peterson of the University of Michigan, publishers Henry Holt and Co., NY, 1949.) And he does talk about technology, but it is not so much an information structure for him: he's thinking about increasing specialization on the one hand and increasing capital goods on the other, but he's certainly not thinking about factories reproducing themselves. Well, actually he's concerned throughout with Capitalism and Collectivism;

With respect to the development of productive power and its efficient use, socialists have their arguments, but it is here that the defenders of capitalism can be most emphatic. The sensational progress in production of the last two centuries has been achieved through a private capitalism that, broadly, has been competitive and free of government restraint. It remains to be seen whether state industry, as in Russia, will be more than imitative in its development of methods...
and so on. Yes, it's fun, and I wish he were still around, but I think that the arguments he laid out for undergraduate consumption in 1949 have mostly been settled. And I should probably produce some code.


Post a Comment

<< Home