will machines render low-skilled workers useless?

Yes! Well, okay, Ryan doesn’t think so.  You should go read that post and the Greg Clark column that prompted this discussion, although I have a feeling I’m about to quote a lot of Ryan’s entry:

Machine and robotic resources aren’t free; they’re resource constrained just like everything else is resource constrained. We have the tecnological know-how to replace millions of human workers with machines right now, but we don’t because the expense of building, programming, operating, and maintaining the machines is too great. It’s not worth it. As demand for human labour falls, the price of human labour will also fall making the hiring of humans more attractive. Meanwhile, as demand for robot labour increases, the price of robot labour will also increase (since the stuff robots are made of is scarce), making the use of a robot for any given task less attractive. There will then be some market equilibrium which will, in all likelihood, involve plenty of employment for low skilled workers.

This all seems generally correct, but only very generally.  It strikes me that the same argument could have been made about the job security of horses at the dawn of the age of the automobile.  And sure, there are still plenty of horses around, but I think you’d have to attribute that to a pretty generous horse-subsidization scheme.

Ryan lists a couple of constraints on mechanization:

One is the analytical constraint — there are some human cognitive and physical processes that we haven’t yet learned to emulated in machines. This will continue to be less binding over time, enabling workers to potentially compete with humans in a steadily broader range of fields.

Certainly some subset of tasks currently performed by humans will be difficult to automate for a long time yet.  But not only will technology improve, as Ryan acknowledges, but we can simply eliminate many tasks that are resistant to automation.  A computer can’t have a conversation with a person the way that a customer service representative can.  But instead of resigning themselves to this fact, businesses have simply moved to eliminate this class of commercial interaction, substituting for it with websites, IVR systems and brutal indifference.

Then there is the energy constraint. Machines require power to operate, and the more machines we build, the more power they’ll need. This constraint might eventually be overcome, but until then energy costs will rise with the machine share of the labour force, helping to keep humans at work.

I think this is overstated.  It’s a real issue, sure, but it takes energy to run a clerk, too — it’s just that it has to be slowly collected from the sun over a period of months, then the resulting chemical artifact collected, processed, shipped, stocked, sold and prepared.  That cost is going to have to be covered by the clerk’s salary, and on a per work-unit basis it’s going to be a lot higher than burning coal somewhere and shoving the electrons into a high-tension wire.  And this doesn’t even account for the fact that our outrageous social norms demand that the business owner pay to power the clerk even when he’s not making the firm any money.

And then there is the materials constraint. So long as it costs money to build things out of other things, it will be difficult and costly to scale up the machine share of the labour force.

Raw materials seem to be holding pretty steady — in the last few years we’ve seen a noticeable blip in prices of copper and a few other things, and I admit I don’t feel qualified to assess the scare stories about peak chromium/lithium/whatever that arise from time to time. But in general I don’t think there’s much indication that supplies of automation-enabling materials will limit further mechanization.  The raw inputs — metal, motors, plastics, microchips — are already produced at enormous scale for many other sectors.  If demand for automation products spikes, those resources can probably be reallocated without any enormous price shocks.

Besides, raw materials just don’t make up that much of the price tag for this stuff.  A large chunk of the cost is the engineering effort that goes into it, which is a knowledge product and therefore not something we can think of as subject to quite the same constraints as physical materials.  As patents expire and as modularity and abstraction spread, I think the absolute price of automation technology is likely to continue to drop.  You only have to look at the rise of Roomba to see that robots have been getting dramatically cheaper at a rapid clip.  Certainly the burden of proof should rest with those who contend otherwise.

The crucial thing to realize about all this is that the less-than-fully-automated status quo is not proof that we’ve reached an equilibrium.  There are cultural hurdles to overcome, for one thing.  These days if I order a sandwich at Wawa, I do so through a computerized interface.  Could this have been done fifteen years ago? Sure, probably — maybe not with a fancy touchscreen, but pictographic buttons would’ve gotten the job done.  I don’t think that my grandparents’ generation would’ve liked that, though, and I doubt that Wawa wanted to be the first chain to bet on such an approach.

Perhaps more importantly, it’s also the case that the size of the total, cross-societal capital investment that we’re talking about is pretty large.  It seems reasonable to me to expect that it may take businesses decades to finance the purchase of as many robots as they should optimally have.  It’s a complicated undertaking for a business, and in many cases the specific, non-technological design of the systems is still being worked out.  Consider automated parking garages: the necessary technologies — magstripe cards, vending machines, robotic gates — have existed for a long time, but the pay-near-the-exit kiosk setup seems to only have emerged in the last decade or two as the best way to organize it all.  Or look at McDonald’s: they’ve been experimenting with automation technology for at least a decade, slowly rolling out fry-loading machines and the like.  Does anyone doubt that the basic technologies for robotic burger assembly already exist?  But the firm’s got a profitable business to run, and franchisees are no doubt slow to finance these upgrades; there’s no hurry.  It took decades to complete the electrification of the country, and its robotification — which promises gradual economic benefits rather than immediate quality of life improvements — may take even longer.

Broadly, I’m sure that Ryan’s right: there’ll be an equilibrium reached at some point.  But I don’t see much indication that we’re there yet, and it seems conceivable that wages for unskilled human labor at that equilibrium point will be below the threshold that’s necessary for funding a minimally-acceptable existence (some would argue that this is already the case).  I’m no economist, but it seems like weird things are likely to occur as labor costs are converted to capital costs, and an unavoidable shift toward a vast welfare state seems like a totally plausible possibility to me.

The good news is that this can be inferred to be more or less what happened in the world of Star Trek*, and that seems to have worked out okay.

* Though of course our perspective on the matter is distorted by the focus on Starfleet

YOU SAY ARTISTS, I SAY STARFLEET: Ezra weighs in.  Also check the unhyperlinkable comment by “thedavidmo” about how standards of living will be higher in our robot-filled future.  Yes!  And that’s not even counting the benefits to be had by reducing the per capita level of toil.  Not that I’ll be alive for a revolutionary reduction in necessary work, mind you, but I think there’s a decent chance of seeing many businesses move to a 4-day work week, which will leave the total number of hours worked mathematically pristine, but which in practice will mean a whole lot less time spent on (or thinking about) the job.

This entry was posted on Sunday, August 9th, 2009 at 7:45 pm and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.