Thursday, September 22, 2005

Your and Your Robot Overlords Make Me Cry

Machines are not going to replace human labor anytime soon. Why?

Long Answer: The Solow Growth Model.

Short Answer: Human labor is really goram cheap.

Long Answer Redux: I don’t know if Ezra or Neil took much economics, so I can’t say I blame them. But positivist lackeys at TechCentralStation have no excuse.

The SGM is, right or wrong, the fundamental assumption for how economic growth happens. Basically, the amount of consumer goods a society produces is its labor times its capital:


This theory supposes that when you add more labor, they don’t try to just make goods by themselves, but they get some of the machines other labor was using, and spread it amongst themselves. The efficiency of marginal labor is directly proportional to how much capital there is. Conversely, when you get new machines, you spread out the workforce to operate them

Now the model can be more complicated than this. L and K have exponents for diminish returns, there’s a term for savings that shows how fast K grows over time, a term for technology improving the efficiency of K, and a way to model a stratified labor force (some labor who do low skill stuff, some labor who make more technology and capital). But that one equation is really what it comes down to. That, and the realization that all technology is a way of making K more efficient.

What’s the point? We could have replaced almost all of humanity with machines a long time ago. There are many many many human jobs that could be done by machines if we wanted to spend enough money on them. More farming machines, more automated factories, more self-service machines at grocery stores, etc. It’s just the more humans you replace, the more expensive it is to replace them. And even when you come up with a design for a machine to do a human job, that doesn't mean it's free to make.

And you know what? Most human labor is really cheap. Factories in Taiwan, customer service reps in India, and teenagers and ex-cons who bag your groceries at CostCo – all are things that could be 95% replaced by capital and technology, if you wanted to spend ten times as much per job.

Now as technology progresses (and I’m an economic renegade who believes the technology growth constant is convex instead of concave – ie, even more wild-eyed about technology than I should be) and savings increase, it will become cheaper than it previously was to replace those workers. But it will be a very long time before it’s actually cheap enough to replace them entirely, and it will have a lot more to do about the economics of the world and the global savings rate, than it will about some new AI code.

To be clear: I am not making the argument that there will always be a role for educated people to run and train the machines. Educated people may be harder in some situations to replace than a low-skilled labor, but certainly not absolutely. Spend one day as a research assistant, and one day as a plumber, and tell me which you think will be automated first. It's a matter that there are still a lot of cheap humans are out there and will be for a long time.

PS: I know Ezra just started this by talking about Singularity a new hip book, but far too many of his comments and other technology-triumphalists showed the predilection to believe that mass-replacement of much labor was right around the corner.


Post a Comment

<< Home