"How Asia Works"  was an interesting read because it gave me a new perspective on state controls in nurturing a developing economy. Joe Studwell's central thesis is that governments in developing countries must follow a set pattern to speed up progress : agriculture (with land reforms) → manufacturing (with export discipline) → financial policy (with incentives for technological progress and long term gains).

My main point of contention is with the specific role of manufacturing in this setup. Studwell claims that even today, a developing country must go through the rigors of scaling their manufacturing employment through a 1-2 process:

  1. Purchasing technology which is easier to use than to create.
  2. Use the cheap labour (high supply in developing countries) to generate exports which are competitively priced with the rest of the world.

This theory absolutely has a strong basis in history, as the successes of the East Asian powerhouses have repeatedly demonstrated. My concern is that the historical context that catalyzed this process — when humans in the loop could boost the productivity of machines — is rapidly diminishing.

Studwell claims that scaling manufacturing is vital because it can employ (1) a large population; and (2) an "unskilled" population. However these apparent advantages dangerously complement each other in an unsavory way - "unskilled" factory jobs are likely to be the first to be eliminated by automation.

In my opinion, machines in manufacturing are charting the following path:

  1. No machines : Humans doing tasks themselves. This has been the state for almost the entirety of human history.
  2. Specialized Machines + Human Operators: At this point, humans design machines to be exceptional at specific tasks. More humans operators ⇒ more machines can be put to use ⇒ more output. In this world, one scales their output by buying more machines and hiring more human operators. This is largely what we've seen in the 20th century, and this seems like the worldview for the author's suggestions. I believe this age is already over.
  3. Specialized Machines + Interfacing Machines: At this point, humans have created protocols for specialized machines to communicate with each other. For manufacturing a product on this pipeline, a human lays out plans for how the output of one machine flows into the input of another machine with the aid of the interfacing protocols. The operational need for humans in this manufacturing process is vastly reduced — largely to adhoc tasks like monitoring for issues. I believe we are only just getting started in this era.
  4. Abstract Machines: The ultimate goal. At this point, a human need only give a specification for "what" they want (eg designs, counts, tolerances) in a declarative manner. The abstract machine will automatically create the right combination of specialized machines and interfaces to accomplish that goal in a manner that provides the best bang for your buck. We are already seeing some early signs of such machines in the (super small scale) world of prototyping - eg. 3D printing.

Scaling up humans in an industry which is actively making leaps and bounds of progress in efficiency by eliminating the humans in the workflow is backward-looking and shortsighted — and in today's world, any new policy that prioritizes manufacturing to generate large scale, long-term "unskilled" employment is nothing short of irresponsible.

Then how else can we create a system that intrinsically supports large scale employment? I don't know yet. I hope we figure it out eventually