When I previously posted about Adam Smith, we were discussing the conflation between wants and demand; that is, we considered the potential misunderstandings which would arise from assuming that demands necessarily reflect wants. Now, however, there are some other ideas to discuss.
As of now, I am on the first 15 pages of the book, past the editor’s notes and introductory comments, and into the division of labor.
Smith’s analysis of labor is that specialization is key to economic growth. To divide a general task, like the production of some good, and assign to each person a special task that contributes to the production of the good. The example he gives relates to the production of pins; that when each task required to make a pin is divided amongst men, they can each specialize at their task and outperform those who would perform all the tasks alone. This leads to economic growth in three ways: increased performance, less time, and the application of machinery. This, then, is the idea that specialization leads to economic growth.
There a few thoughts which provoked me to cram the margins of the page with pencil marks, and I want to explore them below. One relates to AI and labor and the other is about signal theory and the risks of specialization.
If Smith’s analysis is correct, which it seems to be so, then it would follow that we don’t want a general artificial intelligence. If the goal of society is to harness scarce resources to supply human wants, then spending time trying to develop a general AI is a waste of not only resources but also time.
Indeed, it would seem that the development of highly-specialized tools that can perform their single function at maximum capacity rather than the development of a general intelligence which must divide its resources is a wiser choice according to Adam Smith. The general intelligence would never be able to perform as well as that hyper-specialized tool, and so it would always be outcompeted. Even furthermore, these specialized machines with programmed duties could perform whatever we please; for instance, we could have them specialize in mental health practices or military strategy. This, might I add, draws an interesting parallel to the animal kingdom.
Humans are considered generally intelligent, we can solve a lot of problems in novel ways. However, as Smith points out, many of our societies are specialized; one goes to school to become an expert in one topic, one’s career is oriented towards one industry, and even our daily duties can be incredibly specific. This same thing is found in other animals; that is, ants divide up the labor into specialist units, Gorillas divide up the labor into specialist units, and so to do bees. Even furthermore, the complex system known as the human brain is also composed of specialist units; there are cells that provide nutrition, cells that fight off foreign bodies, and cells that deal with motor activity. Now, this picture from nature gives us something important to think about.
Are humans genuinely capable of general intelligence? or are there a bunch of functional neural structures that work in accordance with one another, giving the illusion of a general intelligence. Perhaps we’ve always been specialized units, and perhaps all things move towards specialization because specialization allows for a more proficient manipulation of the resources in a given environment.
Whether this is true or not requires plenty of empirical research, not only into complex systems but also into intelligence. At any rate, whatever the answer is, it certainly flips the general intelligence paradigm on its head.
To not make this post to long, I will post my thoughts about signal detection and information theory, and what these fields can tell us about the dangers of specialization, at a later date.