On Job Loss Due to Automation
The reason we think about what humans do better than machines, and how we can avoid having our jobs taken away, is that humans work for income. But in the long run, I believe humans should not work for income; rather, they should work in order to accomplish something they themselves find valuable. (A piece written at the time of the 2016 “AlphaGo vs. Lee Sedol” match about the problem of job loss due to automation)
Is this something I should worry about in Korea right now?
Before starting, it would be good to think about whether this is really something that I, right now, in Korea, should be worrying about.
I often come across the claim that since Korea does not have the technological capability to create AlphaGo, there is no need to worry for the time being. But the lack of technological capability is not very important.
First, while it is difficult to create a baduk algorithm that defeats the world’s best professional player, it is easy to obtain an algorithm that defeats 99% of humanity. The same is true of a considerable portion of knowledge work.
Second, it is not even necessary for an algorithm to be better than I am in order to replace me. For example, deciding which products to place at the top of the screen in an online shopping mall was traditionally done by people. But if you pick up even a statistical theory from 100 years ago and make a simple algorithm, it generally produces better results than humans. Why is that? There are roughly two reasons.
- For one thing, unlike humans, an algorithm does nothing but that job 24 hours a day without rest. Moreover, an algorithm can carry out rearrangement once every minute.
- Another reason is that using an algorithm makes personalization easy. A human worker arranges products for a single screen viewed by all users, but a personalized algorithm creates screens for individual users or user clusters and arranges products differently for each of them.
The reason companies are not automating immediately is not that they cannot obtain good algorithms. It is simply that chronic problems—inefficient organizational structures, inefficient communication, inefficient performance measurement and compensation, the ignorance or laziness of those in charge, office politics, inertia, absurd government intervention, and so on—are preventing it from being pushed forward or are making it proceed extremely slowly.
Nevertheless, whether slowly or quickly, it is certain that this is already a phenomenon in progress.
Is it enough not to compete with robots, but to cooperate with them?
One HBR article1 proposes not competition with robots but cooperation (more precisely, augmentation). I am also trying to think in that direction.
But regardless of whether the productivity or employability of individuals who cooperate well with robots or AI temporarily increases, the average employability of the members of society as a whole cannot help but continue to decline. If a few individuals achieve dramatic productivity improvements through cooperation with robots, the result leads to even more job losses.
If production efficiency rises sharply in a situation where purchasing power is concentrated in a small minority, then in the end production volume must be reduced, and that in turn means additional job losses. Some of those reduced jobs will in addition go to robots or to humans who are good at using robots. It is a typical positive feedback loop.
This is a problem that cannot be solved unless we escape the structure of a zero-sum game in which, for me to do well, someone else has to lose.
Universal Basic Income
Put together, it looks like this.
- Creating an algorithm capable of replacing my job is easier than one might think, and
- regardless of whether a particular individual’s employability rises, a considerable number of people will lose their jobs.
People in the technology field who think along similar lines seem to be considering a basic income system as one solution to this.2 Perhaps because I too have been in the IT field for a long time, I think similarly. In particular, there are two reasons I support a basic income system.
Labor for value
First, I believe that humans should not work for income, but should work in order to achieve something they themselves consider valuable. The labor we are doing now is alienated labor. If a basic income system is introduced, it becomes possible to separate labor and income to some extent.
The weaker the social safety net, the greater the share of work done for income, and because of this all kinds of side effects arise. People try to become doctors not because they feel a sense of mission in benefiting human beings through the art of medicine, but because it is the profession through which they can earn the most money. The same is true of politics, administration, the legal profession, business, and most other occupations. As for how enormous the individual behavioral changes arising from the difference between the former and the latter, and the resulting social changes, would be, let each person imagine it for themselves.
If there are oxen, humans can plow fields more easily. If there are automobiles, goods can be transported without humans or horses moving them. If there are algorithms, humans do not have to make repeated decisions. The development of technology is consistently moving in the direction of separating labor and income. Only human preconceptions and the inertia of the existing social system have not yet separated the two. If all people were regularly provided with a certain income regardless of labor, people would be able more easily to separate labor from income and connect it with value.
If this shift in awareness takes place, the importance of worries like “What is it that I do better than robots? How can I raise my employability?” will diminish. Instead, people will ask themselves, “Regardless of whether robots do it better than I do or not, what is the work that I regard as valuable?”
One criticism of basic income is the claim that if income is paid to everyone unconditionally, people’s motivation to work will decline. That may happen. But in the context of the present discussion, that is not a disadvantage but an advantage. If motivation declines because of basic income, that implies that the person had not been doing that work for its intrinsic value, but had been doing it, to some extent, for income. In a world where machines do most of the production, it is not a bad thing for humans to lose the motivation to do alienated labor.
Weakening the rich-get-richer, poor-get-poorer dynamic
Second, a pure form of market economy is self-destructive. One of the most conspicuous phenomena is the rich getting richer and the poor getting poorer. We must guard against this through various measures, such as increasing inheritance tax, gift tax, and capital gains tax, and strengthening all kinds of welfare. Basic income is also beneficial in that it helps the fair redistribution of wealth.
Ah, given the context of this piece, automation tax should not be left out either. For example, it is the idea of levying a tax on a company when it terminates employment contracts due to automation.3 I am in favor of it. Originally, creating jobs was one of the social obligations of companies, and they have received the benefits of support policies on that basis, so I think that if they reduce employment they should bear corresponding responsibility.
Since it has come up, to add a little regarding the inheritance of property: I see political power and money power as both the same in the sense that both are forms of power. If the divine right of kings and the hereditary succession of power that follows from it are absurd, then likewise the inheritance of property is also absurd. On this, refer to Russell’s view on inheritance of property.