Competition necessitates humans

Progress in the Trump Administration has been slow yet erratic, save for some talks regarding Korea. There are these huge, canyon-wide gaps between anything important happening, that the odds of missing anything by not following things, are minuscule.

The question, will technology eliminate jobs–and if so, how many–frequently comes up on in online discussion, especially on sites such as Reddit, Hacker News, and Slate Star Codex. There seems to be this pertinent and what I deem to be irrational fear that, if given enough time, technology will destroy all jobs. Maybe it will–it’s impossible to prove that it can’t happen–but the overwhelming evidence shows that to be unlikely, if not impossible. Despite increasing automation, the size of the labor force has not shrunk. That, and other reasons are discussed in Technology and Job Loss: The Debate Continues. But there is an additional reason which didn’t occur to me until only yesterday.

In terms of microeconomics, competition necessitates humans. A firm that relies purely on automation is at a disadvantage over one that adopts a hybrid approach. The reason being, a purely autonomous system cannot adapt to changing environments brought on by competition. Either a competitor will exploit a flaw or blind spot in the automaton or the automaton will encounter an obstacle it cannot solve. Without competition, this is not a problem but the introduction of competition requires innovation, hence humans. A firm that cannot adapt will succumb to one that does.

This is especially true for e-commerce. Consider there are two sellers on Ebay or Craigslist (or any online marketplace that allows users to submit listings) , which we’ll denote as ‘A’ and ‘B’ that are competing with each other. The more goods they can list for sale, the more money they make. Initially, they list the goods by hand, but this is time consuming and repetitive, so both of them invest in software that will automate the process. Or what happens is, competitor ‘A’ notices that ‘B’ is posting goods far faster rate than would be possible by hand, so ‘A’ surmises that ‘B’ is using an automation tool, so ‘A’ follows suit. Thanks to the software, because ‘A’ and ‘B’ are able to post more total goods, they both make more money. However, there is a point of diminishing returns. Most people do not browse beyond the first or second page of most listings. Thus posting more goods is of little use if they are buried where people cannot see them. One of the employees of ‘A’ notices that profit have stalled despite more listings. He also notices that there is a feature on the website to leave feedback and ‘likes’ and that listings with higher feedback rank higher than products with lower feedback. Armed with this new information, the employee decides to cheat by leaving ‘likes’ and feedback on his own listings. It works, and his rankings improve. He then creates a software program to automate the process. Although ‘A’ and ‘B’ are posting more total goods, ‘B’ is being crowded out by ‘A’. Either ‘B’ adapts, or ‘B’ fails, or the website becomes privy to the deception and penalizes ‘A’. The first and third choices require humans.

The software, no matter how advanced, cannot exceed the limitations of its programming. It can only do what is is programmed to do. The program that posts listing cannot make the ‘leap’ to observing that ‘likes’ and feedback are positively correlated with rankings, or even what a ranking is or why rankings are important. Even if it could ‘think’, its thoughts would still be limited. This does not mean emulation of the human brain and the complex emotions that arise from it, is not possible, but such an emulation would need to programmed to have a purpose, just as ‘A’ and ‘B’ have purpose, and then from that purpose or goal, and with pre-existing knowledge, derive a solution. That seems far beyond the feasibly of any technology known.