AI's Dual Role in the Labor Market

How one man’s complement is another man’s substitute

Image generated by DALL·E 3

We don’t have enough information to predict what the effect of AI will be, but anyone can understand the framework for how AI might impact labor markets. Examining the mechanisms for how technology either augments or supplants human labor reveals that these differing forces can actually be two sides of the same coin. While often lost in the discourse around regulating AI, sometimes “one man’s complement is another man’s substitute.”

Economists first think of technological advancements as impacting the demand for specific skills. AI, like all technology, can either be a complement or substitute for one's labor. If a technology increases demand for a given skillset, we call it a complement. If it reduces demand for a given skillset, we call it a substitute. We know these concepts intuitively. Google Maps is a complement to my productivity, reducing the costs of me getting to work by giving me the fastest route. But it is also a substitute for physical maps and reduces demand for the labor of physical map makers. 

What can we say about the impact of AI though? A new working paper on “The Short-Term Effects of Generative Artificial Intelligence on Employment” investigates the potential for ChatGPT to either complement knowledge workers by increasing their productivity or substitute for them. The authors study freelancers on a large online platform, Upwork, before and after the introduction of ChatGPT. They compare those occupations they expect to be most affected by ChatGPT (writing related services) with those least likely to be affected. Their analysis reveals that “more affected occupations experienced a decrease of 2% in the number of monthly jobs and a decrease of 5.2% in monthly earnings,” as a result of the introduction of ChatGPT.

 While only one paper, the conclusion seems bleak. Online freelancers that were expected to be negatively affected by AI did indeed see a reduction in the demand for their labor. However, our interpretation of these results needs to keep in mind a basic economic insight: it is because new technology confers benefits to some that others are hurt. In other words, it may be that AI hurt writing related freelancers in the short term directly because it benefitted their would-be buyers. Based on this mechanism, the conclusion of this paper becomes a little less bleak and the policy implications become more unclear. 

Let me explain. As both a user of ChatGPT and Upwork, I have personal experience with how “one man’s complement can be another man’s substitute.” Over the past year, ChatGPT has been transformative for me, especially in learning Python. It has not only tutored me in the language but also assisted in refining my code and creating entire scripts. This enhancement in my productivity has been remarkable. Where I once depended on outsourcing tasks like this through platforms like Upwork, I now find that ChatGPT fulfills these needs more efficiently.

So, am I partly to blame for the reduction of employment and earnings on Upwork? Yes, as are many others like me. Because ChatGPT can provide what I need more efficiently, my demand for sellers on Upwork has been reduced. It is because AI is a complement to my production that it substitutes for the sellers on Upwork. If ChatGPT ended up being less efficient, then my demand for Upwork sellers would have remained the same. 

With growing interest in regulating AI, it’s important to keep in mind how the benefits of AI to some are directly related to the costs others experience. AI's complementary role in one area of the labor market can lead to it substituting for labor in another. This duality is not just a theoretical concept but a practical reality, whose implication should be clear: it’s hard to forgo the costs of AI without also forgoing the benefits. As we navigate this evolving landscape, recognizing the intertwined effects of technological advancement will be key to the ensuing AI policy debate.