How will the development of AI affect the labor market?

AI and the Economy

How will the development of AI affect the labor market?

Read a summary or generate practice questions based on this article with the new INOMICS AI toolhere.

Artificial intelligence appears to be the next truly disruptive technology poised to rock our global economy. In December 2022, ChatGPT-3 made headlines. It represented a leap forward in the capabilities of generative AI technology, pushing conversations about this new tool into the mainstream as people began to question how ChatGPT will affect the economy.

It’s expected that AI will change the economy and the world, though few have answers as to how. News stories continue to be written as companies, governments, researchers, and students grapple with AI and learn how to adopt it or respond to it.

In this article, we’ll explore this question from an economic standpoint, focusing on the human side of the economy: the labor market. How will AI affect people’s jobs and the future of work?

How disruptive technologies generally affect the economy

AI in this massively available and accessible form seems truly new and unique. But economically speaking, it could just be considered another advancement in technology that can improve our productivity. This scenario has been happening in our economies since the dawn of civilization. So, what economic impact do these advancements tend to have?

Consider the classic Cobb-Douglas production function, shown below:

        \mathit{Y} = \mathit{A}\mathit{L}^{\alpha}\mathit{K}^{\beta}

In this equation, K stands for capital, L stands for labor, and A stands for “technological progress”. A is often glossed over in introductory courses, but is generally described as a “good thing”. When this factor increases, it means that even with the exact same inputs, the economy can produce more.

In other words, technological advancement leaves room for Pareto improvements. The development of AI can be thought of as an increase in A. Everyone is theoretically made better off due to this increase in A; this is why economists typically consider technological progress to be a good thing. To see why, consider previous inventions that have changed our daily lives and made certain careers obsolete.

One such example is the invention of the printing press. Before it was available, books had to be copied by hand, which professional scribes were hired to do. As books were hand-copied, they were quite scarce, and the cost of books was much higher (relatively speaking) than they are today.

Undoubtedly, the invention of the printing press was a boon for society. It allowed a greater quantity and wider range of books to be printed, which enabled more people to become literate, and for those who were literate to make use of their literacy more often. It also caused scribe careers to become obsolete. But nobody would argue that we should re-introduce those jobs, and stop typing or printing.

AI tools could have similarly beneficial effects as other disruptive technologies like the printing press did in the past. It can potentially help us create things much faster and more cheaply, resulting in a productivity boom.

The hidden cost of progress

But, innovation has a cost that the simple Cobb-Douglas model of the economy does not capture. As mentioned, professional scribes barely exist today due to technological progress. But, once books were easier to produce and literacy was more widespread, becoming an author became accessible and doable.

This process of new technology creating new opportunities while making old jobs obsolete is known as “creative destruction” in economics. This is one of the main reasons why people may be hesitant about a new technology’s impact on the economy. Nobody wants to become unemployed and (temporarily) lose their source of income, even if it means that society overall is becoming more efficient and productive.

Another valid concern is the ability of workers to find new employment. This is especially concerning for people who were working in a now-obsolete job. Finding new work takes time and might require retraining, which can be costly. Workers also may have liked their now-obsolete jobs, and might not enjoy being forced to find a new career. Some workers close to retirement age might struggle to find new employment at all.

Suggested Opportunities

The impact of artificial intelligence on the labor market

Some fears of losing jobs to AI have already proved valid. In the first half of 2023, AI quickly started to impact the labor market as businesses looked to it as a revolutionary cost-cutting tool. For example, IBM announced that up to 7,800 jobs could be replaced by AI, and instituted a hiring freeze on those positions, presumably until AI is advanced enough to eliminate those jobs entirely.

Other companies, including DropBox and Buzzfeed, announced job cuts. DropBox alluded to AI, citing “developing technology” as one reason for its job cuts, while Buzzfeed explicitly cited AI as one of the reasons they were letting workers go. In a similar vein, Wendy’s announced a partnership with Google to replace human employees with an AI chatbot for its drive-through ordering system. The system is expected to begin launching in the summer of 2023 after being trained sufficiently.

In response to situations like these, some workers are fighting to gain assurances that their jobs won’t be replaced by AI. A major writer’s strike enacted by the Writers Guild of America, the union that oversees writers across most of Hollywood, began in early May 2023.

Interestingly, the writers did not demand for AI to be banned completely. Rather, they began to strike for better working conditions in general, including the promise that humans would be prioritized in the creative process. The Guild made it clear that AI can be a useful aid in the writer’s room, but should not be seen as a replacement for human professionals. As of this writing, negotiations are ongoing.

Some firms have banned the use of AI, for marketing or for safety

It’s clear that some firms aim to use AI to cut costs, increase productivity, and even replace workers. At the same time, some companies see value in promoting their lack of AI usage – or are holding off for safety reasons.

For example, the tabletop game company Paizo made waves when it announced a ban on the usage of AI-generated content in all of its products. Paizo employs many writers and artists – which fulfill two functions that AI is able to replicate. Paizo’s announcement was seen as a good thing by fans and hobbyists, and has helped to drive record interest in their products in early 2023.

Typically, the adoption of new technology is needed to remain competitive. It’s possible that Paizo’s competitors end up utilizing AI to produce quality content much faster and cheaper than Paizo can, eventually forcing them to adopt AI or driving them out of business. But, if having 100% human-made content becomes important for product differentiation, or as a signal of the company's values – similar to the fair trade label in chocolate – then companies like Paizo could carve a niche for themselves and remain competitive even without utilizing AI.

Other firms have moved to ban the usage of AI for data privacy or security concerns. Because these tools are hosted by other companies and connected to the Internet, sharing sensitive information with them could pose a security risk. Samsung is an example of this, as the firm banned AI on its company devices after employees leaked sensitive information to an AI tool.  

In the future, firms like Samsung could opt to develop their own internal AI tool for their employees to use. This would allow employees the benefits of an AI tool without running the risk of sensitive company information leaking. But developing and training a useful AI takes time, development expertise, and training data, which can be expensive. And, inference costs mean that using an AI tool can be expensive, too – one estimate puts the daily cost of running ChatGPT at about $700,000.

The creative side of creative destruction: new income streams from Internet text

AI must be trained on data. For these large language models (LLMs)  specifically, this usually involves massive volumes of text taken from the Internet. Normally this text doesn’t belong to the companies developing these AI tools. Consequently, some large websites like StackOverflow and Reddit have made plans to start charging AI developers for using their data, which so far they’ve been doing without paying for it.

These types of sites are also beginning to charge for access to their data for future developers, which could become a new income stream for these companies. The CEO of Stack Overflow, Prashanth Chandrasekar, said it well himself: “Community platforms that fuel LLMs absolutely should be compensated for their contributions so that companies like us can reinvest back into our communities to continue to make them thrive”1.

Indeed, this new paradigm of charging for text data could even create jobs to carefully build and maintain forums (or other content databases) that would be sold to developers as training data. In this way, human-only online spaces could become an important and lucrative new aspect of the digital economy. Certainly, other new services and careers will appear due to the widespread usage of AI – such as possible AI detection tools and “prompt engineer” job positions.

There is already some disruption in the labor market due to AI, and there’s likely to be more coming. As with every instance of creative destruction, there will be some growing pains while society and the economy is restructured to take advantage of the new technology. But, it’s also highly likely that new careers will emerge as the dust settles.

On the way, people will claim that artificial intelligence is going to impact the economy in transformative ways, while others will view it as a threat to their livelihoods. The true answer likely lies somewhere in the middle, though only time will tell.


You need to login to comment


The INOMICS AI can generate an article summary or practice questions related to the content of this article.
Try it now!

An error occured

Please try again later.

3 Practical questions, generated by our AI model