AI’s Impact on the Job Market

The widespread adoption of artificial intelligence (AI) has sparked debates around its impact on society. Some argue that our society will inevitably be required to move to Universal Basic Income (UBI) due to widespread automation across many professional fields, while others feel that AI offers just a slightly better search engine. We believe that neither of these futures will play out, and that history provides a more nuanced and informed view of where this is all headed.

This week, we want to dive into our philosophy on AI’s impact on the jobs market in the long run, and also use a recent report by Anthropic to infer what that will look like in the near-term.

A Long-Term View

While there are differences in how AI is impacting society compared to how other forms of automation have worked in the past, we think that fundamental economic principles will still dictate the long-term impact on jobs. Historically, the impact that technology has on the job market has followed something similar to these steps:

1. Technology is applied to production, and productivity increases
2. This increase in productivity makes the creation of goods and services cheaper
3. Cheaper goods and services increase disposable income for other goods and services
4. New industries and jobs emerge to meet this new demand

This process is reinforced by the “Lump of Labor” fallacy, which is the idea that there is a fixed amount of work to be done. This zero-sum view has been refuted time and time again throughout history as new jobs at higher abstraction levels continue to emerge following technological innovation.

Additionally, when labor becomes more efficient, firms can pay those employees more. This idea is called Marginal Productivity Theory. If an employee at a business generates $10 in profit, the business should be willing to pay them up to $9. If an employee augmented by technology can generate $100, the business should be willing to pay them up to $99.

To be clear, these are economic theories, and in reality, other factors are at play. Even if you agree with these ideas, they also do not necessarily answer the most challenging questions that we must ask in the near-term: what jobs will AI take and when will it take them.

Automations’ Historical Impact On Jobs

Automation has existed forever and will continue after AI, but it can still hurt certain parts of the job market in the short term.

Some of the more popular examples of automation in recent history are industrial robots and computers. Both provided enormous benefits to productivity but replaced jobs such as assembly line workers, switchboard operators, among others.

The chart below shows occupations ranging from low to high wages and how they have evolved over time. The middle-wage jobs have seen the largest impact from historical automation.

Michael Strain, Director of Economic Policy Studies at the American Enterprise Institute, argues that the jobs that have been replaced in the past are middle-wage jobs that have required the repetition of tasks done with expertise and quality. These are the types of tasks that machines and computers are good at.

AI’s Impact Mirrors Historical Automations

As with past cycles of automation, AI will undoubtedly take some jobs. We can review recent Anthropic reports to see which jobs this is likely to affect. Anthropic lists out the 6 major categories and job titles where AI has been used, and the percentage of prompts within these categories:

The top 5 professions using AI to date have been:

1. Computer programmers
2. Systems software developers
3. Application software developers
4. Bioinformatics technicians
5. Copywriters & technical writers

When comparing these results to previous automation, we see both similarities and differences.

What Is the Same This Time Around?

When looking at these professions, we see a similar trend to historical automation. Software development (the careers in the first category which make up 37% of AI usage) has meaningful repetition. Before AI, it was a common joke to say that you took a chunk of code from Stack Overflow and dropped it into your project. The whole concept of coding libraries (which is a collection of prewritten code) is an example of repetition within the industry. This is not to say that all software development is a repetitive task, but there are clearly pockets of repetition.

Technical writers & copywriters also fall into the category of repetition. The vast majority of thinking on the internet is not novel, and even the novel thinking builds on previous work. Again, this does not mean that all writing is repetitive, but there are certainly pockets of repetition.

What is Different with AI?

“Bioinformatics technicians” (in the Office & Administrative category) play a crucial role in analyzing and interpreting complex biological data. While AI may help streamline some repetitive tasks here, there are also high-profile cases of AI being used to analyze large datasets and make predictions. This is extremely powerful, but it is more of an extension of big data than it is of replacing critical thought.

Similar to previous eras of automation, “both low-paying and very-high-paying jobs had very low rates of AI use,” according to Anthropic. But what is uniquely different is that, “it was specific occupations in the mid-to-high median salary ranges, like computer programmers and copywriters, who were—in our data—among the heaviest users of AI.”

This shift towards higher paying jobs is an essential difference between automation today and automation in the past, and perhaps what is fueling the idea that AI is coming for “half of white-collar jobs”. But when also considering that 57% of use cases have been categorized as “augmentation” versus “automation,” the prediction that AI is coming for such a substantial segment of white collar jobs feels unsubstantiated.

Takeaway: The transition ahead with AI automation’s impact will be difficult for certain pockets of the job market. People will definitely lose jobs (some already have), and unlike past waves of automation, this one could reach further into higher-paying, white-collar roles than previous technology advances have in the past. That said, there is a reason for optimism when it comes to AI’s long-run impact on work. In the past, technologies that made workers more productive ultimately benefited the broader economy and improved living standards. We believe that pattern is going to continue, bringing a net-positive impact on the economy and the jobs market.