GenAI is already here, but how is it affecting the labor market?

Since Oxford University's Carl Frey published his groundbreaking paper about the impact of technology on the labor market in 2014, concerns have grown about automation displacing jobs, particularly in fields susceptible to robotics and AI.

For much of the intervening decade, the impact has been negligible, and indeed, studies have shown that investment in technology like robotics has led to more jobs being created as firms become more productive.

However, the rise of generative AI has spawned a fresh wave of concern. The latest of which comes from Imperial College London research, which looks at the impact genAI had on freelance jobs in areas such as writing and graphic design.

Testing the impact

The researchers analyzed nearly 1.4 million job posts on an online freelancing platform from July 2021 to July 2023. They chose this particular Petri dish because of the task-oriented and flexible nature of work on these platforms, which arguably places them at the front line of any disruption. Of course, platforms such as Upwork have also tended to trend towards the low-value end of the market in recent years, which may lend itself to more automatable tasks.

The results aren't promising, as the introduction of ChatGPT has led to a 30% fall in the number of writing jobs, with a slightly smaller drop in software and web development jobs. The researchers also found a similar drop in demand for graphic design and 3D modeling freelancers after the introduction of tools like Midjourney and DALL-E 2.

The data doesn't suggest that this is a temporary pattern as companies experiment with the technology (despite other studies suggesting that's largely all companies are doing thus far). The researchers compared the decline as a result of automation with more regular seasonal fluctuations and believe that the automation trend is more likely to endure.

A bigger impact

The authors compare this impact with previous waves of technology, such as the introduction of robots into factories. For instance, they cite research showing that a 20% increase in robot adoption in France resulted in just a 3% fall in employment in those sectors.

On the surface, the impact seen on writing jobs on the freelancer platform is markedly more worrying. It does beg the question of what kind of jobs they actually were though. We've already seen a number of publications, such as CNET row back considerably on the use of ChatGPT to generate content.

Aside from obvious issues around originality, there are also sufficient concerns around ChatGPT's ability to research reliably to suggest using it to automate content that is designed to make you stand out as a thought leader somewhat naive.

Lower quality

As this article for the BBC outlines, many applications of ChatGPT for content creation are little better than the content farms of old, but with AI deployed instead of a low-cost human being. Few would argue that this is high-quality content, and research from the University of Alberta memorably refers to it as "botshit."

As research from the University of Zurich illustrates, the palming off of AI-generated content is far from risk-free. The study found that when content was labeled as AI-generated, it was perceived by readers as less accurate. What's more, they were also much less likely to share that content with their social networks.

When the researchers delved into why this was, a key concern was that the AI-generated content was delivered without human oversight, and was therefore less trustworthy as a result.

A changing role?

As the BBC article outlines, this is prompting some organizations to change the role of writers rather than do away with AI altogether. They're now responsible for overseeing what AI creates and making sure it is fit for public consumption. Of course, this change in role is often accompanied by a pay cut, as editing content is seen as less arduous than producing something from scratch.

Such a change in role fundamentally misunderstands writers' motivations. Researchers have shown that these tend to revolve around self-efficacy, the desire to showcase their skills and expertise, the inherent attractiveness of the task, the outcome of actually writing something of value, and the identity that writers have, both of themselves and in society more broadly. It's hard to argue that any of these motivating factors remain if they're asked to perform such a diminished role.

Early insights into the corporate deployment of generative AI suggest it is being used to cut costs more than drive value. Given the often mediocre output in terms of content and the decimating impact in terms of employee engagement, one has to question whether this is really worth the trouble.