I stumbled upon a study discussing the influence of AI on the freelancing landscape, echoing our fears and unveiling some unexpected findings. The study highlights a significant 33% reduction in writing positions, with translation (19%) and customer service roles (16%) also experiencing declines. The extent of this impact is startling, though the areas affected are somewhat predictable. Counterintuitively, there’s been a growth in video editing, graphic design, and web development opportunities since ChatGPT’s debut. The article speculates that AI tools are still not good enough to replace these jobs.
Let me theorize further about what might be driving some jobs to evaporate while others to remain resilient.
In a lecture by AI pioneer Yann LeCun, he outlines challenges AI faces, particularly with “planning” that involves hierarchical dependencies. For example, a web development project manager must outline the primary steps needed for product completion—such as wireframing, design, coding, testing—and then detail the sub-steps required for each. This complexity seems infinite. Each coder needs to decide on the functions to create, the sequence of actions each function must perform, and so on. LeCun argues that a significant breakthrough in AI technology is necessary to address these intricacies.
Viewed through this lens, the patterns in job displacement and stability become clearer. It might seem logical to assume that “soft skills” like sales and support would be challenging for computers, but this isn’t quite accurate. Actually, AI struggles with tasks that are also difficult for many humans, such as complex mathematical problems and precise memory tasks.
The impact of AI on the job market, I believe, has less to do with the nature of tasks and more to do with the complexity of dependencies. Simplifying these dependencies makes a job more vulnerable.
Writing and translation, for instance, are more susceptible to AI disruption due to their straightforward nature and lack of external dependencies. They are self-contained. With a clear job outline, both human writers and AI systems can commence work immediately. Hence, roles with easily definable parameters are likely the first to be disrupted.
So, for instance, although web design remains resilient (because it is highly collaborative), I suspect that illustration is not because it has no external dependencies.
This is my hypothesis.
I will email you when I post a new article.