Why ubiquitous AI will mean more, not fewer, white-collar jobs
While the spectacular improvements in the performance of generative AI (GenAI) models in the last couple of years have led to justified excitement, there is also widespread concern regarding the impact of all these new tools on so-called white-collar jobs, that is, jobs mostly to do with some kind of information processing. This includes programmers, but also copywriters and designers, as well as call center employees. All of these occupations have important components that, a couple of years ago, could only be done by a human but can either already be automated using GenAI or are easy to imagine being so automated within a year or two. Another occupation that belongs on that list, which is not as frequently mentioned, is middle managers. A lot of their work is transforming information, and much of it is relatively rule-based.
Will the inevitable spread of GenAI tools for information creation and processing put most of these jobs at risk? In some cases, certainly. For example, if all that call center workers do is follow a script, and escalate if the script doesn’t fit, without any chance to apply their judgment or understanding, their job can now be done better (cheaper and more reliably) by AI.
But most knowledge jobs are not like that. In the work of any programmer, designer, or middle manager, there is a component (maybe an irreducible one) of judgment and common sense, going beyond any particular rule set. So while AI may be able to complete a code snippet, create an image from a prompt, or draft a report from a bunch of source documents, it can’t (at least at the moment, and I believe not for a while yet) understand the bigger picture of which code, images, or reports are worth creating and why. Thus, for most knowledge work, AI is much more likely to appear as an assistant rather than a full-scale replacement for humans.
Transformational changes encourage job growth
This in and of itself is not enough to allay the fear of widespread job losses: If the knowledge workers become more productive thanks to AI assistants, will this mean the economy will need fewer of them? To see why this is unlikely, let’s consider another profession: car mechanic. Back when cars first appeared, each one required constant maintenance, so the only practical way to own and use a car was to employ a full-time mechanic. Over time, cars became much easier to operate and maintain, to the point where professional intervention is needed once a year, if then. Does that mean that fewer people are making a living as car mechanics now? Not at all. As the total cost of car ownership (including car mechanic services) plummeted, cars became ubiquitous, so the total demand for car mechanics increased instead of decreased.