Picking up on framing issues in a binary way being a trap when it comes to unravelling the complexity of likely impact of AI, I think binary thinking survives because choosing a side is less effort than holding four positions , each with an element of truth , and admitting none of them cancels the others out.
Polanyi’s observation was that we know more than we can tell — that real competence, the kind a skilled clinician or engineer has, doesn’t reduce neatly to a procedure you could hand someone in writing. I think the same is true of judgement about AI. You can train someone to use the tools in an afternoon. You cannot train them, in an afternoon, to notice when the speed of the answer has quietly become a substitute for the quality of the thinking behind it.
That’s the actual shortage. Not AI skills — there’s no shortage of those, and there will be even less of one by Christmas. The shortage is in people who can sit with an AI-generated answer and ask whether it’s good, rather than whether it was fast.