In the earlier thread about how to advise prospective PhD students these days, philosopher Tim Maudlin reports the advice he got as an undergraduate at Yale from Professor Robert Fogelin in 1980:  "He said—and I quote verbatim because one does not forget something like this—'Not even if you were Immanuel Kant would I suggest going to graduate school in philosophy.'" What insane advice to give to an undergraduate, at Yale no less, which has sent many dozens of students on to successful academic careers in philosophy since that time.  Even now, that advice would be crazy, and the situation is much more grim in the U.S. due to the Republican war on higher education.  Despite all that, there will still be jobs teaching philosophy, but fewer than even in the last thirty or forty years.

The earlier thread also included various prognostications about AI taking over many pedagogical tasks.  Count me skeptical, although for reasons we have noted before, AI will certainly change teaching in various ways.  But we should not be sanguine, and the effects are hard to predict with any confidence.   

An illustrative anecdote:  when I was a first-year litigation associate at large NYC law firm in the late 1980s, one of the most dreaded assignments was a "document production" as part of the discovery process:  a team of associates would go to a large warehouse full of hundreds of boxes with tens of thousands (or more) pages of documents, and we'd have to go through them all looking for the "smoking gun" evidence needed for our case.  This was hugely time consuming, often involving thousands of hours of time of dozens of associates.  Now, since almost all documents are digitized, the entire task can be done by two or three associates using search engines on the digitized documents.  The reduction in labor time is on the order of 75-80%.  Now large language models (trained on proprietary legal databases) are also being used to generate legal memos surveying the law in an area, with links to cases (a check on the accuracy of the results).  This, too, will reduce the demand for junior litigation associates.

That being said, I don't foresee AI taking depositions, or displacing trial lawyers, or handling negotiations…at least not in the next forty or fifty years.  Many tasks that professors perform are not at risk from AI.  The bigger risk is that universities facing budgetary pressures will turn to AI for pedagogical and evaluative functions that can be done better by human beings, but can be done passably by AI. 

Since AI is not actually "intelligent" (it neither thinks nor reasons), but is rather a mimicry machine, one does wonder how AI will perform when most of the output it is training on is AI output.  I guess we will soon find out.