I suspect many parts of this will not age well (saying "AI will never do X" is a recipe for future embarassment) but others might well be right on the spot, or close to it. The cool thing here is that being creative allows you to redefine what you do freely and constructively.

My standard quip is "if you can explain your job succinctly it can be automated". But there is a counterpart: "if you can change what your job is about, it cannot be automated" (at least until AGI).

This is why philosophers and creatives have job security against automation. At least as long as they critically and cleverly change what they do as a response to automation ability. There is a diffuse core purpose of the fields, but what you do near it can change.

This is very different from ditch-digging, where the only likely move is to become a bot-herder rather than stay in the soil and try to make artisanal ditches or complex high-tech watercourses. The purpose is simple and clear, and the work cannot be redefined much.

The problem with diffuse purpose jobs is that success may be equally diffuse and hard to judge (another reason why automation is hard there). That allows a lot of faking and pretension, and if people escape automation by going there there will be much more.

However, many of actual tasks done can have clear success criteria. It is easy to "be" a creative or intellectual... but only some art is breath-taking, only some papers make you think in new ways. I suspect this often happens near the complex border between automatable and not.

This is because this works harnesses the power of automation to go beyond standard human ability, but it is guided by a very creative or strategic insight that is both enabled by automation and sees possibilities and values that are not automated.

This has been going on for ages. Without AI we "automated" logical deduction in philosophy, and the great papers were the ones using it cleverly on problems that mattered (or seeing new problems). A lot of art is routine, but Michelangelo directed his helpers to fill in.

A lot of "AI will take our jobs" is pointless handwringing; it is better to do like Alex and analyse where the new opportunities lie. Including figuring out if there actually are opportunities or one should get out. This looks tricky to automate right now.