McDonalds is also at the far end of the spectrum of "human service jobs that are less-skilled but difficult to automate". There is plenty of demand for higher quality versions - requiring a higher degree of skill and creativity - of the same general type, like fine-dining.
Truly creative (in the sense of it having a high degree of novelty and quality) work is not at risk of being automated any time soon. What is at risk of being automated is the category of "creative" work that requires some skills but is mostly assembly line. Category-defining or truly novel art almost by definition can't be produced by existing AI in any form, because AI can only remix the content it's seen already. "Generic rock song" or "clip-art like picture of a guy yelling at a computer" are at risk of going away, but I hardly think that means humanity will no longer flourish - producing that kind of stuff is romanticized as a cool, highish status thing to do, but functionally I don't really see it as any different or more worth preserving than obsolete skilled labor of the past like carriage-makers or human-computers.
I also think people tend to make the "Lump of Labor" fallacy when thinking about this stuff - economically speaking, if human workers are no longer needed to produce some high-value output, in the long run the excess labor/"talent" that gets freed from that task finds other value-producing tasks to do: https://en.wikipedia.org/wiki/Lump_of_labour_fallacy. Long-term unemployment and underemployment is completely a solvable economic problem; both can increase due to short term shocks like technological development and shifts in supply/demand, and in some cases underemployment is more a matter of "wrongly skilled", but long term they're both a matter of ensuring there is enough money, liquidity, and capital deployed to drive demand for marginal increases in both jobs and quality-of-jobs (without ruinous inflation).