A good chunk of the creativity theory literature throws into question the idea that creativity without skills is actually "a thing". Now of course, we're not necessarily talking about a total lack of skills, but rather about transferring those skills to a machine—something akin to how a conceptual artist might work with an architect, carpenter, or other fabrications expert in the construction of a new sculpture or installation.

However, what's more likely is that a new crop of skills will appear that complement this off-loading of skills specific to 3D illustration/design to AI. There will be skills specific to effective prompting, for one, but I'd imagine it won't be long before new interaction/inferface concepts arise to complement the use of prompts. After all, describing things in language is often not the most effective way of getting an idea across (hence the old adage that "a picture is worth a thousand words").

I think we're very much in a honeymoon phase with text prompting. Once the novelty wears off the fact is that—from a user perspective—text prompting is indistinguishable from search. It's only our understanding/assumption that searched images pre-exist our search that makes prompting experientially different. Obviously there's the novelty/bespoke nature of prompted images, but it's somewhat debatable that this will be a meaningful distinction for long. And of course, the more prompt generation is used, the more surplus images will be created, incentivizing providers to save the GPU cycles and just pump out previously generated, but unused (i.e., unselected) images. It's also possible (though unlikely) that generating may become more efficient than searching, in which case image search will simply be supplanted by generation. Either way, the harsh reality is that you're typing in text and getting what you asked for; i.e., search.