In regard to AI's use in illustration(though you can extrapolate it to other mediums).

Professional artists work with writers to convey intent. Some art styles are better than others at supporting nuance in writing, which has largely driven their development within otaku media. These nuanced art styles, encourage more sophisticated writing. Writing and art, have a mutually edifying relationship.

An algorithm has no conception of intent applied within art, e.g. gesture, composition and characterization. All three, can be affected by every individual line. An algorithm which statistically composites images together, cannot produce anything with intent behind it, let alone on a individual-line basis.

What do lesser artists have over an algorithm? Intention behind art can exist even at lower skill levels. It can come through via subconscious decisions even. The prompter may have an intent, but it can never be correctly expressed by the composite of images with generalized labels. Labels which cannot convey anything close to the full intent of their subject.

A subject such as "drinking water", has certain intrinsic traits, but these do not include the gesture, composition and characterization, which allow art to describe narrative intent.

How about the use of an algorithm to generate concepts? This is often used as the most "diplomatic" argument in favor of image generation's addition to the creative workflow. The reality is, a professional artist, can invent compelling concepts far quicker than a prompter can coax anything interesting out of a model.

These models therefore add nothing in terms of expediency or quality to the creative workflow. Their utility is in allowing unskilled workers to cut corners in a project. Those who gain from the normalization of unskilled labor being selected over skilled labor, will ceaselessly peddle the notion of their output being at least equivalent, and incentivize it within creative industries.