The main thing here is that image generation trough an llm doesn’t even count as creating.

Asking is not creating. In these systems people ask an llm to use a genai tool, the people never actually touch the tool themselves. (They wont even allow it lol)

Thats why ComfyUi with stable diffusion and not chatgpt is the standard for serious art work using ai.

They are fully open source, offline and they don’t require any more energy then playing a video game.

Also workflows look like this, more accessible means a different set of skills can now get you similar results. But it is still skill.

you indeed dont need to know how to hold a pencil to build that.

(Also there are more and more models exclusively trained with artist consent)

So yes, ai does make art more accessible to a small group of technical people. Most people know no one in this group.