[coldcode]

If you make art in postcards, then maybe you will be replaced. Diffusion based AI image generation is limited to a little over 1 megapixel, or double if you use an expander, or you can infinitely grow an initial image with more of the exact same sort of thing. 1 MPx printed at 300 dpi is similar in size to a postcard. Diffusion being able to generate meaningful printed images at 50-200 Mpx is impossible. My art is made at that size (that's 7200 to 14400 pixels per side approximately @ 300 dpi). I don't use AI. I do use my own code.

Try making a Jackson Pollack using AI, AI can only effectively make objective images, given it requiring linking similar images via limited text descriptions.

Making headers for newspaper articles or blog posts on the web might be OK if you don't care much. Check out https://www.quantamagazine.org to see how real artists do things. Their header images are really imaginative given the subject matter, something AI today can't do. It might be instructive to take their articles and try to make header images for them.

If you make videos, then its 1080p which can be done by AI. But video has its own issues with video diffusion generators having limits on continuity. They are getting better, but the longer the clip, the less control you have and the longer it takes to fiddle with the prompt(s). It won't ever replace real movies unless you enjoy the SciFi channel or Sharknado. That would require an entire new kind of AI which we don't have today.

[ilaksh]

I will give you the possibility that some text to image generator models of the moment may not be able to create images that compete with Quanta magazine artists. Maybe. These systems have limitations but I think if you study your own logic you will see that you can't be sure that they will not overcome them in the next 2-20 years.