If you think about this generation of text/media AI as being a kind of generalized decrompressor for some input prompt, the output has more detail than the inpu

If you think about this generation of text/media AI as being a kind of generalized decrompressor for some input prompt, the output has more detail than the input but never more information. So as any model becomes widespread, people will become increasingly sensitive to the sameness of easy, simple prompts. So the graphic designer at the law firm may lose their job tomorrow, but the law firm is going to have to hire a AI Image Technician in a few years so that their stuff looks like somebody actually invested in it (which matters to the law firm’s brand image or they’d not have a graphic designer in the first place). So yeah, the job market will shuffle around, but it’s unclear what that means in net number of jobs or what share of AI Image Technicians will literally have just been Graphic Designers five years earlier.