[RichieAHB]

Taking a picture is deterministic. You aim at something, configure the settings and the cells / CMOS will capture that light in the right manner. You can argue that autofocus and portrait-mode are eroding this but ultimately, if you, the human are not pointing at something good in the first place, well you’ll very likely not have a very good picture. I feel like with AI and a random seed you can ask a question of which you have added none of your own reflection into and it will produce something somewhat in the ballpark that may pass as real. I.e. you don’t even have to point the camera at anything good …

Edit: not to say you can’t use it as a muse or for inspiration, but it isn’t a prerequisite.

[jstarfish]

You do have to provide novel input though, otherwise it's just a vending machine for motel art. Everybody is capable of generating the same generic crap by default, but if you're sufficiently creative in your prompt, it's no different than an artist having an assistant.

The real problem is the long tail. You found a unique form of expression? Your work gets trained on and you get cut out of your own agency.

I fully expect deliberately-bad art will become a thing, because it would be ruinous to datasets to be trained on, will stand out in a sea of KincaidDiffusions, and go generally unexamined by AI crawlers. Lomography was ahead of its time.