[hansword]

I would say it helps to first think what you want to get out of it.

If your task is "show me something that breaks through our hyperspeed media", then I guess some obscure museum is a better place than an ML model.

If your task is "find the best variation on theme X" or "quick draft visualization", they are often very useful. I am sure there will be many further tasks to which current and future models will be well suited. They are not magic picture machines. At least not yet.