I guess the review is a lot of little things like that. Like, it's good as a game, in a way; but there's too many little mistakes in the wrong areas. If I actually wanted to use either image as something other than "hey, let's see what it can do", I would spend so much time fixing those little mistakes in Photoshop or whatever that at that point I might as well start from scratch. Like, people goof all the time; think Bob Ross and his "happy mistakes". Or, hell, the infamous "stormtrooper bumped his head on the door" in the background of Star Wars is an even better example; it's a goof, sure, but it's in the background and where your attention is focused elsewhere, so it goes unnoticed for years (or it was before George Lucas decided to actually highlight it in his latest "special editions"). Generative AI's fault is it is incapable of recognizing mistakes, period, so it doesn't know how to choose which ones to ignore and which ones it does need to fix (which it is probably incapable of). Also, no human would make the mistake of giving their character a button down mid-riff, for example.
But even when there aren't big glaring mistakes, at least not front and center, AI generated images look generated. Right? I'm not alone here, am I? Like, it can make fake photos; that needs all the detail it can get, but when you're making fake drawings, the sheer amount of detail can be off-putting. Like, it's the "why are they all in a tree" problem of the first set; a white void would have been fine, perhaps even preferable, as all those branches and leaves with the exact same level of detail and effort as the actual subject is distracting. Like, sometimes human laziness is there for a reason; a high level of detail may be exhausting for the artist, but it's also exhausting for the viewer.
I guess the review is a lot of little things like that. Like, it's good as a game, in a way; but there's too many little mistakes in the wrong areas. If I actually wanted to use either image as something other than "hey, let's see what it can do", I would spend so much time fixing those little mistakes in Photoshop or whatever that at that point I might as well start from scratch. Like, people goof all the time; think Bob Ross and his "happy mistakes". Or, hell, the infamous "stormtrooper bumped his head on the door" in the background of Star Wars is an even better example; it's a goof, sure, but it's in the background and where your attention is focused elsewhere, so it goes unnoticed for years (or it was before George Lucas decided to actually highlight it in his latest "special editions"). Generative AI's fault is it is incapable of recognizing mistakes, period, so it doesn't know how to choose which ones to ignore and which ones it does need to fix (which it is probably incapable of). Also, no human would make the mistake of giving their character a button down mid-riff, for example.
But even when there aren't big glaring mistakes, at least not front and center, AI generated images look generated. Right? I'm not alone here, am I? Like, it can make fake photos; that needs all the detail it can get, but when you're making fake drawings, the sheer amount of detail can be off-putting. Like, it's the "why are they all in a tree" problem of the first set; a white void would have been fine, perhaps even preferable, as all those branches and leaves with the exact same level of detail and effort as the actual subject is distracting. Like, sometimes human laziness is there for a reason; a high level of detail may be exhausting for the artist, but it's also exhausting for the viewer.