Spotting a fake used to mean spotting edits, such as a repeating background, a shadow in the wrong place, or artefacts around a pasted-in object. That worked because manipulated images usually started with a real photo, leaving clues behind.
AI-generated images are different. They’re created from scratch, with no original image underneath, so those kinds of editing mistakes often don’t exist.
Even this is out-of-date, thanks to edit-models. This is why ‘replace her clothes with a micro-bikini’ is a problem. The model also doesn’t clone-tool from nearby in the same image, like us Photoshop Phriday veterans. They’re fully capable of generating an image from scratch, but replace just enough to blend into a given photo.
An image used to be reasonable proof that something happened.
Even this is out-of-date, thanks to edit-models. This is why ‘replace her clothes with a micro-bikini’ is a problem. The model also doesn’t clone-tool from nearby in the same image, like us Photoshop Phriday veterans. They’re fully capable of generating an image from scratch, but replace just enough to blend into a given photo.
Ehhhh.
Yeah, photos haven’t been reliable for 20-25 years. Hell, people fake Twitter posts as YouTube thumbnails all the time.
The invention of photography just barely predates the first photographic hoaxes.
I keep coming back to Death of a Fantastic Machine.