This above comment is a great example about the limitations of language when dealing with image manipulation and generative image enhancement. “People added ad

This above comment is a great example about the limitations of language when dealing with image manipulation and generative image enhancement. “People added additional craters” … no they didn’t. Without knowing anything about what people interested in stress-testing Samsung’s software did, I can say that because if they did really add additional craters, it’d be front-page news. I assume what people did was added artifacts which are not real (let’s call it ‘fictional’) to an image. Samsung’s software enhanced it as long as these fictional artifacts conformed to its notion of what a blurred crater looks like. So it turned fictional artifacts into higher-res fictional artifacts. At this point, you might as well generate the moon image in a terrain generator like World Machine. Put it another way, if Coca-cola tomorrow puts a giant billboard that says “Drink Coke” on the moon, and it turns out Samsung’s software didn’t expect that, it’ll need a software update. More realistically, if tomorrow you have a lunar impact crater with a shape distinctive enough that Samsung’s software doesn’t expect, again, you’ll need a software update. And that’s the heart of the issue: the camera is no longer capturing reality, but instead injecting detail that someone at Samsung decided would be appropriate for your shot.