Worse, Xerox scanners specifically meant for digitizing documents have changed text for a long time. The compression algorithm they used (I think even in the default settings) sometimes replaced e.g. 6 with 8, and similar things. See: https://www.youtube.com/watch?v=7FeqF1-Z1g0 (german, but there should be news articles from back then in english as well, somewhere)

▲

gruez 6 days ago | parent [-]

That's not really "computation photography" in any meaningful sense, closer to "digital processing". It's not impossible for such glitches to occur with modern smartphone cameras, but it's implausible. I don't think there's ever a confirmed instance of such a gaff happening. Meanwhile a few years ago there was a photo with a misplaced leaf that made the rounds, and people were complaining about how it was caused by computational photography, but it turned out the photo was accurate. The leaf actually there.

	▲	matrss 6 days ago \| parent [-]
		My point was that you don't have to take a photo of a receipt to run into this issue, actual machines specifically build to digitize receipts and other documents already made this kind of mistake. No idea if this can happen with what modern smartphone cameras do to photos. If "AI" is involved then I would expect such issues to be possible because of the basic nature of them being random generators, just like how LLMs hallucinate stuff all the time. Other "enhancement" approaches might not produce issues like this.