IJzeren vibes, Victorian vibes, and other stuff

lmao ja dat was dus ook mijn eerste gedachte

@sappystrawberries

Yes it's fair use, but this is still a crime under I presume every countries archival laws.

Because we don't just wanna preserve the text, we wanna preserve the para-text. Pages where text has been underlinded by older users, annotations, printing mistakes, differences between different prints of the same book. These AI companies are not "preserving" these books. THIS IS NOT DIGITAL PRESERVATION. Digital preservation would be PDFs. Because with PDFs, you're preserving everything that can be preserved digitally. ("But these are PDFs". No, these are not PDFs of the books. They are PDFs of the text. That's not the same.) The only thing you'd miss is the physical material, which you can't preserve online. So something like carbon dating would become impossible for example, but usually that's not a major problem.

Now, these companies are doing transcriptions. Which on its own is fine. It is an absolute crime agianst humanity that they do not share their transcriptions with historians. Even though the field of history would tremendously benefit from these databases. But hey, that's by the by. But calling transcriptions "digital preservation" is super problematic. Like I said, transcriptions are just plain text. They don't make the difference between a paragraph and a title usually, and leave out small things that make each book 'used' and 'have a history' and such.

Also, most transcriptions aren't perfect. Even with printed books, you need to make choices. Do you preserve letters and especially punctuation that is no longer used? Do you keep abbreviations? Do you make a differences between titles and normal text, if so, how? And that's besides the fact that it's really easy for even the most clear text and the best scanners to come up with a bad transcription. And that's presuming you're even dealing with English text.

We know most of these AI researchers have an IT background and not a background in linguistics. I don't trust them to take into account historical writing, even the writing used in early modern western printed books, or how their scanners might not be so great for languages beyond English. So not only do they commit crimes against historical preservation, it wouldn't surprise me if their output is shite.