Re: [RFCv3] Counter-Proposal -- Interpretation of DFSG on Artificial Intelligence (AI) Models

[Aigars Mahinovs <aigarius@gmail.com>]

On Sun, 11 May 2025 at 00:52, Thorsten Glaser <tg@debian.org> wrote:
> >then go up. If I run "wc" on a copyrighted work, the number of words
> >in the document is *not* a derived work from the original document.
>
> If you JPEG-compress a photo of the original document then uncompress
> it, it *is*.


Please, restore a document from the output of "wc". Or from the output
of sha256sum. An average or aggregate is data mining and is exempt
from copyright.

Even substantially similar and recognisable, automated data
transformations have been ruled as fair use. For example:
https://www.pinsentmasons.com/out-law/news/google-thumbnails-are-fair-use-says-court-of-appeals

A very simple creation of a thumbnail from a larger image and then
redistributing that thumbnail (even if the individual, original work
is still recognisable there) was ruled as fair use. Even in commercial
use. Because the purpose was sufficiently transformative. The purpose
of an LLM for generating new works is *far* more transformative than
making a thumbnail.

And on EU, ilulu has already replied as well.
-- 
Best regards,
    Aigars Mahinovs