[philipphutterer]

Or given the number of unscanned books, even just give it the controls for a book scanner, the books and probably some robot arms. Then let it figure out the scanning first in some layers. Shouldn't be that hard.

[whimsicalism]

RGB->YUV is literally an affine transform, of course it falls to the bitter lesson.

[Llamamoe]

Does it? Because I'm not sure the model has any intrinsic incentive to learn to follow how human perception works.