Isn’t it more a problem of having algorithms do what used to be done by humans?

I.e. when you look at a Google map it is using the latest GIS data to construct a “good” representation for your screen resolution using your specified zoom factor and possibly even with dynamic overlays or highlights.

Likewise, when you read an e-book you can read it on many different display sizes, pick your own font, and font size. So where we used to have the book typeset once by a person, it is now done on-the-fly using user and device specific parameters.

The result won’t be as good as when you have a human do it, and even when a human does it, there might be limitations in the e-book markup language, just like with HTML.