[tldoc] TeXLive plans in writing the documentation: will be human or chatgpt generated text?

[Carlos]

On Fri, Jun 09, 2023 at 06:38:39PM -0700, Boris Veytsman wrote:
> C> From: Carlos <linguafalsa at gmail.com>
> C> Date: Fri, 9 Jun 2023 10:56:02 -0400
> 
> >> 
> >> As far as I know there are no plans of using LLMs in documentation
> >> writing or code generation for TUG.  If these projects appear, it
> >> would be interesting to see the results - and we cannot exclude the
> >> possibility they might be useful.  Especially if we can solve the
> >> hallucination problem.
> 
> C> What results specifically would you expect to find? Karl wrote earlier:
> 
> 
> I have no idea. I just meant that if (1) someone applies to Devfund
> with a project involving LLMs,

If you have a fund with a wishlist layout set in place, and don't have an idea, how would you expect me to have one.

> and (2) we will find this project
> interesting and feasible, we probably would fund it.  Nothing
> different from a project involving any other technology.

Some of the items from the wishlist are possibly beyond what llm's can
do anyway. And I'm just speculating now, but if llm's are not smart
enough even with additional human input for some of these wishlist
items on TeX, perhaps you ought to try your own recipes from the same
output you'd expect to find from more ‹modest› tools (such as the
one as outlined on the TuG paper you authored recently and linked on
this thread earlier). Perhaps some, some, not all of these results
found therein would shed light on what can be and cannot be combined
with llm's.  But I don't know.

Wishlist
Some substantial projects we'd be most pleased to see undertaken; see also how to help the TeX community.

Collaborative editing support in a TeX front end, such as TeXworks.
Direct graphic support for more images in Dvips (jpg, tiff, etc.), i.e., interpreting bmp: specials a la PCTeX, perhaps using the bmeps library.
More free documentation. Much of the TeX system still has no comprehensive free documentation.
More math fonts. Only a few typefaces are available, even commercially, for math typesetting. A Braille font would be useful for the low-vision community.
Page breaking: apply something like the same optimization to breaking pages in the document as is done for breaking lines in a paragraph. See Michael Plass' thesis Optimal pagination techniques for automatic typesetting systems, Stefan Wohlfeil's “On the pagination of complex documents”, Paolo Ciancarini et al.'s “High-quality pagination for publishing” (SPE 42:6), and discussions on texhax of the subject. Perhaps something can be done by having TeX output the boxes in the log file with \showlists, and then writing a separate program, as described in Jonathan Fine's article TeX Forever in the EuroTeX 2005 proceedings. (Research level.)
Line breaking: an extended implementation of TeX's algorithm was implemented by Alex Holkner in his thesis Global Multiple Objective Line Breaking, as an approach to minimizing rivers, widows, hyphenations, and more (hence the “multiple objective”). The “global” is because he optimizes over the whole document, which could point toward an implementation for page breaking, as above.



> 
> I would like to use this occasion to remind people about Development
> fund.  If you have ideas about TeX and friends, welcome to apply!  See
> the details at https://tug.org/tc/devfund/.
> 



-- 
Digital circuits are made from analog parts.
		-- Don Vonada