https://bugs.documentfoundation.org/show_bug.cgi?id=37134

--- Comment #114 from Beedell, Roke Julian Lockhart 
<[email protected]> ---
(In reply to Konstantin Kharlamov from comment #113)
> (In reply to Senna-sport from comment #112)
> > And why not ask Generative AI Developer systems to do so to accelerate and
> > help the community? (This is just a comment from someone - me - who is not a
> > developer, but the most important is the creation and implementation of the
> > feature itself as expressed above in the two points).
> 
> Writing code is the least problematic part of implementing any big feature.
> You'd need to study the existing codebase and figure out how to connect
> various parts of the code to make sure it's both working and robust, and as
> of now AI wouldn't help with that.

Yeah, non-local LLMs currently can't process massive token counts, so they
can't process a codebase of the size of LO's. They're better for single files.
Irrespective, they're merely (subsequent) word probability predictors, so they
generally create a bare minimum functional result; not robust code.

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to