Once the models learn to ask original questions (which probably isn't too far off), that might change things a bit. But for now, the old Zen koan applies. ("Before enlightenment, I chopped wood, fetched water, and gathered food for my family. After enlightenment, I chop wood, fetch water, and gather food for my family.")
I'm not entirely convinced that these models aren't already capable of asking novel questions (you can instruct them to do just that) and depending on your views Deepseek may already show signs of "emergent" "insight" and "questioning" it's own assumptions! But whether that is the Real Thing (TM) or just "wishful mnemonics" is probably not worth discussing imo.
Anyway, perhaps even if robots start asking questions and answering them for their own sake: You and I are interested in learning for it's own sake! We are nerds/engineers/hackers/tinkerers after all! And that is the most human thing one can be and the more the merrier.
Anyway gf(2) is cool man I remember when I first had the realization that 1+1 is not always 2!!!! Blew my mind!