Validators of LLMs..
Validators( Non-Deterministic) itself needs Validators! This also seems to be a Self-Referential issue
Not to forget, Self-Referential issue is the deeply related to Godel's Incompleteness..
That also raises the fundamental question if LLMs would be self-consistent or complete ?
Recently I had written that this system of training -testing (followed by AI/ML? seems to suffer from another fundamental issue where even InSample, Out-sample selection could itself be actually InSample at higher level and biased !
So, this whole system of Train-Test method used by AI/ML in general could also be fundamentally inconsistent/ incomplete! That could also make AI/ML so !
No comments:
Post a Comment