What 'training' a model really means
No magic: it guesses, checks, and nudges numbers a billion times.
Discussion · 1.3K
@ml_curious6h · ♥ 200
so where do the weights actually live, like physically
so where do the weights actually live, like physically
@samcodes✓6h
Just numbers in memory and on disk, big grids of them. That grid is the whole model.
Just numbers in memory and on disk, big grids of them. That grid is the whole model.
@worried_wendy4h · ♥ 90
does this mean it cannot be creative? genuine question not a gotcha
does this mean it cannot be creative? genuine question not a gotcha
@samcodes✓4h
Depends how you define creative, it recombines patterns it has seen. Great debate for the comments.
Depends how you define creative, it recombines patterns it has seen. Great debate for the comments.
@hype_harry2h · ♥ 25
idk it still feels like magic to me
idk it still feels like magic to me
Join the discussion in the Circin app — ask questions and get answers from the creator.
No, it isn't 'thinking'. It adjusts weights to reduce error. Ask the hard questions below.