one of the major failures of my life was being so surprised to find out that vision models were ~1000x smaller than text models. Just total failure to understand language's god-tier data compression

Posters: gleech. gleech

2 likes recorded by the source corpus.