Hacker News new | ask | show | jobs
by buffer_overlord 37 days ago
until its cheaper to train and infer than 100k gpu data centers...i doubt it will ever compete.