Post by Quantіan on X
Quantіan@quantian1
XIt hasn't really been reported as such, but DeepSeek is the final nail in the coffin for the popular but obvious dumb "Hapsburg AI" theory: once a LLM reaches a certain point, having its own output added to the training set improves, rather than degrades, performance
505 likes22 repliesPosted Jan 27, 2025