Comment by Animats

Worse, the massive data set may not help much with mistakes. Larger LLMs do not seem to hallucinate less.