1 comment

[ 0.16 ms ] story [ 18.9 ms ] thread
the important part of hype LLMs like gpt-4 (the pre-training) is unsupervised