1 comment

[ 4.7 ms ] story [ 14.2 ms ] thread
What if AI safety didn't rely on censoring the AI's "thinking" process?