1 comment

[ 3.6 ms ] story [ 14.4 ms ] thread
What if AI safety didn't rely on censoring the AI's "thinking" process?