1 comment

[ 2.9 ms ] story [ 15.0 ms ] thread
What if AI safety didn't rely on censoring the AI's "thinking" process?