x
Thoughts on implementing corrigible robust alignment — AI Alignment Forum