High-level hopes for AI alignment — AI Alignment Forum