x
AI interpretability could be harmful? — AI Alignment Forum