We need to discuss how we can correct for biases in the training data of AI alignment teams

→ View original post on X — @plinz