Rohin Shah published a short summary of the CHAI paper “On the Feasibility of Learning, Rather than Assuming, Human Biases for Reward Inference”, along with some discussion of its implications on the Alignment Forum.