Daniel Kokotajlo comments on AGI Safety and Alignment at Google DeepMind: A Summary of Recent Work