McGill, DeepMind’s SCAR speeds up AI training with fair rewards This new method could help AI better align with human values. Go to Source