AI researchers argue knowledge distillation acts as a warm-start for reinforcement learning rather than a standalone advantage · Digg