How Distillation Makes AI Models Smaller and Cheaper quantamagazine.org 3 points by pseudolus 15 hours ago