NewsDialy
Distillation, a process that takes the knowledge embedded in a large AI model and transfers it into a smaller, cheaper one, has moved out of research circles and into the center of a live policy fight.
Silicon Valley and Washington, D.C. are now arguing about how the technique should be regulated. Technology executives and lawmakers are both at the table. What distillation actually means Start with the cost problem.
A large AI model is expensive to build and expensive to run. Every query consumes computing resources. Distillation addresses that by producing a smaller model that learned directly from the larger one.
A capable model, called the teacher, generates outputs across a wide range of tasks. A smaller model, called the student, trains on those outputs rather than on the original data.
Keep reading