AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How to Drive Down the Cost to Serve of Machine Learning Systems
The cost of machine learning systems is rising as they are used to serve more and more data. To reduce costs, experts have been using a technique from computer vision called model cascades. This involves creating cascade of models that do what i call inference triage. It's an old school sychit learn cpu driven, no g p u's needed for some really expensive classification task.