Will Machine Learning Outgrow Human Labeling?
by Mike Schaekermann, Christopher M. Homan, Lora Aroyo, Praveen Paritosh, Kurt Bollacker, Chris Welty
Some machine learning (ML) rhetoric seems to imply an assumption or expectation that, at some point, machines will outgrow the need for human labeled data. Today’s reliance on such labeling is a sort of dirty little secret of AI, and some view it as a necessary means to a larger end. This bet is an attempt to formalize that attitude into a concrete question, whose answer can be measured over time.