Interpretability

Neural networks are famously incomprehensible, so Been Kim is developing a “translator for humans.”

A new idea is helping to explain the puzzling success of today’s artificial-intelligence algorithms — and might also explain how human brains learn.