AI Models Training is Fundamentally Flawed

Why do near-perfect AI models from labs often fail in the real world? A group of 40 researchers across seven different teams at Google have identified a second major cause for the common failure of machine-learning models.

The way AI models are currently being trained is fundamentally flawed. The process used to build most of the machine-learning models we use today can’t tell if they will work in the real world or not. And there lies the problem. Data shift is a known problem but now there’s a new one called underspecification.