Did the Hardest A.I. Problem Just Get Much Easier?

A prominent researcher claims to have a revolutionary approach to “interpretability,” or the science of why A.I. models do what they do. Other experts aren’t so sure.

Summary from New York Times