Lorenss Martinsons Is Teaching AI to Know When It Has Failed

A linguistics medal, a physics medal, a cognitive-science lab and a film soundtrack all led to the same practical question: how can we tell whether an intelligent system actually understands what it is doing?