The illusion of a perfect metric: Why evaluating AI's words is harder than it looks

Open in new window