The Next Word Problem
Every machine in this course is asked exactly one question: what word goes in that blank? What changed over a century is not the question but the evidence admitted. The first models saw nothing and returned raw frequency. Later generations saw one word, then a window of words, then everything through fading memory, then everything at once through attention, and finally paused to think before answering. Below, the same prediction table at six moments in history. Watch the distribution sharpen; that sharpening is the entire field.