The “Good Old Days” of Question Answering: BASEBALL, LUNAR, SHRDLU, and the Abbott & Costello Problem
Long before large language models could answer almost any question with near-human fluency, question-answering (QA) systems were narrow, rule-based, and domain-specific. In the 1960s and early 1970s — the golden age of symbolic AI — researchers built impressive but brittle natural-language interfaces to structured knowledge. These systems worked beautifully within their tiny worlds… and failed […]









