How do you teach a computer what a 'chair' is - when even humans can't fully define one?

The answer

Knowledge representation and reasoning (KR&R) is the foundational AI challenge of encoding information about the world in a form that computers can use for intelligent behavior. Early approaches included semantic networks (Quillian, 1968), frames (Minsky, 1974), and formal logic systems, each attempting to capture the web of concepts, relationships, and rules that humans use effortlessly. The Cyc project, launched by Doug Lenat in 1984, attempted to manually encode all common-sense knowledge - after over 40 years and millions of assertions, it remains incomplete. The difficulty lies in the sheer volume and context-dependence of human knowledge: a chair can be a piece of furniture, a position of authority, or a verb meaning to lead a meeting, and humans switch between these meanings instantly using background knowledge that proved extraordinarily hard to formalize.