Knowledge representation is the area of AI concerned with structuring information in a formal, machine-readable way, so that a system can reason over it, draw inferences from it, and apply it to solve problems rather than just storing it as inert data. Common approaches include semantic networks, which represent concepts as nodes connected by labeled relationships, logic-based rules, and structured knowledge graphs.
ChatGPT and other large language models take a fundamentally different approach from classical knowledge representation: instead of explicit, hand-built structures, the model's "knowledge" is encoded implicitly across billions of learned parameters, developed from patterns in training text rather than curated facts. This gives language models remarkable flexibility across topics, but it also means they lack the kind of clean, explicit knowledge structure that makes classical knowledge representation systems easier to verify, audit, and reason about explicitly.
