SHRDLU was the 1968 AI system that identified objects and navigated a simulated blocks world.
Terry Winograd created SHRDLU at the Massachusetts Institute of Technology as part of his doctoral research. Users could type natural-language commands and questions, such as asking the system to move a block or describe what was on top of another object.
The program combined a language parser with a symbolic representation of a small virtual environment. Because the world was limited and precisely defined, SHRDLU could connect words with objects, actions, and relationships more effectively than general-purpose language systems of its era.
Its apparent fluency created excitement about machine understanding, but later analysis showed why the achievement did not generalize easily. SHRDLU’s knowledge was tightly tied to the blocks world, so it lacked the broad background knowledge needed for ordinary conversation. It is therefore remembered both as a landmark and as a lesson about the limits of narrow symbolic systems.