A personal essay by the author of the 1975 PhD dissertation "An N-Space Model for Visual and Verbal Concepts," reflecting on how that work — which represented word meanings as vectors in a multidimensional semantic space — anticipated core ideas now found in modern large language models, including distributed vector semantics, transformation-based composition, and execution-based grounding. The essay traces the arc from the work's dismissal at the 1974 ACL conference, through fifty years of parallel development in the field, to the author's recent return to the ideas in the form of LATN (Layered Augmented Transition Networks), a modern Python reimplementation that extends the original architecture into explicit syntactic and grounding layers. Written for a general technical audience.
Jeffrey F. Eastman (Sat,) studied this question.