Large Language Models vs Large Language Machines

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This work introduces the Large Language Machine (LLMA), an experimental approach to language processing based on explicit memory and algorithmic inference rather than learned neural weights. A minimal LLMA prototype stores observed language sequences and retrieves or generalizes from them through an explicit, inspectable heuristic. The experiment compares this architecture with a deliberately small Transformer trained on the same text, with the aim of exploring the boundary between memory, retrieval, generalization, and inference. This first publication documents the architecture and experimental setup before comparative testing, together with the minimal source code used in the experiment.