Xuenan Cao

Assistant Professor of Cultural Studies at the Chinese University of Hong Kong, and author of the 2025 arXiv paper "How Large Language Models (LLMs) Extrapolate: From Guided Missiles to Guided Prompts" (arXiv:2501.10361).

Matters to this vault as the author of the primary source that finally gave Norbert Wiener's WWII-era prediction work a concrete, documented technical descendant in modern AI: her paper argues that LLM "hallucination" is extrapolation, and that the underlying mathematics is the same apparatus Wiener carried, largely unmodified, from classified missile-trajectory prediction into communication engineering in 1941-42 (claim-cao-2025-traces-llm-hallucination-to-wieners-1941-42-missile-extrapolation-math). She also credits Andrey Kolmogorov's independent, near-identically-titled 1939 extrapolation theory as a parallel discovery (claim-kolmogorov-1939-parallel-extrapolation-theory-cao-2025). Unknown to the vault before this session.

References

• claim-cao-2025-traces-llm-hallucination-to-wieners-1941-42-missile-extrapolation-math
• claim-kolmogorov-1939-parallel-extrapolation-theory-cao-2025