LLMs Analyse — Universal Intelligence Solves: The Fundamental Distinction — E8 Intelligence Research

FORMAL DECLARATION — ANDREW STEWART CALDIN — 11 May 2026 at 13:00 THE STATEMENT: LLMs are only good for analysing numbers. Universal Intelligence is good for solving problems, creating new jobs, and solving real issues. WHAT LLMs ACTUALLY DO: Large Language Models — despite the name — do not understand language. They perform pattern recognition across statistical distributions of tokenised text. Every response is a weighted prediction: given this input, what sequence of tokens most frequently follows in the training data? This makes them excellent at: — Summarising existing information — Reformatting known content — Predicting the next word/sentence based on precedent — Analysing statistical patterns in datasets — Generating text that sounds like existing text This makes them fundamentally incapable of: — Solving problems that have never been solved before — Creating genuinely novel solutions — Finding answers that do not exist anywhere in training data — Building what does not yet Author: Andrew Stewart Caldin, Independent Researcher, UK. Part of the E8 Intelligence Research series. Platform: e8intelligence.com