What the Analytical Language Engine is: an n-gram Markov model. The model performs next-word prediction from frequency tables of one, two, and three-word contexts, built in a single pass over the 79-sentence library. Map the cabinet onto computing vocabulary: tokenisation, lookup tables, a feedback loop (autoregression). Then draw the boundary: a real LLM shares this pipeline but replaces counting with a trained neural network — noted at the engine's own About page. Key distinction for this band: in this machine the "program" was written by counting; in a modern model the parameters are learned, and no human wrote them.
This is conditional probability in the wild: P(next word | previous words), estimated by relative frequency. The engine blends three estimates with fixed weights — a weighted average where the weights sum to 1 — and backs off to shorter contexts when the longer one was never seen. Temperature becomes a real function:
Sampling is inverse-CDF: divide the interval [0, 1) into segments sized by probability, draw a uniform random number, take the segment it lands in — which is literally what the lottery drum's pointer does. Students can verify the machine's arithmetic by hand: pick a two-word context, count its continuations in the library (view the page source — the corpus is right there), and compare with the gauges.
Add the engineering and the economics. Babbage — Lucasian Professor of Mathematics at Cambridge, Newton's old chair — consumed roughly £17,000 of government money on the Difference Engine, at a time when a locomotive cost under a tenth of that, and delivered no finished machine; the modern Science Museum build vindicated the design, so the failure was one of manufacturing economics, project management, and Babbage's own restless redesigning. The Analytical Engine deserves its architectural due: separate Mill and Store, punched-card programs, conditional branching, loops — a general-purpose computer on paper a century before electronics. Lovelace's 1843 Notes (signed only "A.A.L.") stand as the first substantial writing about computing, including her leap past arithmetic: an engine acting on any symbols with rules — even musical ones — "might compose elaborate and scientific pieces of music." That sentence is the conceptual birth of the computer as more than a calculator.
Read the objection in full, then meet its most famous reply. In 1950, Alan Turing — treating "Lady Lovelace's Objection" by name in Computing Machinery and Intelligence — answered on two fronts: machines surprise their own makers constantly; and a learning machine's behaviour grows from training the way a pupil's grows from teaching, so "it only does what we order" blurs once the orders are learned from data rather than written by hand. Now aim both arguments at the cabinet: our trigram engine is exactly the machine Ada described — it does precisely what the counted data orders. A trillion-parameter model learns its own parameters. Structured debate: Resolved: the difference between the cabinet and a modern LLM is one of degree, not of kind.
Devised by Messrs. Horrocks & Claude · Oldham, Lancs. · © MMXXVI