04
The chess machine with a man folded inside it
For eighty-four years an automaton beat the best players in Europe. Everyone could see there was no room for a person in the cabinet, because they were shown that there was not.
6 min read
In 1770, Wolfgang von Kempelen presented a machine to the Habsburg court: a wooden figure in Ottoman dress seated at a cabinet, which played chess. Well.
It toured Europe and America for eighty-four years, beat Napoleon and Benjamin Franklin, and was examined by many people who concluded it was genuine.
There was a man inside.
The demonstration was the deception
Before each game the operator opened the cabinet doors, one section at a time, and held a candle inside to show the gearing. The audience saw clockwork, and empty space where a person would have to be.
They saw it because the operator inside slid between compartments as the doors opened in sequence, on a rolling seat, behind a screen that moved with him.
The reveal was not a risk taken despite the secret. The reveal was the mechanism. Nobody who had not seen inside would have been half as convinced.
The most persuasive part of any illusion is the moment you are invited to check.
Why people believed a machine could do it
Partly the age. The eighteenth century was producing genuinely astonishing automata — mechanical ducks, writing boys, musicians with working fingers. A chess machine was a step further along a road people could already see.
And partly because the alternative was insulting: to say it was a fraud was to say you had been fooled by a cabinet, in public, along with everyone else.
Edgar Allan Poe, more or less right
In 1836 Poe published an essay arguing the Turk was operated by a hidden human. His reasoning was mostly behavioural rather than mechanical — the machine did not play with the invariance of a machine. It took different lengths of time. It made mistakes. It sometimes lost.
A real automaton, he argued, would play the same way every time.
He got some of the internal details wrong. The core inference was right, and it was reached by watching how the thing behaved rather than by prising it open.
Why it belongs on this site
Because the same shape reappeared two centuries later, twice.
Amazon named a service after it — a marketplace where tasks that look automated are performed by people, paid per task. The name is an admission, and a fairly candid one.
And it is the recurring pattern behind a certain kind of technology claim: a demonstration presented as autonomous, with humans in the loop who are not mentioned. It has a name now — the phrase is pseudo-AI or, less politely, AI washing — and regulators have brought cases about it.
The lesson from 1770 still applies. Do not evaluate the box. Evaluate the behaviour, over time, on cases nobody prepared for.