The Mechanical Turk

History of AI — The Mechanical Turk
History of AI — Imagined AI
Course 1 · Chapter 3 · The Mechanical Turk

♟️ The Mechanical Turk

Chapter 2 ended on Vaucanson producing both a genuine mechanism and a staged fake within the same year. This chapter is dedicated entirely to the most famous fake of all — a "thinking machine" that toured Europe and America for over 80 years, defeated some of the most powerful people of its era at chess, and fooled almost everyone, almost the entire time.

🎩 What Von Kempelen Built (1770)

Wolfgang von Kempelen, a Hungarian engineer, unveiled an automaton to Empress Maria Theresa of Austria: a life-sized figure dressed in robes and a turban, seated behind a wooden cabinet, one hand resting on a chessboard built into the cabinet's top. Before each demonstration, Von Kempelen would open the cabinet's doors in sequence, revealing an interior densely packed with gears, levers, and machinery — apparent proof there was no room for a person hidden inside.

The Turk then played chess. Genuinely well — well enough to defeat most challengers, occasionally pausing to shake its head or tap the board in response to an illegal move.

The Trick

The cabinet's machinery was a stage prop — real gears and levers, but arranged to leave a concealed compartment just large enough for a skilled human chess player to sit, hidden, sliding silently within the cabinet as each door was opened in turn so the occupied section was never the one currently on display. The hidden operator watched the board through a mechanism connected to magnets beneath the chess pieces, and controlled the Turk's arm directly by hand, inside the machine.

🌍 Decades of Deception

The Turk toured for over 80 years (1770–1854), changing owners several times, reportedly defeating Benjamin Franklin and playing against Napoleon Bonaparte (a game the Turk is said to have won, after Napoleon reportedly attempted illegal moves to test it). It survived Von Kempelen's death, was purchased and toured extensively by Johann Nepomuk Mälzel, and was eventually destroyed in a fire in Philadelphia in 1854 — its secret still not definitively publicly confirmed at the time.

⚠ Even Skeptics Struggled to Prove It

The Turk wasn't unquestioned — many contemporaries suspected trickery, and several published essays and pamphlets attempting to explain the mechanism, with varying accuracy. Edgar Allan Poe wrote a lengthy 1836 essay, "Maelzel's Chess Player," reasoning carefully from observed behavior toward the (correct) conclusion that a human operator must be involved — real chess playing, Poe argued, could not follow the purely mechanical, deterministic logic he expected from an actual machine, because it varied and adapted too much between games. Poe's specific reasoning about "true machines never vary" was itself somewhat flawed logic, but his conclusion happened to be right.

🤔 Why Did This Fool Everyone For So Long?

The Turk's deception worked through a combination of clever engineering (a genuinely well-hidden compartment, exploiting how doors were opened in sequence rather than all at once) and something less mechanical: audiences wanted to believe a machine could think. A chess-playing automaton, arriving at the height of Enlightenment fascination with mechanism and reason, offered a spectacle people were primed to find plausible — much as Vaucanson's Duck had been a decade earlier for a less intellectually demanding claim.

This is the chapter's central lesson, and arguably one of the central lessons of the whole "Imagined AI" course: a sufficiently convincing demonstration of apparent intelligence tends to get believed, even by careful and skeptical observers, long before anyone can actually explain how it works. That gap — between "this appears intelligent" and "I understand the actual mechanism producing that appearance" — recurs constantly throughout the real history of AI covered in Courses 2 and 3, right up to modern debates over whether a large language model "understands" anything at all.

📦 A Deliberate Modern Echo: Amazon Mechanical Turk

🔮 The Name Was Reused on Purpose

In 2005, Amazon launched a crowdsourcing platform called Amazon Mechanical Turk — a marketplace where businesses submit small tasks ("Human Intelligence Tasks") that are completed by real, paid human workers behind an API that looks, to the requesting software, like a single automated service. The name isn't a coincidence or a stretch: Amazon deliberately named it after Von Kempelen's automaton, because the underlying pattern is genuinely the same one — a system that presents itself as automated while real human labor operates, unseen, behind the interface.

This pattern didn't stop at Amazon's naming choice, either. Multiple times in the years since, startups marketed as "AI-powered" have been revealed to actually rely heavily on hidden human workers performing tasks manually behind an automated-looking front end — a direct, if usually unintentional, echo of the 18th-century original.

Von Kempelen's Turk (1770)Amazon Mechanical Turk (2005)
What's presentedAn automaton that plays chessAn API that completes tasks
What's actually happeningA hidden human chess master, operating the machinePaid human workers, completing tasks behind the API
Was the deception intentional?Yes — the whole point was to appear automatedNo — openly marketed as human-powered crowdsourcing

That last row is the meaningful difference: Amazon's version is openly, honestly human-powered — the name is a wink at history, not a repeat of the original deception. The genuinely uncomfortable modern echoes are the "AI startups" that weren't upfront about the humans behind the curtain.

🤔 Questions to Sit With

Reflection 1

Poe correctly concluded a human was involved, but reasoned from a flawed premise (that a "true machine" could never vary its play). Is a correct conclusion reached through flawed reasoning still worth taking seriously — and does the answer change depending on the stakes involved?

Reflection 2

The Turk's operators changed the hidden chess master over the decades, meaning the "automaton" was, in a sense, run by many different real experts over its lifetime. Does that make the Turk more or less impressive as a piece of engineering, compared to if a single genuine machine had actually played chess?

Reflection 3

Amazon named their platform after a famous historical hoax, openly and knowingly. Why do you think they made that choice, rather than picking a name with no association to deception at all?

🎯 What's Next

Next chapter: Birth of "Robot" — Karel Čapek's R.U.R. (1920) coins the word itself, and how Frankenstein's Creature a century earlier already explored many of the same anxieties.