Ancient Dreams of Thinking Machines

Long before electricity, people imagined artificial servants, built moving figures, and wondered whether reasoning itself could be done by a machine. This page separates the myths from the machines — and shows which old ideas still shape AI.

A bronze giant strides around the coast of Crete three times a day, hurling rocks at approaching ships. He was made, not born, and the single vein that carries his life runs down to his ankle, sealed with a nail. When the sorceress Medea finally defeats him, she does it not with strength but by outwitting him. The story of Talos is well over two thousand years old, yet it already contains the questions we ask about AI today: who builds these beings, what are they for, and can they be controlled?

Myths of made beings

Ancient Greek poets imagined a whole workshop of artificial helpers. In Book 18 of Homer’s Iliad, the smith-god Hephaestus is attended by servants made of gold that look like young women. Homer says they have understanding, speech and strength, and that they learned their crafts from the gods. In the same scene Hephaestus is building tripods on golden wheels that roll by themselves to the gods’ assemblies and back. These are some of the oldest descriptions in literature of something like a robot: made objects that move and act on their own.

Talos belongs to the same tradition, though he is not in Homer. His best-known telling is in Apollonius of Rhodes’ Argonautica (3rd century BCE); other versions credit Hephaestus or the inventor Daedalus with making him, and he appears on Greek vase paintings from around 400 BCE. The classicist Adrienne Mayor, who gathered these stories in Gods and Robots (2018), points out that the Greeks thought hard about what artificial life would mean — including its dangers.

The Jewish legend of the golem asks the same questions from another angle. The Babylonian Talmud (Sanhedrin 65b) tells briefly of a rabbi who creates a man who cannot speak. Later folklore told of golems shaped from clay and brought to life with sacred words, and of the danger when they grew too strong. The most famous version places a golem in 16th-century Prague, made by the great scholar Rabbi Judah Loew ben Bezalel (died 1609) to protect the Jewish community.

Popular myth The Prague golem is often presented as a story told since Rabbi Loew’s own day. In fact, no known source links him to a golem until the 1830s and 1840s, more than two centuries after his death, and the popular account that claimed to come from his lifetime — Yudl Rosenberg’s The Wonders of the Maharal (1909) — is regarded by scholars as a literary invention. Earlier golem stories were told about another rabbi, Elijah of Chełm.

China has its own story. The Liezi, a Daoist text, describes an engineer named Yan Shi presenting King Mu of Zhou with a life-size figure that walks, sings and dances. When it flirts with the king’s ladies, the furious king orders Yan Shi executed — until the engineer takes it apart to show it is made of leather, wood, glue and lacquer. Debated The Liezi is traditionally credited to a philosopher of the 5th century BCE, but most scholars think the text we have was compiled around the 4th century CE, so the story tells us about imagination in that later period, not about real Zhou-dynasty engineering.

Real machines that moved

Alongside the myths, real engineers built self-moving machines. In Roman Egypt, Hero of Alexandria (probably 1st century CE) described temple doors that opened when a fire was lit, a coin-operated machine that dispensed holy water, and a miniature theatre of moving figures. One of his carts was driven by a falling weight on a string wound around its axle; by changing how the string was wound, the builder could change the cart’s path. Some engineers today describe this as a programmable machine — the “program” is the pattern of the string.

In 1206 the engineer Ismail al-Jazari, working for the Artuqid rulers in what is now south-eastern Turkey, completed The Book of Knowledge of Ingenious Mechanical Devices. It describes some fifty machines with careful instructions and drawings: water clocks (including the famous elephant clock), pumps, a hand-washing device with a figure that offers soap and towel, and a boat of mechanical musicians to entertain guests. Debated Some modern writers call one of his musical devices, whose drummer was driven by pegs on a rotating wheel, an early programmable robot; others think that stretches the word, since changing the pattern meant rebuilding parts.

Europe’s great age of automata came in the 18th century. Between about 1768 and 1774, the Swiss watchmakers Pierre Jaquet-Droz, his son Henri-Louis and Jean-Frédéric Leschot built three figures — the Writer, the Draughtsman and the Musician — that still work today in the Museum of Art and History in Neuchâtel. The Writer dips a quill in ink and writes a message of up to forty characters; by changing the cams on a wheel inside, you can change the text. That is a genuine stored instruction sequence, nearly two centuries before electronic computers.

Myth, real machine, or trick? Tap each tile.

The great hoax: the Mechanical Turk

In 1770 the Hungarian-born inventor Wolfgang von Kempelen unveiled a life-size figure in Ottoman dress, seated at a cabinet with a chessboard, for the Empress Maria Theresa in Vienna. It played chess — and usually won. For more than eighty years “the Turk” toured Europe and America, beating many challengers; it is reported to have played Napoleon in 1809. Before each show the operator opened doors to reveal gears and wheels. But the gears were a decoy. A skilled human player sat hidden inside, following the game through magnets under the board.

The Turk was destroyed by fire in Philadelphia in 1854. It matters to AI history for two reasons. It shows how eager audiences were to believe a machine could think — and how easily a clever presentation can hide a human doing the real work. Amazon later named its crowdsourcing service “Mechanical Turk” for exactly that reason: tasks that look automated are done by people. The question “is there a person behind this?” is worth asking about AI products even today.

Can reasoning be calculated?

Automata imitated the body. A separate line of thinkers asked whether the mind could be mechanised — whether reasoning itself could be turned into steps a machine could follow. This is the deeper ancestor of AI.

ARSRamon LlullMajorca, from the 1270s

The Majorcan philosopher Ramon Llull designed his Ars (“Art”), developed over decades and summed up in the Ars generalis ultima (1305–08). It used paper discs that rotated against each other, each marked with letters standing for basic concepts. Turning the discs produced every combination, which Llull hoped would generate all true statements about God and persuade non-Christians by logic. It was not a calculator, but it was an early attempt to produce reasoning by a mechanical procedure of combining symbols.

1651Thomas HobbesReasoning as reckoning

In Leviathan (1651) the English philosopher Thomas Hobbes argued that thinking is a kind of calculation: “REASON, in this sense, is nothing but Reckoning (that is, Adding and Subtracting)”. If he was right, a machine that could add and subtract the right things could, in principle, reason. That idea sits at the root of symbolic AI.

∫Gottfried Wilhelm LeibnizThe calculus ratiocinator

Leibniz, co-inventor of calculus, admired Llull and named his 1666 dissertation after Llull’s art of combination. He dreamed of a universal language of thought, the characteristica universalis, and a calculus ratiocinator to compute with it. Disputes, he wrote, could then be settled by saying “Let us calculate.” He also built a real machine: the stepped reckoner (designed from 1672), which could add, subtract, multiply and divide. He championed binary arithmetic, too — the number system all digital computers use.

1637René DescartesThe sceptic’s test

In his Discourse on the Method (1637) Descartes argued that animal bodies are machines — but that a machine could never truly think. He offered two tests a machine would fail: it could not arrange words flexibly to reply sensibly to whatever is said to it, and it could not act well across the endless variety of life’s situations, since each of its parts is set up for a particular action. More than three centuries later, Turing’s imitation game would turn the first of these tests around.

This guide’s framing Notice the two separate dreams here: machines that behave like living things (automata) and procedures that reason like minds (Llull, Hobbes, Leibniz). Modern AI is what happened when computers finally joined them.

Fiction asks the hard questions

In 1818 Mary Shelley published Frankenstein; or, The Modern Prometheus. Victor Frankenstein builds a living being from dead matter and then abandons it in horror; the creature, articulate and lonely, turns to revenge. The novel is not about machines, but it gave the modern world its master story about creation: the maker who fails to take responsibility for what he made. That theme runs through almost every later debate about AI safety and alignment.

In 1920 the Czech writer Karel Čapek wrote the play R.U.R. (Rossum’s Universal Robots), first staged in Prague in January 1921. His artificial workers are made of synthetic flesh, not metal, and they eventually rise up against humanity. Čapek credited his brother Josef with the word robot, from the Czech robota, meaning forced labour. It spread into English and many other languages within a few years. The robots of R.U.R. fixed a picture in the public mind — the machine servant that rebels — that still colours how people react to AI.

Quick check

Where do Hephaestus’ golden attendants, with understanding and speech, appear?

They appear in Book 18 of the Iliad, in the scene where Hephaestus forges new armour for Achilles.

How did the Mechanical Turk really play chess?

It was a hoax. The cabinet’s visible gears were a decoy for the hidden player.

Who argued that reasoning “is nothing but reckoning”?

Hobbes, in Leviathan (1651). Descartes took the opposite view: that machines could not truly reason.

Where does the word “robot” come from?

Robota means forced labour; Karel Čapek credited his brother Josef with the coinage.

// Pause and reflect

  1. Talos, the golem and Čapek’s robots are all servants that become dangerous. Why do you think that story keeps returning across such different cultures?
  2. The Mechanical Turk fooled audiences for decades. What would you check today before believing that a product is “fully automated by AI”?
  3. Hobbes said reasoning is calculation; Descartes said a machine could never reason. Which side do today’s chat assistants support — or do they unsettle both?

Your notes are saved privately in this browser.

What happened next

Leibniz’s calculating machine could do arithmetic, but it could not follow a program. In the 1830s an English mathematician designed a machine that could — and a young woman wrote the first serious account of what it might mean. Continue with Babbage, Lovelace and the Engine, or step back to AI History at a Glance.

Key terms on this page

Sources & further reading