A history of thinking machines
AI didn't arrive from nowhere.
It is not an alien landing, a magic trick, or a deception. It is the latest chapter in a story more than two thousand years old — of people, patiently, teaching things to count, to reason, and at last to help us think. Walk the river of it, and the fear tends to fall away.
scroll to walk the timeline ↓
Long before circuits, people dreamed of thinking machines — Talos the bronze guardian, Hephaestus' mechanical helpers. And the dream was already practical: the Antikythera mechanism, a bronze Greek device of over thirty gears, could predict the movements of the heavens.
The wish to build a helper is as old as myth itself.
Blaise Pascal, aged nineteen, built a brass machine to add and subtract for his tax-collector father. Leibniz went further with a machine that multiplied — and dreamed of a “calculus of reasoning” that might one day settle arguments by calculation.
“Let us calculate,” said Leibniz — three centuries early.
The Jacquard loom wove patterns from punched cards — programmable cloth. Charles Babbage designed his Analytical Engine, and Ada Lovelace wrote what many call the first algorithm, then imagined something bolder: that such a machine might one day compose music and make art.
A poet's daughter saw the whole future in a machine that didn't yet exist.
Alan Turing described a simple universal machine that could, in principle, compute anything computable — the idea underneath every computer since. In 1950 he asked, plainly, “Can machines think?” and offered a game to test it.
The question was never whether it was possible — only when.
McCulloch and Pitts sketched a mathematical neuron; Frank Rosenblatt built the Perceptron, a machine that learned to recognise shapes by adjusting itself. The idea of a network that learns from examples — the seed of today's AI — was already growing.
Today's “neural networks” are this idea, grown up.
At a summer workshop in Dartmouth, a small group coined a phrase for what they were chasing: “artificial intelligence.” They were optimistic — famously too optimistic — but the naming made it a field people could pour their lives into.
A name, a summer, and seventy years of patient work to follow.
ELIZA, a simple program mimicking a therapist, unsettled people with how human it felt — a first hint that conversation is powerful. “Expert systems” captured specialists' rules to give real advice in medicine and industry.
People have felt “this one's alive” since 1966. It passes.
Twice the promises outran the results, funding froze, and headlines declared AI dead. It wasn't — researchers kept working quietly through the cold. These slumps matter to the story: AI has never been a sudden, unstoppable arrival. It has stumbled, and got up.
A thing with winters is a thing that's real, not magic.
A method called backpropagation let neural networks learn far better, reviving the field. In 1997 IBM's Deep Blue beat world chess champion Garry Kasparov — a machine outplaying the best of us at a game we thought was ours alone.
Each “impossible” fell, on a schedule of ordinary effort.
The internet gave machines oceans of examples, and gaming chips (GPUs) gave the muscle to learn from them. In 2012 a network called AlexNet crushed an image-recognition contest, and “deep learning” went from fringe to everywhere.
Not a breakthrough from nowhere — data and patience, finally meeting.
A research paper introduced the “transformer” — a way for a network to weigh which words matter to each other. Dry as it sounds, this is the architecture beneath the AI you can talk to today.
The engine of the moment we're in — published in plain sight.
Language models grew, learning from vast amounts of human writing, until in 2022 one you could simply chat with reached everyone at once. It felt sudden — but you've just walked the two thousand years it stood on.
You arrived not at the start of something strange, but late to something long.
So, then
Every clever thing these tools do stands on the shoulders of counting frames and looms, of Ada's notes and Turing's questions, of decades of patient, ordinary mahi — including the winters when it all seemed to stall. It is a tool, made by tāngata, for tāngata. The same mind that carried you this far is exactly the one that thrives with it.
You are not late to something strange. You are early to something old.
Kaiārahi