Before there were computers, how did people imagine thinking machines?

The first program was written for a machine that was never built. In 1843 Ada Lovelace translated an article about Charles Babbage's Analytical Engine, a mechanical computer made of gears, and added notes much longer than the article itself. One note lays out, step by step, how the engine could calculate a series of numbers called Bernoulli numbers, which is why it's often called the first published program. She also guessed the engine could work on more than numbers, perhaps even music. But she was firm that it could originate nothing: it would only do what someone knew how to order it to do.

Almost a century later, a young Claude Shannon noticed in his 1937 master's thesis that the on/off switches in telephone relays behave exactly like true/false logic. Two switches in a row make AND; two side by side make OR. Any logical rule can be wired up out of switches, and every chip in your phone still works this way. In 1948 he went further and showed how to measure information, counting it in bits, the number of yes/no answers you need to pin something down.

In 1950 Alan Turing asked "Can machines think?" and then swapped the question for one you could actually test: the imitation game. A judge trades typed messages with two hidden players, a person and a machine, and tries to tell which is which. Turing took on Lovelace directly. He called her doubt "Lady Lovelace's objection," pointed out that machines surprised him all the time, and suggested a machine might be built to learn rather than be told everything.

Shannon's 1951 guessing game is the quiet hero here. Once people had a little context, they guessed most letters on the first or second try, and he estimated English carries only about one bit per letter, far less than the nearly five bits you would need if all 27 symbols were equally likely. Text is full of patterns. A modern language model plays the same game at enormous scale: it is trained to guess the next token, and each guess is a probability, just like a person's hunch.

By 1955 these threads were ready to be tied together. A proposal by John McCarthy, Marvin Minsky, Nathaniel Rochester and Shannon asked for a summer workshop at Dartmouth College to study how machines could use language, form ideas, improve themselves and be creative. That proposal coined the phrase "artificial intelligence", and the workshop met in the summer of 1956. The field had a name.