The Dream of Building a Mind — The Birth of AI and Its Long Winters

Oil came from beneath the earth; semiconductors came from sand. So where did the strategic resource this story follows—intelligence—come from? It was not a resource at all. It first appeared as a dream, inside the human mind. Could a machine be made to think the way a person does? The long struggle over that question begins here.

”Can Machines Think?”

Standing at the doorway of this story is the British mathematician Alan Turing. Known for his deep involvement in cracking the German cipher machine Enigma during the Second World War, Turing turned after the war to a more fundamental question. At the opening of a paper he published in 1950, he posed it plainly: can machines think?

But Turing was shrewd. Any attempt to define philosophically what “thinking” means would lead to an argument with no end. So he reframed the question. If a person, communicating through text alone, can no longer tell whether the other party is a machine or a human—then perhaps the machine may be said to be “thinking.” This is the famous thought experiment that would later carry his name.

By shifting the standard from “what is happening on the inside” to “how something behaves when seen from the outside,” Turing made a quiet but decisive turn that would shape the very nature of artificial intelligence to come. The goal was not to ask what intelligence really is, but to build a machine that behaves intelligently. From this practical stance, humanity’s long challenge began to run.

1956: A Word Is Born at a Summer Workshop

The question existed. But the term “artificial intelligence” did not yet exist in the world. It was born in 1956, at a small summer workshop held at Dartmouth College in the eastern United States.

The organizer was a young mathematician, John McCarthy. In a proposal the year before, he had put forward the phrase artificial intelligence as the name for a new field of study. This was, in all likelihood, the term’s first public appearance. Only about ten people gathered—among them Marvin Minsky, who would go on to lead the field. Claude Shannon, renowned for information theory, was also named on the proposal.

Their ambition was bold in the extreme. Every aspect of learning and intelligence, they held, could in principle be described so precisely that a machine could imitate it. On that hypothesis, they set out over a single summer to explore machines that could use language, form concepts, and solve problems thought to be reserved for humans. Sometimes called the founding convention of AI, the gathering did not complete anything in one summer. But it gathered scattered researchers under a single banner and brought the field itself into being—and in that sense, it was indeed the starting point of artificial intelligence as a discipline.

Promise and Disillusionment: The Recurring “Winter”

The heat of its birth quickly swelled into excessive optimism. Through the 1950s and 1960s, researchers offered one bold prophecy after another. In just a few more years, they said with a straight face, machines would stand shoulder to shoulder with humans. Expectation drew funding, and funding fueled still greater expectation.

Reality, however, was harsh. The computers of the day were far too weak, and the data they could handle was scant. The harder the problem, the more explosively the computation ballooned, until the machine ground to a halt. In 1973, a report published in Britain coldly observed that AI research had fallen utterly short of its grand goals. This became one trigger among several, and research budgets were tightened in country after country. University labs were shut, and promising researchers drifted to other fields to make a living. This stagnation, which began around 1974, would later be called the first AI winter.

Winter did not end after just one season. In the 1980s, expert systems—which translated specialists’ knowledge into mountains of “if this, then do that” rules—drew the spotlight, and a second spring arrived as companies rushed to invest. But the approach of writing rules by hand eventually hit its limits, and the fever deflated once again. Expectation swells, reality disappoints, funding retreats—the history of AI has dutifully repeated this cycle of spring and winter.

Why Did Humanity Try to Place a Mind Inside a Machine?

Even after two winters, why did this dream never fully die?

One reason is humanity’s inexhaustible curiosity about intelligence itself. What exactly is the “power to think” that makes us human? If we could reproduce it on a machine, we would come one step closer to understanding our own nature. What drove the researchers was this question, one that ran deeper than profit or loss. At the same time, if machines could be entrusted with calculation and judgment beyond human reach, then medicine, science, and industry would all leap forward by orders of magnitude. Intelligence was a power that amplified the source of all other powers.

Between intelligence and the oil and semiconductors this series has followed, there is a decisive difference. Oil’s whereabouts were largely fixed by underground reserves, semiconductors’ by factory capacity. But intelligence is bound to neither land nor factory. With only equations, data, and computers, it can in principle be generated anywhere—which is why it would later transform into a strategic resource capable of deciding the balance of power among nations. At the time, however, no one could yet see that far.

Through the winter, many gave up on the field. But a small handful of researchers, indifferent to the world’s verdict, kept patiently refining one “neglected technique.” It mimicked the neural circuitry of the human brain, and was called the neural network. In those days it was seen as an outdated idea, far from practical use, and was pushed to the margins of research.

And yet—it was not only the technique that lay sleeping. The two conditions that would wake it, namely vast quantities of data and computing power of a wholly different order, were quietly maturing in corners of the world. When, in the 2010s, those three finally met, the long winter ended all at once. The spark was an upset at a competition over a single image. The story moves to 2012.