The Spark of Curiosity

The Spark of Curiosity

Pip rolled forward three inches, paused, then rolled back two.

It was a small gesture, almost nothing — a tiny wobble in a quiet kitchen at 11:47 PM. But everyone in the lab was watching, and the slight back-and-forth motion felt, somehow, like the most important hesitation in the world.

Pip was a robot the size of a large breadbox, sitting on six rubber wheels. It had a single camera for an eye, a microphone for an ear, and a small computer in its chest that hummed gently as it thought. Around it, the kitchen was the kind of ordinary kitchen you would find in any apartment: a wooden floor with scratches, a refrigerator with a dent in the door, a counter cluttered with cereal boxes and a banana, a stove with one burner ringed in old grease, and a window above the sink showing the dark outline of trees against a streetlight.

For a human child, a kitchen is a place of memory. The smell of toast. The way the refrigerator hums and clicks. The knowledge that the bottom cabinet, the one with the loose hinge, will swing open if you yank it, and that the top shelf of the pantry holds the good crackers, but only if you are tall enough or clever enough. We absorb a kitchen the way a sponge absorbs water — quietly, and without much thought.

Pip had no such memory. It had been wheeled into this kitchen for the first time eleven minutes ago. It had never seen this room. It had never seen any kitchen. And so it did the only thing a thinking machine can do when it does not know what to do: it hesitated.

The lead scientist, a woman named Dr. Adaora Okonkwo, watched the small back-and-forth motion on her monitor. Behind her, three graduate students held their breath.

"It's trying to make a decision," she said softly. "Look at it go."

Pip rolled forward again, this time five inches, then stopped. It tilted its single camera upward, the way a dog tilts its head. A slow whirr came from its chest computer. It rotated in place, very slightly, by about fifteen degrees, as if scanning the room. Then it rolled — slowly, carefully — toward the counter.

What was Pip thinking? We will spend much of this book trying to answer exactly that question. But for now, the important thing is this: Pip, the robot, was facing the very first question of intelligence. The same question that human babies face when they open their eyes to a brand new world. The same question that has haunted philosophers, scientists, and engineers for at least two thousand years. The same question that, in 1950, a quiet British mathematician named Alan Turing wrote down on a single sheet of paper, in a way that would change the world.

*What does it mean for a machine to think?*

The question is older than computer science, older than electricity, older than the Industrial Revolution. It is, in some sense, as old as the mirror. As soon as human beings had the words and the imagination, they began asking whether the things they made could one day look back at them and understand what they saw. The answer to that question is what this book is about, and it is not yet a settled answer. But we have made extraordinary progress in the last seventy years, and the most promising progress, we will argue, has come from an idea that is as old as Aristotle and as new as the latest reinforcement learning paper: the idea that a thinking machine is not just a brain. It is a brain in a body, in a world, and the body and the world are part of the thinking.

From Myth to Machine: Two Thousand Years of Artificial Minds

Long before there were robots, there were stories about them.

In the myths of ancient Greece, the god Hephaestus — the divine blacksmith, the great inventor of Olympus — built golden handmaidens who could speak and move and even think on his behalf. They were, in some sense, the first imagined robots. They were made of metal, but they were not simply metal. They had minds. They had purpose. They could anticipate what Hephaestus wanted and do it before he asked. The poet Homer describes them walking alongside the god, "a stir of golden maids, like living girls, with sense and reason, voice and strength, and skill in the works of the immortal gods." Note the word *sense*. The Greeks did not imagine these machines as calculators. They imagined them as beings that could perceive, that could understand, that could act with judgment.

Another Greek myth tells of Talos, a giant bronze automaton forged to protect the island of Crete. Talos was given to King Minos, and his task was to circle the island's shores three times each day, throwing stones at any ship that came too close. Talos had no heart, no lungs, no brain — only a single vein of ichor, the divine blood of the gods, that ran from his neck to his ankle, sealed with a single bronze nail. His mind, if he had one, was a magical thing, not a mechanical one. And yet he could recognize threats, decide when to attack, and pace his patrols. He was, in the most ancient sense, an intelligent machine.

The Greeks also told the story of Pygmalion, the sculptor who fell in love with an ivory statue he had carved. He prayed to Aphrodite, the goddess of love, and she took pity on him. She brought the statue to life, and the statue, Galatea, opened her eyes and stepped down from her pedestal. Galatea was not a machine — she was a person made of ivory. But the dream behind her was the same dream that drives the makers of robots today: the dream that something we have made ourselves might one day look back at us and understand what it sees.

These are myths, of course. They are not science. But they are the long, slow, mythological background of the question Pip was facing in the kitchen. The dream of an artificial mind is one of the oldest dreams humans have. And the dream has been remarkably consistent across cultures and centuries. In Jewish folklore, the Golem of Prague is a clay figure brought to life by a rabbi to protect the Jewish community, and it is intelligent enough to follow instructions but not wise enough to know when to stop. In Indian mythology, the *Yantra* are mechanical devices of various kinds, including war machines and service robots. In Chinese legend, Yan Shi presents King Mu of Zhou with a mechanical engineer who can sing, dance, and wink at the king's concubines, and the king is so impressed that he mistakes the engineer for a real person.

The consistency of the dream tells us something. It tells us that the question of artificial minds is not a quirky interest of modern computer scientists. It is a deep, recurring theme in human thought, present in almost every culture, in almost every century. We are drawn to this question because we are drawn to ourselves. We want to understand our own minds, and one of the best ways to understand a thing is to try to build it.

The First Real Automata

For most of human history, the dream remained a dream. People could not actually build a thinking machine. But they could build moving machines — automata — that performed a few clever tricks. These were not intelligent in any modern sense, but they hinted at what might one day be possible, and they prepared the cultural ground for the more ambitious projects that came later.

In the thirteenth century, in the city of Isfahan, a Muslim engineer named Al-Jazari wrote a remarkable book called *The Book of Knowledge of Ingenious Mechanical Devices*. In it, he described dozens of machines: water clocks, fountains, hand-washing servants, and even a programmable humanoid robot that could be set to perform different actions at different times of the day. Al-Jazari's most famous robot was a kind of mechanical musician — a four-piece band that floated on a small boat in a lake, playing music at fixed intervals, with the rhythm controlled by a rotating drum with carefully placed pegs.

Al-Jazari was not trying to make his machines think. He was trying to make them perform, and to perform reliably, hour after hour, in a garden. But