Genius Makers Summary

In the 1980s, many serious computer scientists treated neural networks like a bad old smell in the lab. Cade Metz, a technology reporter for The New York Times, tells how that rejected idea became the center of modern artificial intelligence. His angle is not just technical. He follows the stubborn people, strange rivalries, and sudden money that turned a fringe belief into a global industry.

The book's clearest lesson is simple: the AI boom did not arrive because machines suddenly became clever in one clean leap. It arrived because old ideas met huge data sets, cheap powerful chips, and companies rich enough to burn money until the method worked.

That changes the story. Genius Makers is less a tale of lone magic than a tour of the workshop where a dismissed tool became an industrial furnace. The sparks are exciting, but Metz keeps asking who owns the heat.

In 1958, Frank Rosenblatt showed the press a machine called the Perceptron, and The New York Times suggested it might someday walk, talk, and think. That promise was too large for the hardware of the time. By 1969, Marvin Minsky and Seymour Papert had published a sharp attack on what perceptrons could not do, and the field cooled fast.

Metz uses this long winter to show why the later boom mattered. Neural networks were built on a strange bet. Instead of telling a computer each rule, you give it many examples and let it adjust its own inner settings. The idea was rough, hungry, and hard to explain to funding boards, which is not a great dating profile for a research program.

A field can die in public and keep breathing in small rooms.

Geoffrey Hinton becomes the keeper of the flame. In the 1980s, while many AI researchers chased systems based on hand-written rules, Hinton kept pushing networks that learned from data. He was not alone, but Metz presents him as the clearest symbol of the type: brilliant, difficult, and willing to look foolish for longer than most careers can survive.

This matters beyond AI because many breakthroughs pass through an embarrassing stage. They look wrong before they look early. The danger is that institutions often cannot tell the difference. Neural networks survived because a few people kept feeding the little basement fire when the official building had turned off the lights.

The future sometimes begins as a bad grant application.