Thinking machines among humans

How a brisk history of AI helps us see the future more clearly

- SRINATH SRIDHARAN

WHEN ADA LOVELACE mused that the first mechanical computer “could compose elaborate pieces of music” if instructed properly, she planted the seed of a question that has since obsessed generations: what happens when the tool starts to think? In The Shortest History of AI, Toby Walsh compresses nearly two centuries of that pursuit into a brisk, lucid 203 pages—a kind of intellectual time-lapse of humanity’s most ambitious experiment.

The book opens not in Silicon Valley, but in Dartmouth college in New Hampshire, US, on June 18, 1956. The day when the legendary AI guru John McCarthy convened a group of like-minded academic colleagues for an eight-week-long workshop to build intelligent machines.

Well, it is him who coined the term “artificial intelligence”.

Walsh argues, the story of artificial intelligence is not about sudden leaps but about patient accumulation—“many so-called overnight successes”, he writes, “were decades in the making”.

That line captures both the tone and the thesis of the book. Al, in Walsh’s telling, is no meteor that recently struck human civilisation; it’s a slow-burning fire that we have been stoking for generations.

The book’s structure—six essential ideas that ‘animate’ AI—gives this short history its backbone. Those ideas are: symbol-manipulation, search and optimisation, rule-based reasoning, learning from experience, reinforcement and correction, and probabilistic inference. Together they read like the DNA strands of machine intelligence.

Walsh, a veteran AI researcher, has the gift of making complexity conversational. He can sketch in a few lines how the “symbol-processing” dream of the 1950s birthed both optimism and hubris, or how neural networks, once discarded as dead ends, rose again to power the age of deep learning.