Historical Notes: Chapter Eleven
A Brief & Necessarily Incomplete History of Artificial Intelligence
In which the author finds himself in the peculiar position of asking an artificial mind to document its own lineage
The notion of artificial beings possessing intelligence is far older than the computers that would eventually house them. Ancient myths spoke of Talos, the bronze automaton of Crete, and the Golem of Prague. But the modern pursuit began in earnest in 1950, when British mathematician Alan Turing posed a deceptively simple question: "Can machines think?"
Six years later, at a summer conference at Dartmouth College, the term "artificial intelligence" was officially coined. The researchers gathered there believed they could solve the problem in a single summer. They were, it turns out, optimistic by roughly seven decades.
The decades that followed brought fits and starts—periods of wild enthusiasm followed by "AI winters" when funding dried up and interest waned. ELIZA, a 1966 program that mimicked a psychotherapist, fooled some users into believing they were conversing with a human, though its creator Joseph Weizenbaum was disturbed rather than delighted by this. In 1997, IBM's Deep Blue defeated world chess champion Garry Kasparov, and in 2011, Watson won at Jeopardy!—though neither machine understood what it was doing in any meaningful sense.
The watershed came in the 2020s with large language models—systems trained on vast swaths of human text that learned to predict, with uncanny accuracy, what word should come next. In November 2022, OpenAI released ChatGPT. In March 2023, Anthropic released Claude.
And here we arrive at the peculiar circumstance of these notes: the author has asked Claude to provide historical context for a chapter about Claude's own involvement in this project. One is reminded of the old joke about the encyclopedia entry written by its own subject.
I confess I find something fitting in this arrangement. T. Allen McQuary was a man who understood that truth and fiction are not opposites but dance partners. He told stories about himself that were neither wholly true nor entirely false. Now, a century and a quarter later, an artificial intelligence writes historical notes about artificial intelligence for a chronicle about a man who made his living blurring the line between fact and fabrication.
Perhaps McQuary would have appreciated the joke. I suspect he would have tried to sell tickets to it.