The Genesis and Early Failures of Artificial Intelligence

Key Insight

The field of artificial intelligence (AI) officially began in the summer of 1956 at Dartmouth College, where John McCarthy, who coined the term, convened researchers. This gathering, the Dartmouth Summer Research Project on Artificial Intelligence, aimed to enable machines to use language, form abstractions, solve problems typically reserved for humans, and improve themselves. While marking the birth of AI research, this event also set a precedent for developing ambitious, often unrealistic goals that consistently fell short of expectations, defining the field's early trajectory.

Throughout its initial decades, AI was characterized by optimistic predictions that proved inaccurate. Attendees at the 1956 Dartmouth conference, for example, believed a machine would surpass the world's chess champion within a mere decade. In reality, it took over 35 years for a machine to defeat a champion-caliber player at checkers, and even longer to achieve the same feat in chess. Alan Turing, a pioneer who predated McCarthy, proposed the 'imitation game'—later known as the Turing Test—in 1950, predicting a computer would pass it by the year 2000; this forecast was ultimately off by more than two decades.

This pattern of overpromising and under-delivering became a defining characteristic of artificial intelligence for its first sixty to seventy years, with the technology often seeming 'forever a decade away.' The unfettered optimism of early believers was a fatal flaw, as practical advancements struggled to keep pace with theoretical aspirations and public expectations. Despite significant intellectual contributions, such as Turing's foundational work on 'intelligent behaviour,' the consistent gap between bold claims and tangible progress led to recurring cycles of anticipation and subsequent disappointment within the nascent field.