AI has moved in waves of optimism and disappointment. Knowing the history helps put today's excitement in context.
The Beginnings (1950s–1960s)
Alan Turing's 1950 paper asked whether machines can think and proposed what became the Turing test. The term "artificial intelligence" was coined for the 1956 Dartmouth workshop. Early programs solved logic puzzles and played checkers, and researchers expected rapid progress.
Winters and Expert Systems (1970s–1990s)
Progress proved slower than promised, and funding dried up in periods known as AI winters. The 1980s saw a boom in expert systems — large collections of hand-written rules — which worked in narrow domains but were brittle and costly to maintain.
Statistical Machine Learning (1990s–2000s)
Researchers increasingly let systems learn from data. Methods like support vector machines and decision trees, together with growing datasets and computing power, made machine learning practical for spam filtering, search and recommendations. IBM's Deep Blue beat the world chess champion in 1997.
The Deep Learning Era (2010s)
In 2012 a deep neural network won the ImageNet image-recognition challenge by a wide margin, kicking off the deep learning boom. GPUs, big datasets and better training techniques drove breakthroughs in vision, speech and translation. The transformer architecture, published in 2017, became the foundation for modern language models.
Generative AI (2020s)
Large language models trained on vast amounts of text made AI assistants mainstream. The field is now as much about applying, evaluating and governing AI as inventing it.