History of AI as a Business

Artificial intelligence has been "about to arrive" for seventy years, and the business built around it has boomed and crashed several times before the current cycle. A summer workshop at Dartmouth in 1956 named the field and promised that a few months' work would make real progress on machine intelligence. Government money, mostly from the US defence research agency, carried the field through the 1960s, until reports in Britain and America found little to show and funding was cut. Expert systems turned AI into an industry in the 1980s, with specialised Lisp machines, venture-backed start-ups and Japan's national Fifth Generation project, and that industry collapsed in 1987–93 when cheaper computers and brittle software caught up with it. Statistical methods quietly took over speech, search and advertising in the 1990s and 2000s, without much talk of "AI". Then three things arrived together around 2012: large labelled datasets, graphics chips that could train neural networks fast, and a handful of researchers who had kept faith with those networks. The deep-learning results of 2012 started a bidding war for talent, the transformer of 2017 gave the field a general architecture, and ChatGPT in November 2022 turned research into the fastest-adopted consumer product to that date and set off the largest capital cycle in technology history. This section traces that sequence, explains the business models each era produced and the patterns that keep repeating, and shows how the history helps separate a real shift from a re-run of an earlier hype.