AI winter
The periods when the money and the patience ran out — and researchers started calling their field something else.
The first arrived in the 1970s. The Lighthill report in 1973 dismissed British AI research as unproductive, DARPA withdrew funding after speech understanding projects failed, and the early promises of machine translation had gone unmet. The second came around 1987, when the market for specialised LISP machines collapsed within months once ordinary workstations became faster and cheaper.
The pattern is the same both times: inflated expectations, results that work in a demonstration but not in the world, and then an abrupt reversal in funding. The effect on language is worth noting — researchers stopped writing "artificial intelligence" in grant applications and wrote "machine learning", "informatics" or "decision support" instead. The term is now used mainly as a caution in discussions of current AI optimism.