A short history of AI

Three ways to see it

1. 1

   1950, Alan Turing publishes a paper proposing what he calls the imitation game, now known as the Turing test. He does not call the field artificial intelligence; that name comes six years later at the Dartmouth workshop of 1956, where John McCarthy, Marvin Minsky, Claude Shannon and others give the field its name and its early ambitions. The promise was that within a generation machines would do most of what humans called thinking. The promise was wrong on timing. It was not wrong on direction.

2. 2

   Through the sixties and seventies symbolic AI dominates: systems that reason with hand-coded logic. In parallel a quieter line of work emerges in the perceptron (Frank Rosenblatt, 1958), the seed of every neural network you read about today. The perceptron is followed by hype, then a famous critique (Minsky and Papert, 1969), then the first AI winter, where funding dries up and the field becomes a punchline. AI winters return more than once. They are part of the story.

3. 3

   1980s: expert systems. Companies pay handsomely for software that encodes a doctor's, lawyer's, or geologist's rules. Pakistan's earliest forays into machine learning at university labs date roughly to this period. The expert system bubble bursts in the early nineties; another winter. But somewhere in this time backpropagation is rediscovered, GPUs start to appear in graphics cards, and the internet begins producing the data the next era will need.

The why-tree

Try this with Claude

AI-edge prompt: 'Give me five turning points in AI history between 1950 and 2026, each in two sentences, and for each one explain what made it matter to a working professional rather than only to researchers.' Compare the model's answer to the timeline you drew. Where do they disagree?