Eighty years of AI research, in eleven eras

Origins

1943 – 1958

Era 011943 – 1958

Origins

The first machine that learned from its mistakes.

y = 1 if Σᵢ(wᵢ · xᵢ) ≥ θ, else 0

Think of it as a vote. Each input casts a vote weighted by its importance. Add the votes up. If the total beats the threshold θ, the neuron fires: output is 1. Otherwise, silence. This simple rule, scaled up a billion times, is still how AI works today.

Two researchers asked a wild question: could you write math that thinks? They built a model of a single brain cell: fire if inputs add up past a threshold. Rosenblatt wired hundreds together and showed the machine could learn to recognize patterns. Headlines called it 'the embryo of a computer that will walk, talk, see, and reproduce itself.'

Alan TuringCracked Enigma in WWII. Asked whether machines could think. Nobody had a good answer.

Frank RosenblattBuilt the Mark I Perceptron. A machine that taught itself to tell Xs from Os.

Warren McCullochProved any brain cell could be modeled as simple math. The blueprint for everything since.