The perceptron is the earliest practical model of a learning artificial neuron. It takes several numeric inputs, multiplies each by an adjustable weight, sums them, and fires an output if the total crosses a threshold. What made it special was that it could adjust its own weights from labeled examples, so it learned to separate one class of pattern from another rather than being hand-programmed.
It was introduced by Frank Rosenblatt in 1958 in the paper “The perceptron: A probabilistic model for information storage and organization in the brain,” published in Psychological Review, volume 65. Rosenblatt framed it as a model of how the brain might store information and recognize patterns, and it became the conceptual seed for nearly all later neural networks.
The perceptron also taught the field a hard lesson: a single perceptron can only separate patterns that are linearly separable, a limitation later highlighted in 1969 by Minsky and Papert. Overcoming that limitation, by stacking many neurons into layers, is what eventually produced modern deep learning.
Why business readers should care: the perceptron is the direct ancestor of every neural network behind today’s image recognition, language models, and recommendation engines. Understanding that these systems are layers of weighted sums learned from data demystifies what “AI” actually does and where its limits come from.