“Feed the Perceptron a pile of punch cards and it could learn to distinguish those marked on the left from those marked on the right. Put aside for one moment the mundanity of the task; the machine was able to learn.”

– Ian Sample

The Guardian’s science editor describes the building blocks underpinning present-day AI models that can write, speak, and create. There is, for example, a direct line from the Perceptron, a device designed in the late 1950s, to the “deep” neural networks that power modern AI systems.

The latest building block, known as “transformers,” enables advanced data processing to learn language features, even extending this capability to generate images from text.


Race to AI: the origins of artificial intelligence, from Turing to ChatGPT | THE GUARDIAN | October 28, 2023 | by Ian Sample