AINUA AI Academy
F1.3 · Foundations

Beginner · 20 min core lesson · 4 min quick note

The short history that helps you understand today

Seven useful stops from hand-written rules to generative AI.

AI historyCritical thinkingTechnology literacy
Nia and Kio introducing The short history that helps you understand today
Cold open

AI did not arrive fully formed in a chatbot window. It has had decades of bold promises, useful breakthroughs, expensive winters and computers that would now lose an argument with a kettle.

By the end, you can

place today’s generative AI in a simple timeline, explain why progress came in waves and challenge the idea that newer always means wiser.

Quick note · 4 min

AI history is a series of changing approaches, not one straight march toward a machine mind.

Do now: Sketch seven stops: early ideas, the 1956 naming, expert systems, AI winters, machine learning, deep learning and generative AI.

A breakthrough changes what is possible; it does not remove the need for evidence, safety or people.
Why it matters

History gives you a hype filter. It shows that capabilities grow when ideas, data, computing and investment line up—and that impressive demonstrations do not erase limits.

01

Rules, symbols and a new field

Early AI researchers asked whether parts of reasoning could be represented as rules and symbols. In 1956, the Dartmouth workshop helped establish the name ‘artificial intelligence’ for a field already drawing on mathematics, logic and computing.

Rule-based programs could solve tightly defined problems, but they struggled when the real world became messy. Every exception needed another rule, and common sense proved remarkably difficult to type into a machine.

The first lesson of AI history: a convincing result in a narrow setting is not general intelligence.
02

Booms, winters and learning from data

Expert systems brought specialist rules into business and research, but high costs and unmet promises contributed to periods of reduced funding known as AI winters. The field did not disappear; researchers kept improving methods while headlines moved on.

As digital data and computing expanded, machine learning shifted more work from hand-written rules to patterns learned from examples. Deep learning later improved image, speech and language tasks, especially when large datasets and specialised chips became available.

03

Transformers and the generative moment

The transformer architecture introduced in 2017 made it easier for models to learn relationships across language at scale. Larger foundation models could then be adapted to many tasks and used to generate text, images, audio, video and code.

Today’s systems combine old and new ideas: prediction, search, rules, human feedback and vast computing infrastructure. The interface feels conversational, but the underlying lesson remains familiar—capability depends on the task and responsible use depends on people.

History is not trivia. It explains why capability can leap while reliability still wobbles.