Artificial intelligence, explained without the fog

Ten short pages on what these systems are, how they learn from examples, where they already work, and where they fail people. No maths required to start.

Begin with the basics Look up a term

Attention weights between words in a sentence
Reading	The	machine	learned	to	read	this	sentence
The
machine
learned
to
read
this
sentence

This is roughly what reading looks like inside a language model. Each row is a word being processed; the bright cells are the earlier words it leans on. The empty upper half is the rule that it may never look ahead.

ignored leaned on

Three ideas hold the whole thing up

Almost everything written about AI is an elaboration of these. Get them straight and the rest follows.

It learns from examples

Nobody wrote the rule that separates a cat from a dog. The system was shown many labelled pictures and adjusted itself until it usually got them right. That adjustment is the whole trick, repeated at enormous scale.

It predicts, it does not know

The output is the most likely answer given the patterns in the training data. That is why a model can write a flawless paragraph containing a fact that does not exist.

It inherits what it was shown

The data carries the assumptions, gaps and unfairness of whoever collected it. A model trained on the past will reproduce the past, including the parts we meant to leave there.

Read it in order, or jump

Each page stands alone, but they were written to be read top to bottom.

What AI is

A working definition, the difference between narrow and general AI, and the words people use loosely.

History

From a summer workshop in 1956 to models that write code — including the two winters in between.

How it learns

Training data, features, loss, and the loop that turns a pile of examples into a prediction.

Neural networks

Layers, weights, backpropagation, and why transformers changed what language models can do.

Where it is used

Health, farming, money, language and transport — with the honest limits of each.

Risks and ethics

Where these systems fail people, why it happens, and what responsible use looks like.

Learn it

A six-stage path from spreadsheets to shipped models, plus the roles that hire at each stage.

Glossary

Forty terms you will meet in any AI article, defined without jargon.

Contact

Send a correction, a question, or a topic you want covered.

What this site is not

Being clear about the edges saves you time.

It is not a coding course — there is no Python here, though the learning path tells you where to find it. It is not a news feed; specific model names age badly, so the pages stay with ideas that hold. And it does not argue that AI will save us or ruin us. Both cases are made loudly elsewhere, usually by people selling something.

What it does try to do is leave you able to read an article about AI and tell the substance from the froth.