What AI actually is

For decades, most software followed rules a programmer wrote down: if this, then that. Artificial intelligence is the broad goal of making computers do things that normally need human intelligence, like understanding language or recognising images.

Machine learning flipped the approach. Instead of writing rules, you show a system many examples and it learns the patterns itself. A spam filter trained on millions of emails learns what spam looks like better than any list of banned words.

Deep learning uses neural networks: layers of simple mathematical units whose connections are tuned during training. With enough data and computing power, these networks learn remarkably rich representations of language, images and sound.

The transformer, a neural network design introduced in 2017, made it practical to train on enormous amounts of text. Every major chatbot today, from Claude to ChatGPT to Gemini, is built on it.

Earlier AI mostly sorted or predicted things: is this email spam, what will sales be next month. Generative AI produces new content: an essay, a photo-realistic image, a video clip, working code.

The leap since 2022 is that these models became general. One model can write a poem, debug a program, summarise a contract and plan a trip, because it learned broad patterns of human knowledge and reasoning from its training data.