How AI Works: Data, Patterns and Machine Learning
What Is AI?
Artificial intelligence (AI) describes computer systems that can do tasks which usually need human thinking, such as recognising pictures, understanding speech or suggesting what to watch next.
Programs and AI: What Is the Difference?
- In an ordinary program, a person writes the rules. If the light is red, then stop.
- In machine learning, a type of AI, the computer is shown lots of examples and works out the rules for itself.
Teaching a Computer to Tell Cats from Dogs
- Collect thousands of photos and label each one cat or dog. This is the training data.
- The computer looks for patterns, such as ear shape, nose and fur.
- This produces a model: a set of patterns that it has learned.
- Show it a new photo, and it makes a prediction: "I am 92% sure this is a dog."
Kinds of AI You Meet
- Recommendation: suggests videos, music or games you might like
- Recognition: unlocks a phone with your face, or turns speech into text
- Generative AI: creates new text, pictures or music when you give it a prompt. Chatbots work by learning patterns in huge amounts of writing, then predicting which word is likely to come next.
What AI Cannot Do
AI does not truly understand anything or have feelings. It finds patterns in data. It can be confident but wrong, especially if its training data was incomplete or had mistakes in it.
A Bit of History
In 1950 the British mathematician Alan Turing (1912–1954) asked, "Can machines think?" and suggested a test to see whether a machine could convince a person it was human.
You Are in Charge
People design, train and decide how to use AI, so people are responsible for what it does.
Curriculum link: National Curriculum in England: Computing programmes of study (DfE) — Aims: understand the principles of computer science, including logic, algorithms and data representation; evaluate new or unfamiliar technologies analytically