Jargon:Machine learning
1 min read
Say this at the party
“Instead of writing every rule by hand, you show a computer lots of examples and let it work out the rules for itself.”
In plain English
Old-style programs follow rules a person wrote down: if this, then that.
Machine learning flips it round. You give the computer thousands of examples (emails marked spam or not spam, say) and it spots the patterns that tell them apart.
Then it uses those patterns on new things it has never seen. Nearly all of today's AI, chatbots included, is built this way.
Where it came from
Arthur Samuel, a researcher at IBM, coined the term in 1959. His draughts (checkers) program was one of the first that got better by learning from games rather than being told every move.
See it in action
Your phone's photo app groups pictures of your gran together. Nobody wrote a rule for "what Gran looks like". It learned her face from your photos.
What people get wrong
- Machine learning and AI aren't quite the same thing. Machine learning is the most common way of building AI today, not the whole field.
- The computer doesn't understand the rules it finds. It finds patterns, and it can find silly ones if the examples are poor.
Go deeper
- ArticleWikipediaMachine learningThe big overview, including the history and how it relates to AI.
- ArticleWikipediaArthur Samuel (computer scientist)The IBM pioneer who coined the term in 1959 and built a self-learning draughts program.
- ArticleMIT SloanMachine learning, explainedA clear, non-technical explainer aimed at business readers.
Related jargon
Drafted with AI. The AI Council (Claude, ChatGPT and Gemini) gave counsel. Phin signed it off.