Machine learning from examples — Technology, 14–17 years
See how a system can learn patterns from data, and why its answers depend on the examples it receives.
Learning a pattern from data
In machine learning, people give a computer many examples and a way to measure wrong answers. The computer adjusts a model until it finds patterns that help with new cases. It is not memorising a rule written by a programmer in the usual way; it is fitting a rule to evidence, which can still be incomplete or biased.
Why machine learning was needed
Some tasks are easy to describe in words but hard to turn into fixed instructions, such as recognising a face or predicting whether an email is unwanted. Earlier software needed people to list every useful feature and exception. Machine learning arose to use large collections of examples when the pattern is too complicated or changes too often for hand-written rules.
A tiny prediction
Suppose a model predicts whether a plant needs water using soil moisture: dry examples are 10, 20 and 30, while wet examples are 60, 70 and 80. A new reading of 25 is closer to the dry group, so the model predicts “needs water”. It may still be wrong if the plant, sensor or training examples differ.
The trap: confident is not correct
A machine-learning system can give a neat answer with a high confidence score and still be wrong. People often trust confident speech because, in ordinary conversation, certainty can signal knowledge. But the score describes the model’s internal comparison with its examples, not a guarantee about the real world, especially when the new case is unusual.
Predictions in daily technology
Recommendation systems suggest songs, films or products by finding patterns in what people have viewed and chosen. Phone keyboards predict the next word, and photo apps group similar faces or scenes. These tools can save time, but their suggestions reflect the data and goals chosen by their designers, not a neutral view of what is best.
Keep exploring
Other languages
Loading MyLeoNes™…