AI Trend Research

← AI glossary

AI glossary · Beginner · Language models

Machine learning

Machine learning is a way of having a computer find rules (patterns) on its own by looking at lots of data and use them to make predictions or decisions, instead of people defining every rule by hand.

About this term
Level
Beginner · Core ideas for first-time readers
Category
Basics
Added
2026-10-08
In plain words It is like showing a child many photos of cats so they learn to recognize cats on their own, instead of explaining 'cats have pointed ears and…' The difference is that a computer does not understand the way people do; it learns statistical patterns in the data.

How it works

First, a large amount of example data is collected. The computer looks at this data, makes predictions, calculates how far off it was from the correct answers (the loss), and then adjusts its internal numbers (parameters) little by little. Repeating this process many times is called "learning" (or "training"). The result of training is called a "model," and it can produce answers from the patterns it learned even when it sees data it has never seen before.

Training and using machine learningExample data → Make a prediction; Make a prediction → Calculate error; Calculate error → Adjust parameters; Adjust parameters → Make a prediction(Repeat); New data → Trained model; Trained model → Prediction/decision① Training (repeated)② UseExample dataMake a predictionCalculate errorLossAdjust parametersTrained modelNew dataPrediction/decisionRepeat
Training and using machine learning A model is built by repeatedly making predictions on data, calculating the error and making corrections; the finished model is then used to answer on new data.

Main types

Supervised learning, which learns from data labeled with correct answers, is used for "getting the right answer" tasks such as classifying spam email. Unsupervised learning, which groups similar data together without correct answers, is used for tasks such as dividing customers into groups. Reinforcement learning, which learns through trial and error and rewards, is used in games, robot control, and also for refining LLMs. Deep learning is one branch within machine learning.

Things to know

Machine learning depends heavily on data quality. If the data is skewed in one direction, the model will also make skewed judgments. Another common problem is "overfitting," where the model memorizes the training data and performs worse on new data. That is why it is essential to check performance separately on data that was not used for training.

Example

Spam filters in email services, recommendation features in online stores and video services, and fraud detection at credit card companies are typical uses of machine learning. For problems where people cannot write down every rule, these systems make decisions automatically using patterns learned from past data.

Related terms

Deep learning Supervised learning Unsupervised learning Reinforcement learning Overfitting Loss function