Machine Learning
Machine learning is a way of building AI by letting computers learn patterns directly from data, rather than having programmers write out every rule by hand. Almost every modern AI system you interact with today — from TikTok video recommendations to ChatGPT and video generators — is built on machine learning.
Example: Show it 1 million pictures of cats and not-cats, and the system works out by itself what a "cat" is.
Traditional programming vs machine learning
In traditional software programming, a human developer writes explicit rules for every scenario: "If the order total exceeds a set amount, apply free shipping." That works well for clear-cut rules, but it falls apart when you want a computer to recognise a person's voice, spot spam emails, or tell a photo of a cat from a dog — tasks with millions of tiny variations.
Machine learning flips the process around. Instead of writing rules by hand, engineers feed the computer a large dataset of examples and let the algorithm discover the underlying patterns itself. Once training finishes, that trained model can make predictions on brand-new inputs it has never seen before.
How AI, Machine Learning, and Deep Learning fit together
People often use these three terms interchangeably, but they actually sit inside one another like three nested circles:
- Artificial Intelligence (AI): The broadest umbrella — any technique that enables computers to mimic human-like intelligence or decision-making.
- Machine Learning: The main subset inside AI — systems that improve at a task by learning patterns from data rather than following hand-coded rules.
- Deep Learning: A specialised subset inside machine learning that uses multi-layered neural networks to handle complex, unstructured data — powering today's LLMs, generative images, and generative video.
Where you use machine learning every day in Malaysia
Even when you are not opening a chatbot, machine learning models quietly run behind the scenes in the apps Malaysians use all day:
- Social feeds and online shopping: Recommendation engines that learn which TikTok clips you watch to the end or which products you browse.
- Banking fraud protection: Systems that learn your normal spending habits and flag an alert when a card transaction looks out of character.
- Traffic and delivery estimates: Navigation and food delivery apps that predict arrival times from historical road speed patterns during rush hour.
- Speech and optical recognition: Voice-to-text transcription and licence-plate cameras at shopping mall car parks.
Why data quality makes or breaks machine learning
There is a golden rule in machine learning: garbage in, garbage out. A model is only as reliable as the dataset used to train it.
If a model's training data is unbalanced — for instance, if it contains plenty of English text and Western photos but very little Malaysian Malay or Southeast Asian imagery — the model will show dataset bias. Understanding how machine learning works helps you see AI not as magic, but as a pattern engine shaped by its training data.
Common questions
What is the main difference between AI and machine learning?
AI is the broad goal of creating smart computer systems, while machine learning is the specific method of achieving that goal by training software to learn patterns from data.
Do I need to study machine learning maths to use ChatGPT or AI video tools?
Not at all. Just as you do not need to build a car engine to drive safely, understanding the basic idea of machine learning simply helps you know what AI can and cannot do when you write prompts.
Does an AI model retrain itself instantly every time I send a chat message?
No. Most commercial models go through a heavy training phase first and are then 'frozen' when serving users (called inference), meaning your conversation lives inside the chat's context window rather than rewriting the model's core weights on the fly.
Got the basics down? Start using AI practically at /mula/ or explore the rest of the glossary at /kamus/.