AI All Around You

Long before you typed a question into a chatbot, AI was already quietly working in your phone, your bank and your groceries. Here is where it hides, what it does, and what kind of AI each one is.

Your phone unlocks when it sees your face. A map app tells you the drive to the office will take 38 minutes, not the usual 25, because of an accident on the highway. Your email quietly moves a “You have won RM50,000!” message into the spam folder. Your bank sends an alert about a card payment you did not make. By the time you have finished breakfast, you may have used half a dozen AI systems — and not one of them called itself “AI”.

The invisible helpers

When people say “AI” today they often mean chat assistants such as ChatGPT, Gemini or Claude. But those are the newest and most visible members of a much larger family. Most of the AI in your life is narrow AI: a system trained to do one job — rank videos, read a number plate, judge whether a transaction looks unusual — and nothing else.

Almost all of it is machine learning. Instead of programmers writing every rule by hand, the system learns patterns from huge numbers of examples, as you saw in How Machines Learn. A spam filter is not told “block any email containing the word ‘lottery’”; it is shown millions of emails that people marked as spam or not spam, and it learns the patterns itself.

Key point

A rough test for “is this AI?”: does the system make a judgement or prediction it learnt from data, rather than follow a fixed rule someone wrote? A calculator is not AI. A camera that decides which face to focus on is.

Tap to reveal: twelve everyday AIs

Most of these fall into a handful of families: recommender systems that suggest what you might like, computer vision that interprets images, speech systems, and neural networks that predict from data. Tap each tile to see what the AI is doing behind the scenes. The short code tells you the main kind of AI involved.

REC = recommender · ML = machine learning that predicts from data · CV = computer vision · SPEECH = speech recognition and synthesis · LLM = large language model.

What kind of AI is each one?

On AI, ML, Deep Learning, GenAI you met the nesting dolls of the field. The everyday systems above fall into a few broad kinds. Switch between the tabs to see each.

Machine learning that predicts or ranks. Spam filters, fraud detection, arrival times, e-hailing prices and recommendations all take in data and output a number or a choice: “92% likely spam”, “38 minutes”, “you’ll probably like this”. Many use fairly classic methods; some use deep learning. They do not create anything new — they judge.

Consensus All of these are narrow systems. The face-unlock network cannot recommend a song, and the recommender cannot read an X-ray. Even the most capable chat assistants, which feel general, have important gaps — see What AI Gets Wrong.

Why it matters that you notice

You might ask: if it works, why care whether it is AI? Three reasons.

1It shapes your choicesFeeds decide what you see

Recommendation systems are usually tuned to keep you watching, scrolling or buying. That is not evil in itself, but knowing it helps you notice when an app is choosing for you — and to go looking for things outside your feed.

2It makes mistakesPredictions, not certainties

A fraud system can block your genuine payment overseas; face unlock can struggle in poor light; a translation can garble an idiom. Every one of these outputs is a prediction with a chance of being wrong. When it matters, a human should be able to check and override it.

3It runs on your dataYour clicks are the training set

Much everyday AI learns from how people behave: your location, purchases, searches and voice. Knowing this helps you decide which settings to switch off and which apps deserve access. Staying Safe with AI covers practical steps.

There is a fairness question too. Bias in the data can mean a system works better for some groups than others — face-analysis systems, for example, have been found to make more errors on darker-skinned women than on lighter-skinned men. The next page in this track explores why.

Spot the AI

Test your eye. For each scenario, decide whether AI is likely to be involved.

Spot the AI

A kettle switches off automatically when the water boils.

A thermostat or bimetallic strip follows a fixed physical rule. Nothing is learnt from data — it is automation, not AI.

Your music app builds a weekly playlist of songs you have never heard but mostly like.

Personalised playlists for millions of users are made by recommender systems learning from listening patterns.

Your bank texts you to confirm an unusual late-night online purchase.

Fraud detection models score each transaction against your normal pattern. (Always confirm through your bank’s official app or number, never through a link in a message.)

A parking-barrier camera reads your number plate and opens the gate.

Automatic number-plate recognition uses computer vision to find the plate and read the characters.

A spreadsheet adds up a column of figures.

Arithmetic by formula is ordinary computing. No learning, no prediction.

Your phone’s night mode turns a dark, grainy scene into a bright, clear photo.

Computational photography uses AI to align, combine and denoise several frames.

// Pause and reflect

  1. List the apps you opened yesterday. Which of them probably used AI, and for what?
  2. Which one of those would you most want a human to be able to check or override when it gets things wrong?
  3. Is there one app whose recommendations you would like to rely on less? What could you do about it?

Your notes are saved privately in this browser.

Try this

Open the settings of one app you use daily — video, music, shopping or maps — and look for words like “personalisation”, “recommendations”, “activity” or “history”. Note what data it uses, and decide whether to keep, pause or clear it.

Key terms on this page

Sources & further reading