B3 · Applications of Artificial Intelligence

AI in Smartphones

A smartphone uses artificial intelligence in many small features: unlocking by recognising a face or fingerprint, predicting the next word you type, improving photos, translating speech and text, and filtering unwanted calls. Each feature is a separate model trained for one narrow job.

8 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

How many times did AI help you today before breakfast, without you noticing?

A smartphone uses artificial intelligence in many small features: unlocking by recognising a face or fingerprint, predicting the next word you type, improving photos, translating speech and text, and filtering unwanted calls. Each feature is a separate model trained for one narrow job.

In simple words

Think of your phone as a small office with many helpers. One helper only knows faces. Another only guesses your next word. Another only fixes dark photos. None of them is clever at everything, but together they make the phone feel smart.

Why it matters

The phone is the AI most people carry all day. Knowing which features use AI helps you understand why they sometimes fail, what data they need from you, and which settings you control.

Key terms

On-device AI
AI that runs on the phone itself, without sending your data elsewhere.
Cloud AI
AI that runs on large computers in a data centre, reached through the internet.
Biometrics
Using features of the body, such as a face or fingerprint, to confirm who someone is.
Computational photography
Using software and AI to build a better photo than the lens alone could capture.
Predictive text
A keyboard feature that suggests the next word from what you have typed so far.
Neural processing chip
A part of a phone's processor designed to run AI models quickly using little battery.

Explained in detail

A small computer built for AI

Modern smartphones contain a special part of the processor designed for the arithmetic of neural networks. It lets the phone run AI models in a fraction of a second while using little battery. This is why a phone can blur the background of a photo as you take it, or turn speech into text with no internet connection. Ten years ago these tasks needed a large computer in a data centre.

How the camera cheats, in a good way

A phone camera has a tiny lens and sensor, far smaller than those of a professional camera. It makes up for this with computing. When you press the button in dim light, the phone takes many frames in quick succession. A model lines them up, keeps the sharp parts of each, removes the grainy noise and combines them into one bright photo. In portrait mode a model works out which pixels belong to the person and blurs the rest, imitating an expensive lens. The picture you see was partly photographed and partly calculated.

The keyboard that learns your words

Predictive text is a small language model. It was trained on a very large amount of writing, so it knows that "good" is often followed by "morning". It also learns from you: the names of your friends, the slang you use, the mix of Hindi and English you type. Much of this personal learning stays on the phone. Some keyboards improve their shared model through a method in which each phone works out a small improvement and sends only that summary, not your messages.

AI that opens the phone to everyone

Some of the most valuable phone AI is found in the accessibility settings. A screen reader speaks aloud what is on the screen, and can now describe the contents of a photo to a blind user. Live captions turn any speech playing on the phone into text for a deaf user. Voice typing lets a person who cannot read or type easily send a message by speaking. A camera app can read a signboard or a medicine label aloud. For many people these features are the difference between using a phone and not using one.

On the device or in the cloud?

On-device AICloud AI
Works offline?YesNo
Where your data goesStays on the phoneTravels to the company's computers
Size of modelSmallCan be very large
Typical usesFace unlock, keyboard, cameraChatbots, complex translation, image generation

How it works

FeatureWhat the AI doesWhat it learned from
Face unlockTurns your face into a list of numbers and compares it with the saved oneMany faces, then your own face during set-up
Predictive keyboardSuggests the next wordHuge amounts of text, plus your own typing
CameraSpots the scene, brightens night photos, blurs the backgroundMillions of example photos
Live translationChanges speech or text from one language to anotherSentences that exist in both languages
Spam call warningFlags numbers that behave like spamReports from many users

On the phone, or far away?

Some features run on the phone itself. This is called on-device AI: it works without the internet, and your data stays with you. Face unlock usually works this way. Other features send your request to large computers in a data centre, which is called cloud AI. Cloud AI can use bigger models, but it needs a connection, and your data leaves the phone.

Try it yourself

Follow one face unlock

You lift your phone. Watch what happens in less than a second.

AI feature or ordinary feature?

Does the feature learn from examples, or does it follow fixed steps?

  • An alarm at 6 a.m.

  • Searching your gallery for 'beach'

  • Night mode in the camera

  • Suggesting the next word as you type

  • The calculator app

  • The torch switch

Common mistakes

  • Thinking the phone has one big AI. It has many small models, each for a single job.
  • Thinking every smart-looking feature is AI. Alarms, timers and calculators follow fixed rules.
  • Trusting autocorrect and translation without reading the result.

Check yourself

Question 1

1. What does face unlock compare?

Question 2

2. What is on-device AI?

Question 3

3. Which of these does NOT need AI?

Summary

  • A phone contains many small AI models, each trained for one job.
  • On-device AI keeps data on the phone; cloud AI sends it to large computers.
  • These features are helpful but not perfect, so check important results.

Finished reading?

Sources: Elements of AI (University of Helsinki) · AI4K12: Five Big Ideas in AI