Class 6 Lesson 4 — Types of AI Tools

Story · Meet Raju's Family

One Evening, Six AI Tools 🏠

It is 6 PM on a Tuesday at Raju's house in Hyderabad. Everyone is home.

Raju's Amma asks her phone "Hey Google, what is today's weather?" — a voice assistant answers instantly.

His Nanna is on YouTube, which has lined up a cricket highlights video he never searched for — his watch history guided it there.

His akka (older sister, in college) is using ChatGPT to help draft an essay introduction. She will rewrite it in her own words.

His tatha (grandfather) is on a video call with a doctor in Mumbai. The hospital's AI system scanned his recent chest X-ray and flagged two areas for the doctor to review.

Raju himself is playing a game on his tablet where the AI opponents adapt to how he plays and get harder when he gets better.

And their neighbour downstairs just sent Raju's mother a WhatsApp message with an AI-generated image of their festival invitation card.

👉 Six people. Six AI tools. All in the same hour. All doing completely different things. In this lesson, you will understand the six families of AI tools — and what makes each one different.

Section 1 of 8

🗺️ The Six Families of AI Tools

AI tools are not all one thing. They are built for different purposes, trained on different data, and solve different problems. It helps to group them into families.

💬

Language / Chatbot AI

ChatGPT · Gemini · Copilot

Reads and writes text. Can answer questions, summarise, translate, and help with writing tasks.

🎨

Image AI

DALL-E · Midjourney · Adobe Firefly

Creates images from text descriptions. Also used to enhance, edit, or analyse photos.

🎙️

Voice AI

Google Assistant · Alexa · Siri

Understands spoken language and responds in speech. Also converts text to audio (TTS).

📺

Recommendation AI

YouTube · Netflix · Amazon · Swiggy

Predicts what you want to see, buy, or eat next. Uses your history and behaviour patterns.

👁️

Computer Vision AI

Face unlock · Plantix · Medical imaging

Understands and interprets images and video. Detects objects, faces, diseases, and actions.

📈

Prediction / Forecasting AI

Weather apps · IMD · Credit scoring

Uses patterns in historical data to predict future outcomes — weather, stock prices, health risks.

These families are not perfectly separate — many real products combine several types. For example, a self-driving car uses Computer Vision + Prediction AI + Recommendation AI all at once. But understanding the families helps you recognise what any AI is doing.

Quick check from the story: Amma's voice assistant → Voice AI. Nanna's YouTube feed → Recommendation AI. Akka's ChatGPT essay helper → Language AI. Tatha's X-ray scanner → Computer Vision AI. Raju's game opponents → Prediction AI. The festival invitation image → Image AI. ✓ All six families in one family home.

Section 2 of 8

💬 Language AI — Chatbots That Read and Write

Language AI (also called Large Language Models or LLMs) is the type of AI behind ChatGPT, Google Gemini, and Microsoft Copilot. These tools can write, summarise, explain, translate, and answer questions on almost any topic.

How do they work (simply)? They were trained on a huge amount of text — books, websites, articles, code — and learned the patterns of language. When you ask a question, they predict the most useful sequence of words to respond with.

Important: Language AI does not "know" things the way you know them. It predicts what a useful response looks like based on patterns in its training data. This means it can produce very confident-sounding answers that are factually wrong. This is called a hallucination.

What Language AI is genuinely good at:

First drafts of writing (emails, essays, descriptions)
Explaining difficult concepts in simpler words
Translating text between languages
Summarising long documents
Generating ideas and brainstorming
Helping debug simple code

Good uses for students

Ask it to explain a difficult concept
Generate ideas for a project topic
Get feedback on your own writing draft
Practise Q&A for an exam topic

Uses that can get you in trouble

Copying its text directly as your work
Using it for facts without verifying
Treating it as a reliable encyclopedia
Relying on it to write your original thoughts

India context: Google Gemini supports many Indian languages including Telugu, Hindi, Tamil, and Kannada. You can ask it questions in Telugu and get responses in Telugu. This is a result of India-language training data work by teams like Google's India AI research and initiatives like AI4Bharat.

Section 3 of 8

🎨 Image AI — Turning Words into Pictures

Image AI can generate photorealistic images, illustrations, and artwork from a text description. You type "a farmer in a paddy field at sunrise with modern irrigation equipment" and the AI produces an image that matches your description in seconds.

How does image AI learn? It was trained on hundreds of millions of images paired with text descriptions. Over time, it learned what things look like — what "sunrise", "paddy field", "farmer", and "irrigation" each contribute to the visual output — and how to combine them.

Use Case 1

Visual Content

Posters, social media graphics, book illustrations, presentation visuals — created from text prompts in seconds.

Use Case 2

Education

Teachers use it to create diagrams, historical scene illustrations, and concept visualisations when no ready image exists.

Use Case 3

Business

Small businesses create product mockups, logo ideas, and ad visuals without hiring a graphic designer.

Use Case 4

Medical Imaging

AI analyses real medical images (X-rays, MRI scans) to spot signs of disease — this is computer vision, a related but different type.

The deepfake problem: The same image AI technology can create fake photos of real people — making it appear they said or did things they never did. These are called deepfakes. In India, several cases of deepfake misuse have been reported, including fake videos of celebrities and politicians. Always ask: is this a real photo or AI-generated? Reverse image search can help verify.

Copyright question: Image AI was trained on images from the internet, many of which were created by human artists. The question of whether this is fair to the original creators is still being debated in courts around the world. As a learner, be aware that AI images should not be passed off as your own original artwork.

Section 4 of 8

🎙️ Voice AI — Computers That Listen and Speak

Voice AI covers two related skills: understanding spoken words (called speech recognition) and producing spoken words from text (called text-to-speech or TTS). Google Assistant, Alexa, and Siri use both.

How speech recognition works:

Your voice is captured as a sound wave
The AI breaks the wave into small segments (phonemes — the basic sound units of language)
It matches those phonemes to words using a language model
It figures out the most likely sentence you said
A second AI (Language AI) interprets the meaning and generates a response
A TTS system converts the response back into speech you hear

Why accents matter: Voice AI is trained on voice recordings. If the recordings mostly came from one accent or dialect, the AI will understand that accent best. Early versions of Google Assistant understood American English much better than Indian-accented English or Telugu. This has improved as more Indian voice data has been collected — but regional accents and rural dialects still present challenges.

Voice AI in India today:

Google Assistant supports Telugu, Hindi, Tamil, Kannada, Bengali, and more — fully voice-driven
IRCTC voice booking — you can book train tickets by speaking in Hindi
Bhashini (Government of India) — a national AI translation and voice platform supporting 22 Indian languages
DigiLocker voice access — access your Aadhaar, degree certificates, and driving licence by speaking

Accessibility impact: Voice AI has been transformative for people who cannot read — elderly citizens who never learned to use a smartphone can now use voice commands in their own language to do banking, get government services, and communicate. This is one of the most socially important applications of AI in India.

Section 5 of 8

📺 Recommendation AI — The Invisible Curator

You already know this one from Lesson 2 — but let us go one level deeper. Recommendation AI is used in far more places than YouTube.

Platform	What it recommends	How it personalises
YouTube	Videos	Watch time, completion rate, what similar viewers liked
Netflix / Hotstar	Movies and shows	What you have watched, at what time, on what device, how quickly you skipped past content
Amazon / Flipkart	Products	What you searched for, what you put in cart but did not buy, what similar buyers purchased
Swiggy / Zomato	Food orders	Your past orders, time of day, weather in your area, what nearby customers ordered recently
Spotify / JioSaavn	Music	Songs played fully, songs skipped, listening time, mood patterns (morning vs night)
Instagram / Reels	Short videos	How long you paused, whether you re-watched, whether you shared, your comment history

The business model behind recommendations: These platforms make money when you spend more time on them. Recommendation AI is designed to maximise your engagement (time spent + clicks + purchases) — not to maximise your wellbeing. This is an important distinction that every digital citizen should understand.

Interesting data point: In 2023, Zomato reported that their AI recommendation engine could predict a user's next food order with 35% accuracy just from browsing patterns — before the user even searched for anything. This is how personalisation works at its most advanced: predicting intent before you act on it.

Section 6 of 8

👁️ Computer Vision AI — Machines That See

Computer Vision AI analyses images and video to understand what is in them — objects, faces, text, actions, diseases, defects. It is one of the oldest and most developed areas of AI.

Application 1

Face Recognition

Phone unlock, Aadhaar eKYC, airport boarding, school attendance — identifying people from their facial features.

Application 2

Crop Disease Detection

Plantix analyses photos of leaves and identifies 400+ crop diseases. Free app used by millions of Indian farmers.

Application 3

Medical Imaging

Aravind Eye Hospital's AI screens retinal scans for diabetic blindness. Qure.ai analyses chest X-rays for tuberculosis.

Application 4

Traffic and Safety

Traffic cameras detect violations, count vehicles, and identify dangerous driving. Highway cameras spot potholes automatically.

Application 5

Text Recognition (OCR)

Google Lens reads text in any image — road signs, restaurant menus, handwritten notes. Also translates in real time.

Application 6

Quality Control

Factories use cameras + AI to spot defective products on a conveyor belt faster and more accurately than human inspectors.

India medical AI achievement: The AI screening system deployed by Aravind Eye Care (the world's largest eye hospital, based in Tamil Nadu) has screened over 3 million eyes for diabetic retinopathy. Rural patients who would never have access to a specialist ophthalmologist get an AI-assisted screening in their district health centre. This is AI being used to reduce healthcare inequality — not increase it.

Privacy concern with Computer Vision: CCTV + face recognition is now deployed in many Indian public spaces — airports, railway stations, some schools and malls. Unlike a human security guard who forgets you after you leave, AI records and can track your movements indefinitely. This raises serious questions about surveillance, consent, and civil liberties that Indian courts are beginning to address.

Section 7 of 8

📈 Prediction / Forecasting AI — Seeing Tomorrow in Today's Data

Prediction AI uses patterns from historical data to estimate what will happen in the future. It is one of the most widely used — and highest-stakes — types of AI.

Domain	What it predicts	Why accuracy matters
Weather	Rain, temperature, cyclone tracks	Farmers need advance notice. Cyclone warnings save lives.
Agriculture	Crop yield, pest outbreaks, irrigation needs	India has 140 million farm families — better predictions mean better harvests.
Healthcare	Patient readmission risk, disease outbreak spread	Early warning allows hospitals to prepare and treat sooner.
Finance	Credit risk (will this person repay a loan?)	Can expand or restrict access to credit for millions of people.
Sports analytics	Player performance, injury risk, match outcome	Indian Premier League teams use AI for player selection and strategy.
Traffic	Road congestion, travel time, accident risk	Google Maps, NHAI smart highways, ambulance routing.

High-stakes prediction: Credit scoring
When a bank or fintech app decides whether to approve your loan, an AI often makes or influences that decision. It looks at your income, spending history, repayment record, and sometimes even your phone behaviour (how often you call, your GPS history, what apps you use). This is powerful — but also raises fairness concerns. What if the training data had historical biases against rural borrowers or women? The AI may perpetuate those biases without any human realising.

Agriculture prediction in India: The ICRISAT and Microsoft India collaboration used satellite imagery, weather data, and soil sensors to give farmers in Andhra Pradesh AI-powered sowing advisories. In the first year, farmers who followed the advice saw a 30% higher yield than those who used traditional methods alone. Prediction AI, when it works well, is one of the most powerful tools for rural development.

Section 8 of 8

🧰 Which AI Tool for Which Task?

Now that you know the six families, the practical skill is: given a task, which type of AI should I use?

Task

Write or improve text

→ Language AI (ChatGPT, Gemini). Draft first, then edit in your own voice.

Task

Create a visual / image

→ Image AI (DALL-E, Firefly, Canva AI). Describe what you want in detail.

Task

Control something hands-free

→ Voice AI (Google Assistant, Alexa). Set alarms, send messages, check weather.

Task

Find something new to watch

→ Recommendation AI is already doing this. Use it intentionally — or search directly instead.

Task

Identify something in a photo

→ Computer Vision AI (Google Lens, Plantix). Point, scan, get the answer.

Task

Plan for future conditions

→ Prediction AI (weather app, IMD alerts, crop advisory tools).

The most important rule: No AI tool is perfect for all tasks. A chatbot is not a reliable fact checker. A weather app cannot write your essay. A recommendation engine is not a career counsellor. Always match the tool to the task — and always verify the output if accuracy matters.

Bridge to Lesson 5: Now you have a map of the whole AI tools landscape. In Lesson 5 we will travel across India and meet the real people and organisations who are using all six families of AI tools to solve uniquely Indian problems — from saving crops to restoring ancient manuscripts to guiding ambulances through city traffic.

🎯 Quick Quiz — 10 Questions to Check What You Learned

Q1. Raju's akka uses ChatGPT to help write an essay introduction. Which family of AI tools is she using?

Q2. The hospital's AI scanned Tatha's chest X-ray and flagged two areas for the doctor to review. What type of AI is this?

Q3. A Language AI tool like ChatGPT produces a very confident answer that turns out to be factually wrong. What is this error called?

Q4. Swiggy recommends biryani to you at 1 PM every Wednesday even though you never searched for it. What does this show about Recommendation AI?

Q5. A farmer uses Plantix to photograph a diseased paddy leaf and identify the problem. Which family of AI is Plantix using?

Q6. Which statement best describes the business model behind YouTube's Recommendation AI?

Q7. Why do early voice assistants understand American English better than Indian accents?

Q8. You need to create a poster for your school's science exhibition. Which type of AI tool is most directly useful for generating a visual?

Q9. Aravind Eye Hospital's AI screens retinal scans to detect diabetic blindness in rural patients. This is an example of AI being used to:

Q10. A fake video makes it appear that a public figure said something they never actually said. What is this called?

0/10

questions correct

📝 Activity Sheet — AI Tool Matcher

Tip: in the print dialog, choose "Save as PDF" to download.

For each task below, identify the best type of AI tool and write one real example app or service you know. Then write one risk or limitation to watch out for. Use your notebook.

#	Task	Best AI Family	Real app example	One risk to watch
1	Check if your paddy crop has a disease
2	Get a first draft of a birthday speech
3	Create an illustration for your project cover page
4	Find out if it will rain in your district tomorrow
5	Set a reminder without typing — just speak it
6	Discover a new movie to watch tonight

Reflection (write in notebook):

Which of the six AI families do you use most often in your daily life? Did any family surprise you — one you didn't realise you were already using?
For which tasks would you never trust AI to give you the final answer without a human double-check? Why?
Think about someone in your family who is elderly or not comfortable with technology. Which of the six AI families could most improve their daily life? How would you explain it to them?

Use this table in your notebook today, or print this page directly if helpful.

👨‍👩‍👧 For Parents and Teachers

What this lesson covers: This is Lesson 4 of 12 in the Class 6 full-year AI curriculum. Students take a guided tour of the six families of AI tools: Language AI, Image AI, Voice AI, Recommendation AI, Computer Vision AI, and Prediction/Forecasting AI. Each family is explained with India-specific examples, real applications, and honest discussion of limitations and risks.

New vocabulary introduced:

Large Language Model (LLM) — the AI behind chatbots like ChatGPT and Gemini
Hallucination — when an AI produces confidently wrong information
Deepfake — AI-generated fake images or videos of real people
Speech recognition and Text-to-Speech (TTS) — the two sides of Voice AI
Computer Vision — AI that interprets images and video

Discussion topics for home or class:

Show your child a newspaper article. Ask them: "Could AI have written this? How would you check?"
Look at a social media image together. Ask: "Is this real or AI-generated? How do you know?"
Ask: "Of all the AI tools we discussed today, which one do you think has helped Indian families the most? Why?"

Learning time: Around 60–75 minutes. The AI Tool Matcher worksheet works well as individual work or in pairs.

Safety by design: No personal data collected. No login required. Deepfake and privacy risks are addressed honestly and age-appropriately throughout the lesson.