Class 7 · Unit 2: Domains of AI · Lesson 2.4 · 40 min

Chatbots, image recognition and translation

A tourist points her camera, types a question, speaks into a translator — three apps in two minutes. Are they all the same AI?

Today you will: Watch three AI apps work, live · Log what goes in and what comes out · Name each app's domain — and how to check it

Story

A tourist from Chennai stands at the Dal Lake boulevard with her phone. Hiba and Bilal watch her use three apps in two minutes:

She points her camera at a flower: "Lotus. Nelumbo nucifera."
She types into a chatbot: "Best time to see Dal Lake at sunrise?"
She speaks into a translator and shows the shikara-wala the screen, in Urdu.

Bilal: “Three apps. Are they all the same AI?”

Hiba: “Look at what goes in and what comes out. That will tell us.”

What is really happening inside each of these apps?

Watch

Three apps, three domains — — scan code 7.2.4 in the printed book.

Warm up your fingers

Skill: numbers inside sentences (7 min)

On typing.com, open a number-row review lesson.
Then type these three lines accurately. Accuracy matters more than speed: Sunrise at 6:10 on 15 May. Shikara ride: 2 hours, 4 people. Pahalgam is 95 km away.

On the laptop

You need: LibreOffice Calc · app-detective.ods

Mission 1: Teacher demo (whole class, 8 min)

☐1
The teacher runs three apps on the projector:
- Image recognition on 3 local photos — a chinar leaf, a samovar, a houseboat.
- Translation of "Where is the nearest bus stop?" into Urdu — and back. Did the meaning survive?
- A school-approved chatbot, asked one fact about Kashmir. The class checks it in the textbook.
☐2
Predict before each one: will it get it right?

Mission 2: Your log (pairs, 10 min)

☐3
Open app-detective.ods. One row per app: what went IN · what came OUT · domain(s) · did it make a mistake?
☐4
Add a fourth row for an app your family uses — a voice assistant, a keyboard that suggests words, a map.
☐5
Last column: how could a person check the app's answer?

No laptop today?

Draw IN → MODEL → OUT three times in your notebook, one for each app in the story. Add one app from home. Swap with a partner: did you name the same domain?

Now you know

The tourist's three apps: a photo of a flower goes into computer vision and 'Lotus' comes out; a typed question goes into NLP with data and an answer comes out that may be confidently wrong; a spoken question goes into NLP and Urdu text comes out, worth checking with a round trip. Using an app is not trusting it.

Every AI app is input → model → output.
Image recognition is computer vision. Translation is NLP. Chatbots are NLP, often with data science.
Many apps combine domains.
Confidently wrong happens. Important answers get checked by a person.
Stop – Check – Share still applies — using an app isn't trusting it.

Debate it

The tourist's chatbot tells her that the Mughal Gardens are open at 5 a.m. They are not, and she waits outside for two hours.

Whose mistake is it: the chatbot's, the company's, or hers for not checking? What should the app have said instead?

Check yourself — practice, not a test

1. Match each app to its main domain:
Photo app that names a flower · English-to-Urdu translator · App that predicts next week's apple prices from past sales
Match with: Natural Language Processing · Computer Vision · Data Science
2. Put the parts of an AI app in order:
- 1. **Input
- 2. Output**
- 3. Model
3. If a chatbot sounds confident, its answer must be correct.
- ○ True
- ○ False
4. A voice assistant answers "What is the weather in Gulmarg?" Which domains does it use? (choose all)
- ○ Natural Language Processing
- ○ Data Science
- ○ Computer Vision

Remember

In → model → out. Name the domain, and check the answer.

Story

A tourist from Chennai stands at the Dal Lake boulevard with her phone. Hiba and Bilal watch her use three apps in two minutes:

She points her camera at a flower: "Lotus. Nelumbo nucifera."
She types into a chatbot: "Best time to see Dal Lake at sunrise?"
She speaks into a translator and shows the shikara-wala the screen, in Urdu.

Bilal: “Three apps. Are they all the same AI?”

Hiba: “Look at what goes in and what comes out. That will tell us.”

What is really happening inside each of these apps?

On the laptop

You need: LibreOffice Calc · app-detective.ods

Mission 1: Teacher demo (whole class, 8 min)

☐1
The teacher runs three apps on the projector:
- Image recognition on 3 local photos — a chinar leaf, a samovar, a houseboat.
- Translation of "Where is the nearest bus stop?" into Urdu — and back. Did the meaning survive?
- A school-approved chatbot, asked one fact about Kashmir. The class checks it in the textbook.
☐2
Predict before each one: will it get it right?

Mission 2: Your log (pairs, 10 min)

☐3
Open app-detective.ods. One row per app: what went IN · what came OUT · domain(s) · did it make a mistake?
☐4
Add a fourth row for an app your family uses — a voice assistant, a keyboard that suggests words, a map.
☐5
Last column: how could a person check the app's answer?

No laptop today?

Draw IN → MODEL → OUT three times in your notebook, one for each app in the story. Add one app from home. Swap with a partner: did you name the same domain?

Now you know

Every AI app is input → model → output.

Image recognition is computer vision. Translation is NLP. Chatbots are NLP, often with data science.

Many apps combine domains.

Confidently wrong happens. Important answers get checked by a person.

Stop – Check – Share still applies — using an app isn't trusting it.

Check yourself — practice, not a test

1. Match each app to its main domain:

Photo app that names a flower · English-to-Urdu translator · App that predicts next week's apple prices from past sales
Match with: Natural Language Processing · Computer Vision · Data Science

2. Put the parts of an AI app in order:

1. **Input
2. Output**
3. Model

3. If a chatbot sounds confident, its answer must be correct.

○ True
○ False

4. A voice assistant answers "What is the weather in Gulmarg?" Which domains does it use? (choose all)

○ Natural Language Processing
○ Data Science
○ Computer Vision