• Foundation
  • About 20 min
  • 15 practice questions

Final self-assessment: AI in India and everyday life

Test yourself on the whole course: 15 questions with answers explained.

If something looks wrong, please tell us.

Attempt all 15 questions before looking at any answer. Write your choice down for each, then check them together and read every explanation, including those you got right. The questions cover all five lessons and are a little harder than the lesson checks.

Final self-assessment

  1. 1. A spam filter is trained on thousands of emails already marked 'spam' or 'not spam', and then sorts new emails. Which statement is correct?

    1. It is general AI because it handles many emails
    2. It follows only rules typed in by a programmer
    3. It is machine learning: it learned patterns from labelled examples
    4. It cannot make mistakes because it has seen thousands of emails
    Show answer

    Answer: C. Learning from labelled examples and applying the patterns to new cases is machine learning. Handling many emails of one kind is still narrow AI, not general AI. It did not rely only on hand-typed rules, and any trained system can still make mistakes.

  2. 2. After a system is trained, it is checked on examples it has never seen before. Why?

    1. To find out how well it works on new cases, not just the ones it memorised
    2. To make the system run faster
    3. To reduce the electricity bill
    4. To delete the training data
    Show answer

    Answer: A. Testing on unseen examples shows whether the system has learned general patterns or only memorised its training data. It is not about speed, electricity or deleting data.

  3. 3. A shopping app shows Kavya only products similar to those she bought before, and she rarely sees anything new. Which idea explains this?

    1. Encryption of her data
    2. Face unlock on her phone
    3. A fraud alert on her account
    4. A recommendation system creating a narrow 'filter bubble'
    Show answer

    Answer: D. Recommendation systems predict what you will like from past behaviour, which can keep showing similar items and narrow what you see - a filter bubble. Encryption, fraud alerts and face unlock are unrelated features.

  4. 4. A satellite image shows one part of a wheat field reflecting much less near-infrared light than the rest. What is the most reasonable conclusion?

    1. That part has the healthiest plants in the field
    2. That part is likely to be under stress, such as from water shortage or disease, and should be checked
    3. The satellite camera is broken
    4. That part has been harvested already and needs no attention
    Show answer

    Answer: B. Healthy plants reflect strongly in near-infrared, so a patch reflecting less suggests stressed plants, from water shortage, pests or disease. It is a reason to inspect, not proof of a broken camera or a harvested field.

  5. 5. Which combination of inputs would an AI flood model most need to warn villages downstream?

    1. Mandi prices, crop types and fertiliser sales
    2. School exam results and census data
    3. Mobile recharge records and bank balances
    4. Rainfall forecasts, upstream river levels and the shape of the land
    Show answer

    Answer: D. Flooding depends on how much rain falls, how high rivers are already rising upstream and how water flows over the terrain. Prices, recharges, balances or exam results do not predict where water will go.

  6. 6. A crop app is very accurate in the region where it was trained but often wrong in a distant region with different crop varieties. The best explanation is:

    1. Its training data did not represent the new region's crops and conditions well
    2. AI only works in the morning
    3. The phones in the distant region are slower
    4. Farmers in the distant region take photos upside down
    Show answer

    Answer: A. An AI is only as good as its data. If the training images came from other varieties and conditions, the patterns it learned may not fit the new region. Phone speed, time of day or photo orientation do not explain a regional drop in accuracy.

  7. 7. In a hospital using AI to read chest X-rays, who is responsible for the final diagnosis?

    1. The company that sold the software
    2. The AI tool, because it read the image first
    3. The doctor, using the AI result together with tests and examination
    4. The patient
    Show answer

    Answer: C. AI output on a scan is a flag that supports the doctor. The doctor, who also uses confirmatory tests and the examination, makes and is responsible for the diagnosis. Neither the tool, its seller nor the patient takes that role.

  8. 8. Why is AI screening of retina photos especially useful where eye specialists are few?

    1. It cures diabetic retinopathy
    2. It checks many photos quickly and picks out those who need a specialist
    3. It removes the need to test blood sugar
    4. It works without any photograph
    Show answer

    Answer: B. AI screening sorts large numbers of retina photos so specialists can concentrate on the patients who need them. It does not cure the disease, replace blood-sugar testing, or work without a photo.

  9. 9. Translating study videos and notes into Indian languages mainly helps by:

    1. Letting students learn in the language they understand best
    2. Making exams shorter
    3. Removing the need for teachers
    4. Increasing the price of textbooks
    Show answer

    Answer: A. Translation lets students study in their mother tongue or strongest language, which helps understanding. It does not shorten exams, replace teachers or raise textbook prices.

  10. 10. Match the item to its description: (1) #AIForAll, (2) Bhashini, (3) IndiaAI Mission. Which pairing is correct?

    1. (1) a translation platform, (2) a 2018 strategy theme, (3) a satellite
    2. (1) a 2024 Cabinet approval, (2) NITI Aayog's paper, (3) a payment app
    3. (1) a cyber helpline, (2) an exam board, (3) a weather model
    4. (1) theme of NITI Aayog's 2018 AI strategy, (2) Indian-language translation platform, (3) mission approved by the Union Cabinet in 2024
    Show answer

    Answer: D. #AIForAll is the theme of NITI Aayog's 2018 National Strategy for Artificial Intelligence; Bhashini is the government's Indian-language translation platform under MeitY; the IndiaAI Mission was approved by the Union Cabinet in 2024. The other pairings mix these up.

  11. 11. A city's grievance system uses AI to read complaints and send each to the right department. What remains the job of officials?

    1. Nothing: the AI also fixes the problem
    2. Resolving the complaint and being accountable for the outcome
    3. Translating every complaint by hand
    4. Deleting complaints the AI cannot read
    Show answer

    Answer: B. AI can read, translate and route complaints, but officials resolve them and are accountable. Hand translation of everything defeats the purpose, and deleting unreadable complaints would deny people service.

  12. 12. Which principle is broken if a welfare benefit is stopped automatically by an AI flag, with no official check and no way for the person to ask why?

    1. Data compression
    2. Faster internet
    3. Accountability and transparency
    4. Higher resolution cameras
    Show answer

    Answer: C. Decisions affecting people's rights need a responsible official (accountability) and an explanation the person can ask for (transparency). Internet speed, compression and camera resolution are technical matters, not principles of responsible AI.

  13. 13. A video shows a well-known actor urging viewers to put money into an app that 'guarantees double returns in 30 days'. The video looks real. What is the best response?

    1. Invest a small amount first to test it
    2. Share it with friends so they benefit too
    3. Trust it because the face and voice look genuine
    4. Treat it as a likely deepfake scam, check official sources, and report it
    Show answer

    Answer: D. Guaranteed high returns are a red flag, and deepfakes can make faces and voices look real, so appearance proves nothing. Check the person's official channels and report it on the platform and at cybercrime.gov.in. Testing with money or sharing it spreads the fraud.

  14. 14. Why is agreeing a family code word a good defence against voice-clone fraud?

    1. A cloned voice can copy how someone sounds but not a secret only the family knows
    2. It makes phone calls cheaper
    3. It blocks all unknown numbers
    4. It stops AI from recording voices
    Show answer

    Answer: A. A voice clone imitates sound, not private knowledge. Asking for a code word known only to the family exposes the fake. It does not change call cost, block numbers or stop recording.

  15. 15. A person realises he has just paid money to a fraudster. What should he do first?

    1. Wait a week to see if the money comes back
    2. Post about it on social media and do nothing else
    3. Report it at once on the helpline 1930 or at cybercrime.gov.in
    4. Send more money to 'unlock' the first payment
    Show answer

    Answer: C. Reporting quickly on 1930 or at cybercrime.gov.in improves the chance of tracing or stopping the money. Waiting loses time, a social media post is not a report, and paying again to 'unlock' money is another common trick.

Which lessons to revisit: questions 1-3, lesson 1 (AI around you); questions 4-6, lesson 2 (farming and weather); questions 7-9, lesson 3 (health and education); questions 10-12, lesson 4 (government and public services); questions 13-15, lesson 5 (jobs, skills and staying safe).

Sign in to keep track of the lessons you finish.

Privacy · Terms · Refunds · Delivery