AI Basics
Practice questions
A student uses a chatbot to prepare notes on exam dates and fees. What is the right practice?
- Copy the chatbot's answer because it sounds confident
- Ask the same chatbot again until it gives the same answer twice
- Use the answer only if it is long
- Check each date and fee against the official notice before relying on it
Answer and explanation
Answer: D
Chatbots can hallucinate, and repetition or confident tone does not make an answer true. Facts such as dates and fees must be checked against the official source.
A spam filter lets 12 scam emails into the inbox and moves 3 genuine emails to spam. How many false negatives and false positives are there?
- 3 false negatives and 12 false positives
- 15 false positives and no false negatives
- 12 false negatives and 3 false positives
- 12 false negatives and no false positives
Answer and explanation
Answer: C
Scams let through are cases where the filter said 'not spam' wrongly: 12 false negatives. Genuine emails marked spam are 'spam' said wrongly: 3 false positives.
Why can a chatbot state a wrong fact in a confident tone?
- It produces likely-sounding words from patterns rather than looking facts up in a checked record
- It deliberately lies to users
- Its screen is faulty
- It has general intelligence and disagrees
Answer and explanation
Answer: A
A language-model chatbot generates the most likely next words. It aims to sound right, not to verify facts, so a weak pattern can still produce fluent, false output - hallucination.
A factory arm repeats the same welding movement all day with no sensing or decision. It is best described as:
- Computer vision
- Automation, not AI
- Deep learning
- General AI
Answer and explanation
Answer: B
A machine repeating a fixed motion follows a set program. It becomes AI only when it senses its surroundings and makes decisions, as a robot finding its way around obstacles does.
An online shop groups its customers into types by their buying habits, with no labels given in advance. This is:
- Supervised learning
- Reinforcement learning
- An expert system
- Unsupervised learning
Answer and explanation
Answer: D
Finding groups in unlabelled data is unsupervised learning. Supervised learning needs labelled examples, and reinforcement learning uses rewards and penalties.
A smart speaker hears a spoken question, works out its meaning and replies aloud. Which sequence of branches is used?
- Computer vision → robotics → NLP
- NLP → speech recognition → computer vision
- Speech recognition → NLP → speech synthesis
- Expert system → computer vision → robotics
Answer and explanation
Answer: C
Voice is turned into text by speech recognition, the meaning is worked out by NLP, and the reply is spoken by speech synthesis (text-to-speech). No images or physical movement are involved.
A rules-based system has one advantage over most learning systems. Which?
- Its decisions are easy to explain because the rules can be read
- It needs huge datasets
- It handles messy, varied input better
- It improves automatically with more data
Answer and explanation
Answer: A
In a rules-based system you can point to the rule that caused a decision. Learning systems handle varied input and improve with data, but their decisions are often hard to explain.
Which activity is inference, not training?
- Adjusting a model's weights using thousands of labelled photos
- A phone unlocking when its owner looks at it
- Labelling mango photos as ripe or unripe
- Collecting a dataset of voice recordings
Answer and explanation
Answer: B
Unlocking a phone uses the already trained face model on a new input - that is inference. Adjusting weights is training, and labelling and collecting data are preparation for training.
A company uses the same learning algorithm on two different datasets - one of Hindi emails, one of English emails. What results?
- One model shared by both
- No model, because the algorithm is the model
- The datasets merge automatically
- Two different models, because a model depends on the data it was trained on
Answer and explanation
Answer: D
The algorithm is the learning method; the model is what that method produces from particular data. Different data give different models.
Why did expert systems prove brittle?
- Every rule had to be written and updated by hand, and they failed on cases their rules did not cover
- They needed millions of labelled photographs
- They could only run on smartphones
- They learned too quickly from data
Answer and explanation
Answer: A
Expert systems hold human-written if-then rules. Writing and maintaining those rules was slow and costly, and a case outside the rules left the system stuck.
How many years separate Alan Turing's 1950 paper from Deep Blue's 1997 win over Kasparov?
- 41
- 46
- 47
- 6
Answer and explanation
Answer: C
1997 − 1950 = 47 years. (Dartmouth 1956 to Deep Blue 1997 is 41 years, and Turing's paper to Dartmouth is 6 years.)
Arrange in time order: (1) Deep Blue beats Kasparov, (2) the term 'artificial intelligence' is coined for the Dartmouth workshop, (3) Turing proposes the imitation game, (4) deep learning spreads widely.
- 2, 3, 1, 4
- 3, 2, 4, 1
- 3, 1, 2, 4
- 3, 2, 1, 4
Answer and explanation
Answer: D
Turing's paper is 1950 (3), Dartmouth is 1956 (2), Deep Blue is 1997 (1), and deep learning spreads from around the 2010s (4). Order: 3, 2, 1, 4.
A student says: 'Because a chess computer can beat a world champion, it is general AI.' What is wrong?
- A chess computer is narrow AI: excellent at chess but unable to do unrelated tasks
- Nothing; beating a champion proves general intelligence
- Chess computers are not AI at all
- General AI only plays board games
Answer and explanation
Answer: A
General AI would handle any intellectual task. A chess program excels only at chess and cannot, say, translate or drive, so it is narrow AI.
Which of these is the best example of AI rather than a fixed-rule program?
- A geyser that switches off at a set temperature
- A lift that stops at the floor whose button is pressed
- A microwave timer that rings after two minutes
- A maps app that predicts a route's travel time from traffic patterns
Answer and explanation
Answer: D
Predicting travel time from past and live traffic patterns is judgement learned from data. The geyser, lift and timer each follow a single fixed rule with no learning or prediction.
Which pair is correctly matched?
- General AI - exists in everyday phones
- Narrow AI - also called weak AI
- Strong AI - another name for narrow AI
- Weak AI - can do any intellectual task a human can
Answer and explanation
Answer: B
Narrow AI is also called weak AI: it does one task or a small set of tasks. General AI (strong AI, AGI) would handle any intellectual task and has not been achieved, so the other pairs are wrong.
A fraud filter flags 50 payments: 42 are real fraud and 8 are genuine. It also lets 5 real frauds through. How many real frauds were there in total?
- 42
- 50
- 47
- 55
Answer and explanation
Answer: C
Real frauds = frauds caught + frauds missed = 42 + 5 = 47. The 8 genuine payments flagged are false positives and are not counted as fraud.
Why should a human check AI output before using it for an important decision?
- AI can make errors, lacks context and cannot be held accountable
- AI output is always written in a foreign language
- Checking makes the AI run faster
- The law says AI may never be used
Answer and explanation
Answer: A
AI can be wrong, can miss context a person would notice, and cannot take responsibility. A person catches errors and owns the final decision.
A face-recognition system trained mostly on photos of one group of people works poorly for others. This problem is called:
- Inference
- Bias
- Reinforcement
- Encryption
Answer and explanation
Answer: B
Unfair, uneven results caused by unbalanced training data are bias. The fix starts with more balanced, representative data and human review.
A fraud filter blocks a customer's genuine payment as suspicious. This is a:
- False negative
- Hallucination
- True positive
- False positive
Answer and explanation
Answer: D
The system said 'fraud' (positive) when the truth was 'not fraud', so it is a false positive. A false negative would be a real fraud let through.
In AI, 'hallucination' means:
- The AI system switching itself off
- A camera seeing images in the dark
- The AI giving confident but false or made-up information
- A virus attacking the AI
Answer and explanation
Answer: C
Hallucination is when an AI system, especially a chatbot, produces fluent, confident output that is false or invented, such as a fake reference or a wrong date.