Ai 2

Question 1

True or False: AI can only perform tasks that it has been explicitly programmed to do.

Answer 1

False

Question 2

Fill in the blank: Machine learning is a subset of ______.

Answer 2

Artificial Intelligence

Question 3

Which of the following is a common use case of AI in healthcare? (a) Inventory management (b) Disease diagnosis (c) Scheduling meetings

Answer 3

b) Disease diagnosis

Question 4

What is natural language processing (NLP)?

Answer 4

A field of AI that enables computers to understand, interpret, and respond to human language.

Question 5

True or False: AI can analyze large datasets faster than a human can.

Answer 5

True

Question 6

Short Answer: Name one advantage of using AI in customer service.

Answer 6

24/7 availability

Question 7

Which algorithm is commonly used in supervised learning?

Answer 7

Linear regression.

Linear regression is a fundamental algorithm in AI that models the relationship between input features and a continuous output by fitting a straight line to the data

Question 8

What is reinforcement learning?

Answer 8

A type of machine learning where an agent learns to make decisions by receiving rewards or penalties.

Question 9

What is the Turing Test used for?

Answer 9

To determine whether a machine exhibits human-like intelligence.

Question 10

Which of the following is NOT a type of AI? (a) Narrow AI (b) General AI (c) Super AI (d) Quantum AI

Answer 10

d) Quantum AI

Question 11

Short Answer: What is an example of AI in autonomous vehicles?

Answer 11

Self-driving technology

Question 12

Give an example input and output for linear regression - use case thermostat

Answer 12

Input: Outside temperature

Output: Heating time needed

Analogy: If it gets colder outside, you need to run your heating longer. Linear regression finds the best-fit line to predict how long to heat your home for a given temperature.

Question 13

What is computer vision?

Answer 13

A field of AI that enables computers to interpret and make decisions based on visual data.

Question 14

How does an SDK work

Answer 14

Simplified Answer:
An SDK (Software Development Kit) gives developers everything they need—like pre-made code, instructions, and testing tools—so they can build apps quicker and make sure they work properly on a specific system.

Example:
If you’re making an app for iPhones, you use Apple’s iOS SDK. It gives you code and tools so you don’t have to figure out everything yourself, like how to make the app open the camera.

Question 15

What is the difference between supervised and unsupervised learning?

Answer 15

Supervised learning uses labeled data, while unsupervised learning does not.

Question 16

Which technology is often used in AI chatbots?

Answer 16

Natural Language Processing (NLP)

Question 17

What is a neural network?

Answer 17

A computational model inspired by the human brain, used in machine learning.

Question 18

What is the role of data in AI?

Answer 18

To train, validate (fine-tune), and test the model.

Question 19

Short Answer: Name one ethical concern related to AI.

Answer 19

Bias in decision-making

Question 20

Which of the following is a potential use of AI in agriculture? (a) Pest control (b) Weather forecasting (c) Both a and b

Answer 20

c) Both a and b

Question 21

What is linear regression

Answer 21

Linear regression is a way to find the best straight line through a bunch of dots on a graph, so you can make predictions. For example, if you know how study time affects test scores, you can use it to guess someone’s score based on how long they studied.

Question 22

What is a continuous output in machine learning?

Answer 22

A continuous output is a result that can take any numerical value within a range. It’s common in regression tasks, like predicting someone’s weight or a house price.

This contrasts with a fixed category output, used in classification tasks. For example, predicting whether an email is spam or not spam is a classification problem with fixed categories.

Question 23

Which algorithm is commonly used in unsupervised learning?

Answer 23

K-means clustering

This is a method used when we don’t have the answers in advance. It looks at lots of data and tries to group similar things together – like sorting socks by colour without knowing what each colour is called.

It works in steps and keeps repeating them until things stop changing. This kind of repeating process is called iterative – it means doing something again and again, each time making small improvements, like adjusting a recipe bit by bit until it tastes right.

Question 24

What is a continuous output in AI?

Answer 24

A continuous output is when an AI model predicts any numerical value within a range, rather than choosing from fixed categories.
It is typically used in regression problems.
Example: Predicting house prices, where the output could be £300,500 or £301,750, not just “cheap” or “expensive.”

Question 25

What is an agent (in AI)?

Answer 25

An agent in AI is anything that perceives its environment, processes information, makes decisions, and takes actions to achieve a goal. It might use sensors (like cameras, microphones, or software APIs) to collect data, but the agent is the thing that decides and acts. Example 1: A chatbot is an agent. It reads your message (input), thinks about it (processing), and sends a reply (action). Example 2: A robot vacuum cleaner is an agent. Its sensors detect walls and furniture, but the agent decides how to move and clean the room.

Question 26

Q: What is the opposite of a continuous output in AI?

Answer 26

A categorical output typically handled by classification models. It predicts fixed categories like “spam” or “not spam,” rather than any value within a range.

Question 27

Q: What does binary mean?

Answer 27

A: Binary means something with exactly two parts, options, or states, like “yes/no” or “on/off.”

Question 28

What is deep learning?

Answer 28

Deep learning is a subset of machine learning that uses neural networks with many layers to model complex patterns in large amounts of data.

Question 29

What are common applications of deep learning?

Answer 29

Deep learning is commonly used for:- - image and speech recognition - natural language processing - medical diagnosis - autonomous systems.

Question 30

What is shallow learning?

Answer 30

Shallow learning refers to traditional machine learning algorithms like decision trees or logistic regression, which typically have fewer layers and require manual feature engineering.

Question 31

How do deep learning and shallow learning differ in handling data?

Answer 31

Deep learning:- can automatically learn complex features from large, unstructured datasets Shallow learning:- relies more on predefined features and smaller datasets.

Question 32

When would you choose shallow learning over deep learning?

Answer 32

When: • The dataset is small • The problem is simple • Interpretability matters — meaning we need to understand how the model made its decision

Question 33

Can shallow learning be used in supervised and unsupervised learning?

Answer 33

Yes. Shallow learning can be used in both: • Supervised learning (e.g. logistic regression, decision trees) • Unsupervised learning (e.g. k-means, PCA)

Question 34

True or False: Logistic Regression is only used for regression problems.

Answer 34

False — despite the name, Logistic Regression is used for classification problems.

Question 35

Which of these is an ensemble method that combines multiple decision trees? A. Logistic Regression B. Naive Bayes C. Random Forest D. K-Nearest Neighbours

Answer 35

C. Random Forest

Question 36

__________ Regression models the probability that a given input belongs to a particular category using the logistic function.

Answer 36

Logistic

Question 37

Naive Bayes is especially well-suited for which of the following tasks? A. Image recognition B. Spam detection C. Stock market prediction D. Audio processing

Answer 37

B. Spam detection

Question 38

True or False: Random Forest is more interpretable than a single Decision Tree.

Answer 38

False — Random Forest is generally less interpretable than a single Decision Tree due to its ensemble nature.

Question 39

Which model combines the output of multiple trees to improve prediction accuracy and reduce overfitting?

Answer 39

Random Forest

Question 40

What is the key idea behind the Naive Bayes algorithm? A. It builds multiple decision trees and averages them B. It assumes all features contribute equally and independently to the outcome C. It draws a line to separate data points D. It creates clusters of similar data points

Answer 40

B. It assumes all features contribute equally and independently to the outcome

Question 41

What is AI

Answer 41

The capability of machines to mimic human intelligence

Question 42

Q: What is feature engineering?

Answer 42

A: Manually selecting or creating input variables (features) from raw data to help a machine learning model perform better.

Question 43

Q: How does feature engineering differ in deep learning?

Answer 43

A: Deep learning models learn features automatically from raw data, while shallow models rely on manual feature engineering.

Question 44

Q: What’s an analogy to understand manual vs automatic feature engineering?

Answer 44

Manual feature engineering is like a photographer manually adjusting the camera settings—choosing the right aperture, shutter speed, ISO, and framing for each shot. It takes skill and domain knowledge. Automatic feature engineering is like using a modern smartphone camera in auto mode, which uses AI to detect the scene, optimise settings, and enhance the image without needing you to intervene. It figures out the best way to represent the data (features) by itself.

Question 45

What is a token in language models like ChatGPT?

Answer 45

A token is a small chunk of text, like a word or part of a word, used in processing and generating language. (Example: “unbelievable” might be split into “un”, “believ”, and “able.”)

Question 46

How does ChatGPT generate responses using tokens?

Answer 46

It predicts one token at a time, choosing the next most likely token based on the input and its training data.

Question 47

What does “temperature” mean in AI models?

Answer 47

Temperature controls randomness in the model’s output. • Low temperature = more predictable • High temperature = more creative or random

Question 48

Why might you get a different answer to the same question in ChatGPT?

Answer 48

Because the model includes controlled randomness (via token sampling and temperature), even identical prompts can produce varied answers.

Question 49

What is token sampling?

Answer 49

Token sampling is the process where ChatGPT picks the next token from a list of likely options, not always the top one, to create varied outputs.

Question 50

How can you make ChatGPT responses more predictable?

Answer 50

Lower the temperature (if possible) and use more specific prompts to reduce randomness.

Question 51

What does it mean when we say “interpretability matters” in machine learning?

Answer 51

It means we need to understand and explain how the model made its decision. Examples: • Doctors explaining why a patient was diagnosed • Banks justifying why a loan was denied Shallow models (like decision trees) are more interpretable than deep learning models (like neural nets).

Question 52

What’s the difference between linear and logistic regression?

Answer 52

• Linear regression predicts a number (e.g. house price) • Logistic regression predicts yes or no (e.g. is this spam?)

Question 53

Classification predicts […] labels like “spam” or “not spam,” while regression predicts […] values like house price or weight.

Answer 53

Classification predicts categorical labels like “spam” or “not spam,” while regression predicts continuous values like house price or weight.

Question 54

Overfitting is when a model learns the […], not the actual patterns — like memorising past exam questions without understanding the subject.

Answer 54

Overfitting is when a model learns the noise in the training data, not the actual patterns — like memorising past exam questions without understanding the subject.

Question 55

Embeddings are like turning words into […], so a model can understand relationships — similar words get […] numbers.

Answer 55

Embeddings are like turning words into numbers, so a model can understand relationships — similar words get similar numbers.

Question 56

Underfitting happens when a model is too […] to capture the underlying trend — like trying to fit a […] through curved data. It performs poorly on both training and test data because it hasn’t […] enough.

Answer 56

Underfitting happens when a model is too simple to capture the underlying trend — like trying to fit a straight line through curved data. It performs poorly on both training and test data because it hasn’t learned enough