Artificial Intelligence

**Choose Questions **

*Advice: Open Lectures when you read questions*

 Lecture 1

1. What is the primary focus of Lecture 1 in Artificial Intelligence (CAI 2603)?

a) History of AI
b) Agents in AI
c) Definition of AI
d) Overview of the course
2. According to the course, what would be a more suitable title for Artificial
Intelligence?**
a) Machine Learning
b) Computational Rationality
c) Cognitive Science
d) Robotics
3. How is rationality defined in the context of decision-making in AI?
a) Minimally achieving pre-defined goals
b) Selecting actions to maximize expected utility
c) Making decisions based on emotions
d) Following a fixed set of rules
4. What term is used in AI to refer to entities that perceive and act, selecting
actions to maximize expected utility.
a) Algorithms
b) Agents
c) Actuators
d) Sensors
5. What distinguishes a rational agent from a reflex agent?
a) A rational agent thinks about the consequences of its actions
b) A reflex agent considers the consequences of its actions
c) A rational agent acts solely based on the current state of the world
d) A reflex agent has long-term goals
6. What characterized the period known as the "AI Winter" in the history of AI?
a) A decline in expert systems development
b) A boom in the expert systems industry
c) The defeat of Kasparov by Deep Blue
d) A resurgence of statistical approaches
7. In terms of approaches, what marked the "AI Spring" in the 1990s?
a) A decline in probability-based methods
b) A focus on uncertainty
c) A resurgence of knowledge-based systems
d) The defeat of Kasparov by Deep Blue
8. What task is explicitly mentioned as something AI can currently do?
a) Perform a surgical operation
b) Write an intentionally funny story
c) Translate spoken Chinese into English in real-time
d) All of the above
9. What is the primary goal of a self-driving car in multi-task learning, as
mentioned in the lecture?
a) Identify pedestrians and signs
b) Play tennis effectively
c) Drive safely in Egyptian Media Production City
d) Detect cars and pedestrians, among other tasks
10. What prerequisites are mentioned for the course in terms of skills and
knowledge?
a) Advanced calculus and deep learning experience
b) Linear algebra, basic calculus, and probability
c) Programming in Java and familiarity with C++
d) Statistical analysis and proficiency in R programming
11. What is the term used to describe a self-driving car's ability to perform
multiple tasks such as detecting cars, pedestrians, signs, and lights?**
a) Monotaski
b) Single-task learning
c) Multi-task learning
d) Sequential learning
12. According to the lecture, what are the key lessons learned from the brain in
the context of decision-making?
a) The brain is modular and easily reverse-engineered.
b) Memory and simulation are crucial for decision-making.
c) Brains are perfect in making rational decisions.
d) The brain's decision-making is purely emotional.
 Lecture 2
1. In the healthcare domain, which of the following is NOT mentioned as a use of
Artificial Intelligence?
a) Precise diagnosis of lesions
b) Individualized treatment planning
c) Real-time surgical guidance
d) Development of new pharmaceuticals
2. What can be done with Poe, the Paper Model mentioned in the lecture?
a) Perform complex surgical procedures
b) Create 3D animations
c) Generate paper-based models from text descriptions
d) Simulate therapy outcomes
3. According to the lecture, what are the three pillars of AI?
a) Modularity, Simulation, Data
b) Modeling, Inference, Learning
c) Diagnosis, Planning, Execution
d) Precision, Adaptability, Decision-making
4. What is the main focus of the modeling paradigm in AI?
a) Answering questions about the model
b) Approximating real-world problems using formal mathematical objects
c) Choosing and applying techniques for each problem
d) Studying the complexity of computational algorithms
5. In the context of modeling, what is described as "lossy"?
a) Data collection process
b) The complexity of models
c) Surgical procedures
d) Approximation of real-world problems
6. What is the primary task of inference in AI?
a) Constructing abstract models
b) Answering questions about the model
c) Choosing appropriate techniques
d) Writing down abstract model families
7. In the learning paradigm, what is the focus of machine learning?
a) Constructing a full model
b) Turning abstract model families into concrete models
c) Writing down abstract models
d) Selecting appropriate inference techniques
8. What is the key challenge mentioned in modeling, according to the lecture?
a) Ensuring modularity of models
b) Determining the computational complexity to keep
c) Choosing the most advanced techniques
d) Defining formal ways for problems like Chess, Go, or Sudoku
9. How is the complexity of some models handled in inference?
a) By using lossy approximations
b) By constructing abstract models
c) By focusing on efficient algorithms
d) By collecting more data
10. In the context of learning, what does machine learning turn abstract model
families into?
a) Full models
b) Concrete models
c) Lossy models
d) Inferred models
11. According to the lecture, what is the focus of healthcare applications in
Artificial Intelligence?
a) Pre-surgical planning
b) Real-time surgical guidance
c) Diagnosis and prognosis
d) Therapy simulation and outcome
12. Which foundation model is mentioned for image generation based on text
description?
a) Tesla
b) HAL 9000
c) DALL-E (OpenAI)
d) DeepMind
13. What is the main application of DALL-E as discussed in the lecture?
a) Autonomous driving
b) 3D animation creation
c) Image generation based on text description
d) Surgical navigation
14. In the modeling paradigm, what is described as the process of approximating
real-world problems?
a) Learning
b) Inference
c) Modeling
d) Decision-making
15. What does the lecture define as the "three pillars" of AI?
a) Data, Analysis, Optimization
b) Modeling, Inference, Learning
c) Planning, Execution, Evaluation
d) Diagnosis, Treatment, Recovery
16. What is the primary focus of the learning paradigm in AI?
a) Constructing full models
b) Turning abstract model families into concrete models
c) Writing down abstract models
d) Selecting appropriate inference techniques
lecture 3
1. What does an agent refer to?
a. Only something that acts
b. Something that operates autonomously and pursues goals
c. Something that adapts to change
d. Both b and c
2. In the two views of AI, what does AGI stand for?
a. Augmented General Intelligence
b. Artificial General Intelligence
c. Adaptive General Interface
d. Augmented General Interface
3. What is the main characteristic of reflex-based models in machine learning?
a. Backtracking computations
b. Feed-forward computations
c. Complex computations
d. Recursive computations
4. In supervised learning, what is the task T?
a. Learn a mapping from inputs to outputs
b. Cluster similar data samples
c. Maximize defined reward
d. Explore the environment
5. What is the goal of unsupervised learning?
a. Learn a mapping from inputs to outputs
b. Explore the environment
c. Make sense of data and identify patterns
d. Minimize expected loss
6. What does RL stand for in the context of AI?
a. Reinforced Learning
b. Regression Learning
c. Reinforcement Learning
d. Recursive Learning
7. How is the interaction of an agent with its environment encoded in
reinforcement learning?
a. Through supervised learning
b. By observing human behavior
c. By means of a policy
d. Through unsupervised learning
8. What is the "alignment problem" in AI?
a. Aligning data with labels
b. Aligning computations
c. Aligning preferences and what algorithms optimize
d. Aligning neural networks
9. What is proposed as a solution to the "alignment problem"?
a. Explicitly specifying a reward function
b. Inverse reinforcement learning
c. Backtracking computations
 d. Minimizing expected loss
10. Which learning paradigm involves creating "smart tools" to aid human
decision-making?
a. Supervised Learning
b. Reinforcement Learning
c. Augmented Intelligence (IA)
d. Unsupervised Learning
11. In machine learning, what are inputs x also called?
a. Predictors
b. Labels
c. Responses
d. Targets
12. What is the main focus of inference in the modeling-inference-learning
paradigm?
a. Running fixed computations
b. Efficient algorithms to answer questions about the model
c. Creating abstract models
d. Identifying patterns in data
13. What does the term "reward hacking" refer to in AI?
a. Maximizing expected loss
b. Optimizing the reward function without capturing preferences
c. Creating smart tools
d. Augmented general intelligence
14. Which type of agent is expected to operate autonomously, perceive its
environment, and persist over time?
a. Rational agent
b. Reflex agent
c. Augmented agent
d. Unsupervised agent
15. What is the goal of unsupervised learning?
a. Maximize defined reward
b. Explore the environment
c. Make sense of data and identify patterns
d. Minimize expected loss
16. Which is an example of a reflex-based model?
a. Linear classifiers
b. Inverse reinforcement learning
c. Policy creation
d. Clustering
17. What is the task of reinforcement learning?
a. Grouping similar data samples
b. Making sense of data
c. Exploring the environment to maximize defined reward
d. Learning a mapping from inputs to outputs
18. What does AGI stand for in the context of the two views of AI?
a. Adaptive General Interface
b. Augmented General Intelligence
c. Artificial General Intelligence
d. Augmented General Interface
19. Which paradigm involves the process of turning an abstract model family into
a concrete model through learning?
a. Modeling-inference-learning paradigm
b. Augmented intelligence paradigm
c. Reflex-based paradigm
d. Reinforcement learning paradigm

Lecture 4
1. What is the key idea behind state-based models in artificial intelligence?
- A. Procedural solutions
- B. Modeling states and transitions
- C. Variable-based evaluation
- D. Chess and Go
2. In which application might state-based models be used for motion planning?
- A. Speech recognition
- B. Chess
- C. Robotics
- D. Sudoku
3. Which type of problem involves controlling everything in an environment with
no uncertainty?
- A. Search problems
- B. Constraint satisfaction problems
- C. Markov decision processes
- D. Adversarial games
4. What does a virtual assistant need to do in terms of information?
- A. Forget heterogenous information
- B. Digest heterogenous information
- C. Ignore deep reasoning
- D. Answer irrelevant questions
5. What should a good virtual assistant be able to do in terms of communication?
- A. Speak multiple languages
- B. Remember only recent information
- C. Interact using natural language
- D. Ignore user instructions
6. How do state-based models differ from variable-based models in terms of
solutions?
- A. Variable-based solutions are procedural
- B. State-based solutions specify step by step instructions
- C. Variable-based solutions are abstract
- D. State-based solutions don't consider order
7. Which type of model involves hard constraints like Sudoku and scheduling?
- A. State-based models
- B. Variable-based models
- C. Bayesian networks
- D. Markov decision processes
8. In variable-based models, what matters in the evaluation criteria of Sudoku?
- A. Order of filling squares
- B. Speed of filling squares
- C. Number of squares filled
- D. Type of numbers filled
9. What is the motivation for incorporating logic in virtual assistants?
- A. Reducing heterogeneity
- B. Digesting irrelevant information
- C. Drawing inferences from knowledge
- D. Communicating in multiple languages
10. How does a logic-based system differ from large language models (LLMs) in
terms of consistency?
- A. LLMs are more internally consistent
- B. Both are equally consistent
- C. Logic-based systems have hallucinations
- D. LLMs lack internal consistency
11. Which type of AI provides a class of models that is higher-level and requires
support from machine learning?
- A. Logical AI
- B. Statistical AI
- C. Artificial General Intelligence
- D. Augmented Intelligence
12. What does a Bayesian network involve in terms of variable dependencies?
- A. Hard constraints
- B. Soft dependencies
- C. Random variables
- D. Procedural dependencies
13. Which hardware components are often tuned for AI applications?
- A. Central Processing Unit (CPU)
- B. Random Access Memory (RAM)
- C. Graphics Processing Unit (GPU)
- D. Hard Disk Drive (HDD)
14. What does the acronym GPU stand for in the context of AI hardware?
- A. Graphics Processing Unit
- B. General Processing Unit
- C. Graphical Performance Unit
- D. General Performance Unit
15. What has deep learning systems exceeded in some vision tasks according to
the provided content?
- A. Human intelligence
- B. Machine learning capabilities
- C. Processing speed
- D. Human performance

Lecture 5
1. What does a search problem consist of?
- A. Only a state space
- B. Only a successor function
- C. A state space, a successor function (with actions, costs), a start state, and a
goal test
- D. Only a goal test
2. In the context of a search problem, what is the purpose of a successor function?
- A. To define the start state
- B. To determine the cost of actions
- C. To specify the goal test
- D. To describe the transition between states triggered by actions
3. How is a solution defined in a search problem?
- A. A sequence of cities
- B. A list of actions
- C. A sequence of states
- D. A sequence of actions (a plan) transforming the start state to a goal state
4. What does the state space in the context of search problems represent?
- A. The space between states
- B. A set of solutions
- C. A set of possible states
- D. The sequence of states
5. In the example of traveling in Romania, what does the state space consist of?
- A. Roads
- B. Cities
- C. Distances
- D. Actions
6. In the example of traveling in Romania, what is
the successor function?
- A. Traveling cost
- B. Roads between cities
- C. Cities with distances
- D. Actions with costs
7. What is the primary purpose of a search state in pathing problems?
- A. To include every last detail of the environment
- B. To update the location and dot booleans
- C. To keep only the details needed for planning
- D. To define the agent positions
8. How many world states are there in the specified scenario
with agent positions, food count, ghost positions, and agent
facing?
- A. 120x(230)x(122)x4
- B. 120
- C. 120x(230)
- D. 120x(230)x(122)x4

Lecture 6

1. What is a state space graph in the context of search problems?

a. A mathematical representation of plans
b. A mathematical representation of a search problem
c. A graphical representation of AI models
d. A set of goal nodes
2. In a state space graph, what do nodes represent?
a. World configurations
b. Actions
c. Goals
d. Plans
3. What is the key idea behind a search tree?
a. It represents only the start state
b. It is a mathematical model
c. It shows plans and their outcomes
d. It has no root node
4. How are nodes in a search tree related to states in a state space graph?
a. Nodes and states are the same thing
b. Each node corresponds to a single state
c. Nodes represent abstracted world configurations
d. States are plans within the tree
5. What does the root node of a search tree represent?
a. A goal state
b. The start state
c. An abstracted world configuration
d. The end state
6. In state space graphs vs. search trees, what does each node in the search tree
represent?
a. An entire path in the state space graph
b. A single state in the state space graph
c. A single action in the state space graph
d. The root node of the state space graph
7. How is a search tree constructed in comparison to a state space graph?
a. A search tree is always fully constructed in memory
b. A state space graph is constructed on demand
c. Both are fully constructed in memory
d. Both are constructed on demand
8. What is the purpose of tree search in AI?
a. To find the smallest tree
b. To expand as many tree nodes as possible
c. To maintain a fringe of partial plans
 d. To minimize tree expansion
9. What is the key idea of general tree search?
a. Exploration strategy
b. Fringe maintenance
c. Expansion of tree nodes
d. Root node expansion
10. What does the expansion step involve in tree search?
a. Reducing the number of nodes
b. Creating new nodes
c. Searching for goals
d. Ignoring the fringe
11. What is the main question in general tree search?
a. Which node to ignore
b. Which fringe nodes to explore
c. How to reduce the tree size
d. How to achieve the root node
12. What does the exploration strategy focus on in general tree search?
a. Minimizing fringe nodes
b. Maximizing tree expansion
c. Reducing the size of the search tree
d. Deciding which fringe nodes to explore
13. In the example tree search, what does the path "s → d → e → r → f → G"
represent?
a. A solution plan
b. A goal state
c. A world configuration
d. A root node
14. What is the purpose of maintaining a fringe of partial plans in tree search?
a. To ignore unnecessary nodes
b. To maximize tree expansion
c. To minimize tree size
d. To keep potential plans under consideration
15. What is the primary role of nodes in a search tree?
a. Representing abstracted world configurations
b. Showing states and corresponding actions
 c. Representing entire plans
d. Indicating the root node
16. How does a state space graph differ from a search tree?
a. In a state space graph, nodes represent entire plans.
b. In a search tree, nodes are abstracted world configurations.
c. Both state space graphs and search trees have the same structure.
d. A state space graph focuses on partial plans, while a search tree focuses on
entire plans.
17. What is the relationship between state space graphs and search trees?
a. State space graphs are a subset of search trees.
b. Search trees are a subset of state space graphs.
c. State space graphs and search trees are unrelated.
d. State space graphs and search trees are equivalent representations.

Lecture 7
1. What is the primary strategy of Depth-First Search (DFS)?
a) Expand the shallowest node first
b) Expand a node with the least cost
c) Expand the deepest node first
d) Expand nodes in random order
2. In Depth-First Search (DFS), which data structure is typically used for the fringe?
a) FIFO Queue
b) Priority Queue
c) LIFO Stack
d) Linked List
3. Which of the following best describes the time complexity of Depth-First Search
(DFS)?
a) O(b^s)
b) O(b^m)
c) O(b)
d) O(s)
4. What is the primary strategy of Breadth-First Search (BFS)?
a) Expand the shallowest node first
b) Expand a node with the least cost
c) Expand the deepest node first
d) Expand nodes in random order
5. In Breadth-First Search (BFS), which data structure is typically used for the
fringe?
a) FIFO Queue
b) Priority Queue
c) LIFO Stack
d) Linked List
6. Which of the following best describes the time complexity of Breadth-First
Search (BFS)?
a) O(b^s)
b) O(b^m)
c) O(b)
d) O(s)
Lecture 8
1. What is the main idea behind Iterative Deepening in search algorithms?
a) To increase the depth of the search with each iteration
b) To get the space advantage of DFS with the time advantages of BFS
c) To decrease the depth of the search with each iteration
d) To combine the strengths of BFS and UCS
2. What is the primary advantage of Uniform Cost Search (UCS) over BFS in terms
of finding paths?
a) UCS is faster than BFS
b) UCS always finds the least-cost path
c) UCS explores fewer nodes than BFS
d) UCS is more memory-efficient than BFS
3. In Uniform Cost Search, what does the "effective depth" refer to?
a) The maximum depth of the search tree
b) The depth of the last explored node
c) The depth of the cheapest solution found
d) The depth where the solution is guaranteed to be found
4. What is the time complexity of Uniform Cost Search in terms of effective depth?
a) O(b^C*)
b) O(b^d)
c) O(b^(C*/ε))
d) O(b^h)
5. What does A* Search use as its
evaluation function?
a) g(n) - backward cost
b) h(n) - heuristic cost
c) f(n) = g(n) + h(n)
d) b(n) - branching factor
6. When should A* terminate in a search
process?
a) When a goal is enqueued
b) When a goal is expanded
c) Only when a goal is dequeued
d) When the heuristic value becomes zero
7. Why might Greedy Search fail in finding an
optimal solution?
a) It always chooses the closest node
b) It does not use any heuristic function
c) It does not consider the cumulative cost
d) It can lead to a locally optimal solution
8. What is the key difference between Greedy Search and A* Search?
a) A* uses a heuristic function, while Greedy does not
b) Greedy is always faster than A*
c) A* considers both path cost and heuristic cost
d) A* is less informed than Greedy
9. What makes a heuristic admissible in A* Search?
a) It overestimates the true cost to the nearest goal
b) It underestimates the true cost to the nearest goal
c) It equals the true cost to the nearest goal
d) It is not related to the true cost
10. Why does an inadmissible heuristic break optimality in A* Search?
a) It slows down bad plans but outweighs true costs
 b) It speeds up good plans but underestimates true costs
c) It traps good plans on the fringe
d) It increases the time complexity of the algorithm
11. What is the significance of admissibility in heuristic functions?
a) It ensures the heuristic is always optimistic
b) It prevents the heuristic from being too slow
c) It guarantees the optimality of A* Search
d) It avoids the use of heuristics in search algorithms
12. In A* Search, what is the role of the function f(n) = g(n) + h(n)?
a) It estimates the heuristic cost
b) It calculates the backward cost
c) It combines the cumulative cost and heuristic cost
d) It determines the branching factor
13. What does the "effective depth" represent in Uniform Cost Search?
a) The depth of the cheapest solution found
b) The maximum depth of the search tree
c) The depth where the solution is guaranteed to be found
d) The depth of the last explored node
14. Which algorithm is known for expanding equally in all directions?
a) Uniform Cost Search
b) A* Search
c) Greedy Search
d) Depth-First Search
15. What is the goal of Iterative Deepening in search problems?
a) To decrease the depth of the search with each iteration
b) To get the space advantage of DFS with the time advantages of BFS
c) To increase the depth of the search with each iteration
d) To combine the strengths of BFS and UCS
16. Which search algorithm orders nodes based on the sum of backward cost and
heuristic cost?
a) Uniform Cost Search
b) Greedy Search
c) A* Search
d) Depth-First Search
17. What is the primary advantage of Uniform Cost Search over Greedy Search?
a) UCS is always faster than Greedy
b) UCS always finds the least-cost path
c) UCS explores fewer nodes than Greedy
d) UCS is more memory-efficient than Greedy
18. What does A* Search use to estimate the distance to the nearest goal for each
state?
a) Cumulative cost
b) Heuristic function
c) Depth limit
d) Branching factor
19. Which property ensures that A* Search is complete and optimal?
a) Monotonicity
b) Admissibility
c) Heuristic consistency
d) Optimality
20. What is the primary advantage of A* Search over Uniform Cost Search?
a) A* is always faster than UCS
b) A* always finds the least-cost path
c) A* explores fewer nodes than UCS
d) A* is more memory-efficient than UCS