NEW
Font size
WorksheetsAI - basic 1
Total questions: 40
Worksheet time: 20mins
Which of the following best describes Artificial Intelligence (AI) as introduced in the course?
Programming simple arithmetic functions
Making computers perform tasks that typically require human intelligence
Building only simple search engines
Restricting computation to deterministic tasks
According to the lecture, one way to define AI is making machines that:
Only process input data without output
Think and act like humans
Avoid any human-like tasks
Refuse to interact with external environments
Which perspective of AI focuses on creating models that think like humans through techniques such as introspection and psychological experiments?
Acting rationally
Thinking rationally
Thinking like humans
Acting like humans
Acting like humans in AI includes enabling machines to:
Understand and generate natural language
Restrict all reasoning to mathematical logic
Only perform brute-force computations
Ignore knowledge representation
Which area laid foundational theories for AI by considering human intelligence as mechanical processes?
Philosophy
Control Theory
Linguistics
Mathematics
In AI’s historical context, Search is considered:
An optional technique hardly used in AI
A fundamental technique used to explore possible solutions
A method unrelated to problem-solving
A minor aspect only for trivial tasks
Uninformed Search differs from Informed Search in that it:
Uses domain-specific heuristics
Has no knowledge of how close a state is to a goal
Guarantees the best solution in zero time
Is always more efficient than informed search
In the context of AI applications, Machine Learning involves:
Human operators always supervising computations
Systems that adapt from data and experience
Ignoring past experience in making decisions
Only dealing with simple arithmetic tasks
A Rational Agent in AI is one that:
Acts to maximize expected utility based on its knowledge
Ignores perceptions and acts randomly
Copies actions from any human regardless of rationality
Focuses solely on symbolic logic without action
Natural Language Processing (NLP) in AI deals primarily with:
Designing robot hardware actuators
Interpreting and generating human language
Building only theorem provers
Ignoring any semantics in conversation
Search in AI problem-solving is:
A random guessing method
A systematic way to find sequences of actions that lead to a goal state
Never used in practical applications
Limited to binary decision trees
The Goal Test in a search problem:
Determines the actions available in a state
Decides when a state satisfies the desired objective
Is irrelevant for problem-solving agents
Always returns false
A Search Tree:
Represents states and actions starting from the initial state
Is used only for supervised learning tasks
Has no root or leaf nodes
Cannot be pruned or optimized
Breadth-First Search (BFS) expands:
The deepest nodes first
The shallowest unexpanded nodes first
Only goal states
Nodes randomly
Depth-First Search (DFS):
Is guaranteed to find the shortest solution path in all cases
Uses less memory but may fail to find a solution in infinite state spaces
Always returns the optimal solution
Never explores deep nodes
Iterative Deepening Search:
Combines the completeness of BFS and the memory efficiency of DFS
Is just another name for BFS
Is never used in practice
Ignores solution depth entirely
An Informed Search strategy:
Is also known as a blind search strategy
Utilizes problem-specific knowledge (heuristics) to guide the search
Never uses heuristics
Will always run in constant time
The function f(n) = g(n) + h(n) is associated with:
Breadth-first search
Depth-first search
A* search algorithm
Hill-climbing search
A Heuristic is:
A guaranteed optimal solution finder
An estimate of the cost to reach the goal from a given state
Irrelevant to informed search methods
A random guess that cannot guide search
An Admissible Heuristic:
Overestimates the cost to reach the goal
Never overestimates the true cost to reach the goal
Must ignore all domain knowledge
Prevents the A* algorithm from functioning correctly
Knowledge Representation (KR) in AI involves:
Using arbitrary, uninterpretable codes for data
Encoding information about the world in a symbolic form
Avoiding the use of symbols or logic
Randomly guessing facts
Reasoning in AI:
Manipulates symbols to infer new truths from known facts
Always proceeds without any rules or structure
Is unnecessary if we have raw data
Means guessing without any logical foundation
Propositional Logic (PL) consists of:
Quantifiers and predicates
Truth-functional connectives and atomic propositions
Only existential quantifiers
No concept of True or False
Which of the following is a logical connective used in Propositional Logic?
$\lor$ (or)
$\forall$ (for all)
$\exists$ (there exists)
$\lambda$ (lambda expression)
First-Order Logic (FOL) extends PL by adding:
Only one truth value
Quantifiers and predicates to express relationships between objects
Fewer expressive capabilities
Direct integration with supervised learning
Converting natural language to logic often involves:
Directly copying sentences without interpretation
Using formal syntax, predicates, and quantifiers to capture meaning
Eliminating all structural elements of language
Ignoring universal quantification
Axioms in a knowledge base are:
Statements proven from other statements
Foundational truths or rules assumed without proof
Always false statements
Never used in reasoning
The relationship between universal and existential quantifiers includes:
$\forall x \neg P(x) \leftrightarrow \neg \exists x P(x)$
$P \leftrightarrow Q$ is the only important equivalence
Quantifiers do not have logical relationships
Existential quantifiers are identical to universal quantifiers
Entailment in logic means:
One sentence is syntactically rewritten as another
One set of sentences logically implies another sentence
There is no relationship between sentences
Random guessing of truth values
Higher-Order Logic (HOL) allows:
Quantifying over predicates and functions, increasing expressiveness
Using fewer connectives than PL
No increase in expressive power over PL
Ignoring quantification entirely
Uncertainty in AI refers to:
Having complete certainty about the state of the world
The lack of exact knowledge about outcomes or states
Never using probabilistic reasoning
Eliminating all unpredictability
A Bayesian Network is:
A cyclic graph of variables and probabilities
A Directed Acyclic Graph (DAG) with nodes as random variables
An unconnected set of facts without probabilities
Irrelevant to handling uncertainty
Conditional Probability Tables (CPTs) in Bayesian networks:
Show how child variables depend on their parent variables
Are not required to define a Bayesian network
Are always identical for all nodes
Represent only deterministic rules, not probabilities
Bayes’ Rule is:
$P(A|B) = \frac{P(B|A)P(A)}{P(B)}$
$P(A) = P(B)$ for all events
Limited to deterministic logic
Irrelevant to reasoning under uncertainty
Conditional Independence in Bayesian networks:
Is never assumed
Reduces the complexity of the joint distribution
Has no effect on inference complexity
Eliminates the need for probabilities
Inference in Bayesian Networks aims to:
Determine deterministic rules only
Compute posterior probabilities given evidence
Avoid using observed evidence
Ignore variable dependencies
A Converging connection in a Bayesian network occurs when:
Two variables directly influence a single common effect
One variable influences two different variables
Variables are arranged linearly without convergence
No common effect is present
Exact Inference in large Bayesian networks:
Is always computationally trivial
Can become extremely expensive computationally
Is unnecessary if we have CPTs
Requires no algorithms
Approximate Inference methods like Monte Carlo sampling:
Are used when exact inference is too costly
Guarantee a perfect answer every time
Never consider the given evidence
Cannot be used if networks are large
Gibbs Sampling is:
A method to do deterministic inference in Bayesian networks
A Markov Chain Monte Carlo (MCMC) technique for approximate inference
Never used in uncertainty reasoning
A variant of breadth-first search
