Question 11 pts
Empiricism is
the idea that (sense) data is the ultimate source of all knowledge and intelligence
a theory that rules out innate (e.g. genetic) knowledge
the view that empirical sciences (like physics) are superior to conceptual sciences (like mathematics)
a method for evaluating AI systems by empirically testing their performance
Flag this Question
Question 21 pts
A computational problem is called intractable iff
Group of answer choices
the time required to solve it grows polynomially with the size of the problem instances it cannot be solved by machine learning
the time required to solve it grows exponentially with the size of the problem instances it cannot be solved by an AI system
Flag this Question
Question 31 pts
is best thought of as the study of Group of answer choices companies and business preferences, desires, and utility multi-agent interactions
making money
Flag this Question
Question 41 pts
Which of the following appears to be the best approach to developing medical applications in AI, given the current state of the art?
Group of answer choices
An intelligence augmentation approach where the system and a doctor work together Go with human judgement because domain knowledge means they perform better than machines
focus on diagnosis problems based on images where computers are especially strong Eliminate human bias by using a complete automated solution
1. CONTROL THEORY is
how artifical systems can operate under their own control
how the brain controls the body
how we can best control AI systems
2. Mathematics allows us to rigorously formalize many crucial AI concepts. Which of the following has not been formalized by mathematics?
probability
dualism
logic computation
3. Cognitive psychology, which views COGNITIVE the brain as an information-processing device
4. In the standard model of an intelligent system, the system’s objective is defined by its designer. This raises the problem of value alignment; choose the best definition of the value alignment problem from the list below.
how to align the designer’s values with social values like justice and freedom from bias that human preferences evolve and are hard to quantify
how to align the machine’s values with social values like justice and freedom from bias
how to make sure the value given to the machine match the true value of the human designer
AI design evaluates an agent according to the (expected) outcome of their actions. In the terminology of philosophical ethics, this approach is an example of
Group of answer choices
deontology
rights-based ethics utilitarianism 1consequentialism.
Flag this Question
Question 21 pts
In a multi-agent environment, other entities Group of answer choices
can cooperate with the AI agent
must be viewed as agents
1may be viewed as agents
can interfere with the AI agent
Flag this Question
Question 31 pts
Implementing the agent function with a lookup table is doomed to failure when the set of percepts is even moderately large because
Group of answer choices
the table is too large to be built manually by the designer
there are too many entries to learn their values from experience
1all of the other answers
the table is too large to be stored even in external memory
Flag this Question
Question 41 pts
Which of the following is true of a transition model?
Group of answer choices
1both
it represents how the world changes independently of the agent neither
it represents how the world changes in response to the agent’s actions
Flag this Question
Question 51 pts
A rational agent must know its utility function. True or false? Group of answer choices
1True
False
Every finite game (with finitely many strategies for each player) has a Nash equilibrium (allowing mixed strategies)
True
False
An outcome in a game is a possible payoff pair for both players. (E.g. (0,0) is an outcome in the coordination game.) An outcome is Pareto-optimal if
there is some other outcome preferred by some player
there is no other outcome preferred by all players
there is some other outcome preferred by all players
In a single-move game, a strategy profile
specifies the basic properties of a strategy
assigns to each player a single mixed strategy
assigns to each player a single strategy, which may be pure or mixed assigns to each player a single pure strategy
For any options/strategies s,s¡¯: If s strongly dominates s’, then s weakly dominates s¡¯. In other words, strong dominance implies weak dominance.
7
True
False 22%
Question 11 pts
Suppose that we have an objective function f(x1, x2, x3) with three arguments x1, x2, x3. Then the gradient of f at a point x1, x2, x3
Group of answer choices
is a vector that points in the direction where the objective function decreases
is a vector of three numbers
is one number
Flag this Question
Question 21 pts
Hill-Climbing is
Group of answer choices
like climbing Mount Everest in thick fog with amnesia a favourite way for AI engineers to relax
a kind of tree search
a standard benchmark challenge for AI robots
Flag this Question
Question 31 pts
Which of the following is true of random restart search methods for maximizing a function?
Group of answer choices
Uses random sideways moves to escape from a plateau
Never reaches a local maximum
Helps to find better local maxima
chooses at random from among the uphill moves
Flag this Question
Question 41 pts
Choose all that apply. The Newton-Raphson method max the function
Group of answer choices
avoids the problems of local maxima, ridges, and plateaux
divides the function derivative by its second derivative (for single input variables) finds the root of a function(method itself)
can be adapted to find a function extremum (maximum or minimum)
all of the other answers
find the root of a derivative(max)
In a search problem, if a search tree node is in the frontier, this means that the node
will be expanded in the next step
was generated by the algorithm but not yet expanded
was expanded in the previous step 8 respondents o* & Attempts: 126 out of 126
In a search problem, the term ¡°heuristic¡± refers to
a search strategy that is only sometimes optimal
a rule of thumb
a function that assigns values to states
Very large scale integration for building circuits can be solved by modelling it as a search problem.
True
False
Given the assumptions for single-agent search described in Ch.3.1, the solution to a search problem is,
A branching strategy for dealing with contingencies
a sequence of actions to be executed by the program
2Nash equilibrium
Compared to the minimax algorithm, the alpha-beta algorithm returns for each node
the same evaluation 82 respondents
higher evaluations 7 respondents
sometimes higher, sometimes lower 17 respondents lower evaluations 9 respondents
Attempts: 115 out of 115
To solve sequential games, Al systems build out the game tree 33 respondents
use the game tree as a theoretical construct
The minimax algorithm follows which of the following search strategies?
Depth-First 112 respondents
Breadth-First respondent a 2 respondents
Attempts: 114 out of 115
In a two-player sequential game, the MIN player
always moves second 103 respondents
always moves first 3 respondents sometimes moves first, sometimes
A probability density function
shows the values with the highest
defines probabilities for a continuous random variable
79 respondents
is a uniform probability assignment for 33 respondents
a continuous random variable
The inclusion-exclusion principle is a formula for the following: The probability of a disjunction of
15 respondents
mutually exclusive events
the joint probability of any list of events 16 respondents
The probability of any disjunction 83 respondents
Attempts: 114 out of 114
XK K
Let XY,Z be three random variables. If X and Y are conditionally independent given Z, then P(X,Y|Z) = P(X|Z) x P(Y|Z).
True 103 respondents
False 11 respondents
The domain of a random variable is the range [0,1]
the part of the environment modelled by the variable
the set of entities referred to by the variable
the set of possible values of the variable
: 102 respondents
Ly
90% 10%
5% 89%
For any joint probability distribution, there is a Bayesian network that represents it using the product formula.
True 100 respondents 93%
False 8 respondents 7 &
Attempts: 108 out of 108
Suppose that node B is not a descendant of node A in a Bayesian network. Then A is independent of B given the parents of A.
True a0 respondents 74%