# Turing Test and Robot College Student Test

CE213-5-SP/AU/ZU 2
Question 1
(a) Based on your understanding of the two AI tests: Turing Test and Robot College
Student Test, with no more than 100 words propose an AI test using a Metaverse
scenario.
(b) The A* search strategy uses both the cost of reaching the current state from the initial
state and an estimate of the cost of reaching the goal state from the current state to
select nodes for expansion. It estimates the cost of reaching the goal state from the
current state using a heuristic.
(i) If the heuristic is faulty and always returns a value of 1000, what search strategy
is this A* search equivalent to?
(ii) With such a faulty heuristic, is this A* search optimal and complete?
production systems using the Mycin’s certainty factor system:
(i) If condition 1 is satisfied with certainty 0.9 and condition 2 is satisfied with
certainty 0.7, what is the certainty with which condition 1 AND condition 2 is
satisfied?
(ii) If a conclusion can be drawn from Rule 1 with certainty 0.8 and the same
conclusion can also be drawn from Rule 2 with certainty 0.5, what is the certainty
with which this conclusion can be drawn by using both Rule 1 and Rule 2?
(d) From the perspective of production rule arrangement and execution, compare a reactive
agent with subsumption architecture and an expert system that are designed for mobile
robot navigation control. Point out at least two major differences.
(e) Choose up to two appropriate machine learning methods for solving each of the
following problems through machine learning and briefly justify your choices:
(i) Given a dataset that contains a large number of people’s facial images labelled by
corresponding people’s identifications, you are asked to develop a face
recognition system.
(ii) Given a map describing an indoor environment with a large number of positions,
including initial and goal positions, rewards for reaching some positions, and a set
of actions that a mobile robot can take, you are asked to develop a controller for
the mobile robot to navigate in the indoor environment.
[8%]
[8%]
[8%]
[8%]
[8%]
3 CE213-5-SP/AU/ZU
Question 2
(a) The following figure shows the current game position or state of a Noughts and Crosses
(Tic-Tac-Toe) board game played by two players: Player X and Player O.
Player X will make the next move to put a cross on one of the three empty grids.
(i) Construct a game tree with the above current game state as the root node and all
the leaf nodes representing possible endgame states.
(ii) Assume that the value to Player X of an endgame state is 1 if Player X wins, -1 if
Player X loses, and 0 if it is a draw. Use minimax search strategy with AlphaBeta pruning to find the value to Player X of the current game state. You can
assume the maximum value for any game state is 1. Show your working.
(b) In the following state transition diagram, the number alongside each arrow indicates
the reward associated with the corresponding state transition, and state A is the goal
state. If the discount factor is 0.9, calculate the maximum discounted cumulative reward
values of state D and state E for reinforcement learning, respectively.
O
O O
X X
X
[8%]
[12%]
[10%]
E D
F
0
0
0
0
C B
100 50
A
0
0
0
CE213-5-SP/AU/ZU 4
Question 3
(a) This question is about knowledge discovery from data by machine learning. The data
samples given in the table below represent certain relationships between the Diagnosis
result and the measurements of Blood Pressure, Pulse, and Body Temperature. Assume
these data samples are very informative and among the three measurements Body
Temperature is least relevant to Diagnosis.
Blood Pressure Pulse Body
Temperature
Diagnosis
High High 38 A
Normal High 38 B
Normal Normal 40 C
High Normal 35 A
Normal Normal 35 C
Normal High 35 B
High Normal 40 A
High High 35 A
(i) Find whether Blood Pressure or Pulse has the largest information gain about the
Diagnosis result.
(ii) Use information gain based decision tree induction method to get production rules
(IF_THEN rules) as the knowledge that can be discovered or extracted from these
data samples.
(b) A McCulloch-Pitts neural network of 3 MP neurons is represented by the following
diagram and the values of neurons’ thresholds and connection weights are also
indicated in the diagram. Calculate its output Y when its inputs are S1 = 0.5 and S2 =