Dummy AI4 exam Answer all questions in 1hr 45 minutes. Total 80 marks.

  1. Give 4 categories of definitions of Artificial Intelligence, with an explanation of each which lists the most important motivation and disadvantage of each. [12]
  2. What is ‘rational’ in any given rational agent depends on 4 things. What are they?[2]
  3. An ‘intelligent anaesthetist’ agent is being designed to assist during complex surgical operations. The system will be informed of a patient’s case history from case notes in advance of an operation, and will have access to sensor readings during the operation. Minor changes in drug and oxygen administration can be made by the agent. It can present the medical staff with information about the patient’s state and warn of impending problems It can explain the reasoning behind its actions in natural language, if staff do not understand why it is behaving the way it is.
  1. Describe your interpretation of the agent’s percepts, actions and goals. [3]
  2. Describe the properties of the agent’s environment.[5]
  3. What would the advantages and disadvantages of introducing adaptive behaviour to the agent be?[4]
  4. You wish to build an ‘experiment assistant’ into the agent which will advise the medical staff which test to perform next, given an uncertain hypothesis about a given patient (current state of knowledge E). What would the expected value of perfect information (evidence Ej) about a given test (from N possible tests) be? (Provide an equation which would support this decision process) [6]
  5. Can you forsee any legal, ethical or practical issues which would have to be analysed more carefully before using such a system on human patients?[4]

[36]

  1. Describe the behaviour of depth-first, breadth-first and uniform search, and compare their relative advantages and disadvantages [9]
  2. How does ‘informed search’ try to improve on ‘uninformed search’?[3]
  3. Explain how A* search works, what its advantages are, and identify any restrictions on heuristics which can be used with it. [6]
  4. Write a pseudocode implementation of the MINIMAX algorithm [6]
  5. What is the maximum level of ply would be feasible for MINIMAX on a computer which could process 1000 evaluations per second in a game which had a maximum branching factor of 40, and where moves had to be completed within 2 minutes? [3]

[27]

3.

  1. Why do we sometimes need mechanisms for reasoning under uncertainty? [3]
  2. A nuclear power plant has a safety system which measures the radioactivity of a water outflow. If radioactive levels above a certain threshold are detected the sensor returns a warning signal. If the sensor reliability is 95%, and the probability of a leak is 1 in 1000, what is the probability of a leak having taken place? [5]
  3. Draw a risk-averse utility curve for a lottery. [2]
  4. What is the principle of maximum expected utility?[2]
  5. Describe the explaining away effect in a belief network (you can use a diagram to help make the point clear), and give an example of its possible use [3].
  6. In the network below, if A is known, does inserting hard evidence for B alter the probability of C? Explain. [2]


[17]