Chris Miller

So AI is one hell of a huge course. There was lots to learn, lots that could come up and lots that I didn’t know! The start and end of the course were my strongest points, the middle part was perhaps my worst section of the course.

Question 1

• A – Definition of Turing Test in AI [4]
• B – Advantages and disadvantages of learning capabilities in intelligent agents [3]
• C – Define problem generator give trade-offs [3]
• D – Define memes and how they play a part in agent design [4]
• E – Compare behavior-based, classical and hybrid approaches to robot design [5]
• F – Issues related to evaluation functions in search-based game players [3]
• G – Alpha-beta pruning, effects on computational cost and quality of results [3]

Question 2

• A – Defining precepts and actions [3]
• B – Three novel features to be included in mobile system [6]
• C – Define utility function and give refinations for the features in part b [4]
• D – Fundamental issues in perception, examples with features in part b. Sensor fusion. [6]
• E – Role of emotions in design of agents, relevance to mobile agent [6]

Question 3

• A – 3 reasons for not using deterministic logic in medical diagnosis [3]
• B – Definition of belief networks. Benefits in terms of computational, knowledge elicitation and interpretability [6]
• C – Define maximum expected utility, relationship with AI and human decision making [4]
• D – Explaining away in belief networks [3]
• E – Probability computations [5]
• F – Same as part E, using natural frequencies [2]
• G – Draw and explain risk-averse utility curve for a lottery [3]
• H – Equation to support decision process [4]

The exam was not as bad as could have been expected. I did manage to draw some random graph for the risk-averse utility curve, with axes labeled risk and averse respectivley. Hope I managed to scrape a decent grade.
- Chris